Both AMD and I expect CoolerMaster, need a good slap. It's kinda baffling how they thought to get away with this for the second time and no one in a multi million dollar R&D department noticed this. Reference 6900XT showed the same behaviour but only when it was mounted vertically, where the vast majority would have their card mounted horozontally so the issue didn't seem so big.
Ah, GN had explained the phenomenon of "dry out" at some point where if a heat pipe is overwhelmed it just poops its pants and the liquid never condenses back leading to poor thermal conductivity
Id say its more a matter of resource management. Im sure the engineers behind these products are excellent. They're just working with their hands tied behind their backs by the bean counters.
They are too busy making exaggerated performance charts to pretend they are so power efficient. Meanwhile, my 4090 sips power while annihilating everything AMD has ever made.
@@mattgreenfield8038 Your 4090 must be the only one that "sips power" because they didnt put those fancy high power connectors on them for a joke. Nvidia does good power design typically and these cards are built for a lot of power and have rather high power draw requirements. Jensen would likely not even exaggerate this much. The 4090 is a powerful beast that needs to be well fed.
@Boanerges57 if you had a 4090, you would know they never hit over 400 watts even when you max the power limit to 600 watts, unless you are running something extremely demanding at 4k. Look at the difference in quality between the 2 reference brands. I got a Gigabyte Gaming OC 4090, but the 4090FE is so much more well built than these AMD reference cards. Another domination by Nvidia. Why would anyone buy a 7900XTX instead of a 4090. Makes no sense. If you can't afford it, just get a 4080. This 7900XTX is worse than Vega.
@Okay Yeah damage control? How much damage control can they do for this disaster, which is this RDNA3 generation. This is now worse than Vega. I'm so happy I didn't wait for these dumpster fire 7900 series and bought my 4090 on day 1. No problems. No undervolting. No black magic needed to get the 4090 to demolish these cheap RDNA3 cards. What a disaster for AMD. They deserve it for mocking the 12vhpwr that was caused by accidental user error. Look at the build quality difference between a 4080FE or a 4090FE... compared to these AMD reference cards. OMG, lol 😆
Its insane to believe that AMD behind closed doors doesn't already know about this issue . They are scrambling and panicking to figure out a solution before they say anything
I don't follow Reddit or Twitter just UA-cam for this info. None of the other channels are talking about this detail because of the holiday season, but thank you for your in depth analysis and news spotlight on this topic. This brings me more understanding of what JayTwoCents was possibly seeing with the down clocking under load. I was hopeful for team red this year, but looks like I'm sitting on the sidelines till one of these GPU manufacturers gets their act together.
It's a shame this problem is pretty much irreversible as is. Great diagnosis concluding with the vapor chamber. AMD is going to have to go back to square one. I always wait a certain amount of time before purchasing new tech, as with past launches it saves grief.
@@tigriukasinlove Don't be dumb, you should never just buy into a monopoly. Did AMD fuck this up this time? They sure as hell did. However, if you want Nvidia to continue raising prices to the moon and Intel to go back to its 14n++++++, then by all means be a corporate shill and believe they have your best interest at heart.
I love how ONE person is doing a better job than a full QA department. Or they reported it and management decided to ignore it. Huge kudos to you for exposing those scams.
Yea....QA is like TY for figuring that out for us. He should be working for one of these companies, but I don't blame him for doing his own thing...seems he's doing OK. It's hard to believe that they would ignore that, cost more money in long run.
@@LiveType Hmmmm, makes you think why aib cards custom models actually had that delay launch, huh? Amd clearly knew this was a thing. Shame on them. At least im glad that first with cpus and now with gpus we get to see the true face of that company. Not that i ever had any doubts, but yeah.
1) very limited amount of cards at launch 2) instant rumors about new revision on the way => means that AMD knew there is a problem with vapor chamber for a long time, maybe before launch That really tells a lot about character of people in AMD management. I mean everybody makes mistakes, the difference is how you face them.
10:00 when the acrylic stand didn't work, my thought is that the liquid in the vapor chamber is unable to wick up to the underside where the GPU chip is in contact. My hypothesis is that if it's mounted horizontally, but upside down, the temperatures would be normal. AMD: It's normal. Just turn your PC upside down. That'll fix it.
@@zihechen3111 AMD might send a lawsuit to cease and desist. Its the same issue like nVidia, there is no problem but there are faulty cards out there all the time. Asking for people to send their broken cards in will result in... broken results. Also there might be plenty of other reasons why the hotspot temp is that much higher.. and even if it is higher maybe it is totally fine and it is meant to go that high. Publicly quickly rushing some hot takes video out for clicks is not the way to go imo, just ask AMD about it first atleast.
yeah not that much of a point now . this is 100 percent the problem and not a theory at this point. And it almost sounds like "working coolers" are gonna be rare finds.
Working coolers may also stop working as was mentioned in the video. Worked in one position, was changed to another orientation and it stopped working, but when changed back to the original orientation it still didn't work correctly. Seems like a very poor design where the liquid can be moved to a part of the card that breaks the phase change loop.
@@ComputerGeeks-R-Us right, but not permanently broken right? I think once the whole thing cools the phase change loop works when vertical on cold start again? I already see some hints in reddit that ppl are returning XTX cards that are ok with hotspot in fear that they will develop this issue in the future ... this is very damaging to RDNA3 ... this also shows (to nvidia as well) why AIBs have value and should be able to make decent profits
Bought a Power Color Red Devil recently, ran fine on the provided cooler (which is not reference design as far as I can tell). I then slapped on Alpha Cool's Core cooler and got horrible temps, 105-110c hotspot. Noticed a lot of people on the AC forums having the same issue. I tried remounting, repasting with 3 different pastes. Nothing worked. Then i noticed the card was boostin to 3Ghz and 600W without me even touching the overclock settings.... ! I can only assume the card detected a much, much better cooler and then boosted like mad. I set the clock back to stock clocks manually and it's been running at ~55c core and ~80c hotspot ever since.
they don't test every single card. and if its a low percentage it would be easy to test a bunch of them and miss the few that have this issue. it also seems like they don't 100% consider it an issue as they aren't allowing RMA for this issue. (yet)
You're nuts if you think they did not know this. as i said elsewhere amd released a half baked product for sale to the public. they were hoping that it would go unoticed just like their weird clocks speeds all over the place and their low fps from higher clock speeds. amd entire staffing are shills even the ones that knew releasing the product was wrong but did nothing but sit back and allow it to happen. HInt Hint Lesie Zue LOL.. now they will go into damage control mode and say it's only a few batchs here and there.
Your due diligence is impressive and appreciated Roman! Cant wait for the Vapor Chamber cut aways and maybe some watercooled testing... that would seal this diagnosis for sure and put AMD in no position to ignore this.
yes, I am full of admiration, but in my opinion he could also check how the card behaves when it is placed on its back with the cooling up, it can tell us what exactly is the problem with this chamber. it would be good if the card was tested in all positions, it would give a lot of valuable information
@@konbiex4684 Yeah, but those would be unrealistic mounting options since average user will likely mount one of the two positions he tested. That aside, for the science of it, it would be another way of testing this vapor chamber theory. Yes.
You can't really measure the pressure in the chamber by cutting it open, unless you maybe do it under vacuum. Measuring the amount of liquid in there tends to be difficult as well, as the water tends to be just stuck in the chamber wall coating and doesn't want to come out. You have to boil it off and re-condense it and that's experimentally pretty shaky to do it cleanly not losing or gaining any water, considering air has a fair amount of moisture in it. So only if there was an issue with the wall coating you're gonna likely see it when you cut it open, but not the two more usual and expected heatpipe or vapour chamber faults.
@@SianaGearz indeed, but now after first tear down he knows where the fill port is before its sealed and covered. It would take quiet a bit of prep, but if he thinks it is needed, I believe Roman could create a proper test to see what the volume and pressure are. Would be another great vid!
I can’t quite get over the fact that no one picked this up from any of the manufacturers. Was there any kind of quality assurance testing?? Surely when you make a card you don’t just test it on a test bench. You test it in a normal case that any home user would use?! Mind boggling that this flaw actually went out to market!!!
I had a similar problem many years ago. My AMD R9 390 was mounted vertically (with the IO facing upwards) in my Silverstone FT02 case and it ran very hot and never came close to advertised clocks. While troubleshooting I turned the case on its side and ran a benchmark. It ran much cooler and at its advertised speeds. I think the heatpipes or vapour chamber did not work very well when mounted like that. Ended up having to change cases to something with a normal orientation and the problem went away.
Would have been interesting to see the "bad" vaporchamber swapped onto a known good card. Great investigative work. I was guessing the whole way though.
He should have contacted AMD first and ask about this phenomenon. Making some hot takes clickbait videos is not the way to go imo. Also if the card is working fine otherwise I dont even see a problem here.. for years hardware has been made to work at 100c, it just sucks if its a laptop and it burns your jewels.
And here we are 6 months later from this video and the first I've heard of the problem is from me researching which GPU to buy. So it seems literally nothing is being done about it.
This is huge. Shame gamer nexus has taken Xmas off. Maybe HUB will catch on to this story. Well done Der8auer 4x 1000$ and you already had a XTX so 5K for one story. Amazing dedication.
@@oinodin This right there. Those Cards are 100% fine still. Not everyone can use them but for ppl who would buy Waterblocks anyways this would be basically a none Issue. Just sell them for like... i dont know... 150 Euro less if the Cooler is broken permanantly and GG. It just needs to be clear that the Card is working fine and thats that. If Roman buys Waterblocks and mounts them then he could probably sell them for more even. I dont know about the Card which he slightly modifyed but even that one should be fine if he tells the Person what he did and he sells it slightly cheaper.
Always blown away by your hard work, dedication and excellence in coming up with videos like this. If I was AMD or NVIDIA I'd keep you close and include you in a private pool of testers before openly releasing to the wider global market. Your skill set along with your passion for cutting edge hardware makes you more valuable than an employee with the same skill set but with NO heart. It is unimaginable to think of the potential loss of AMD if this was truly a design flaw. What a headache to start off 2023. Thank you Roman and happy new year. Keep 'em coming.
Dear Der8auer -- Happy New Year! -- Quick question what's the part you use that acts as both a PCIe 180 degree breakout board that also lets you read the power draw from the PCIe power cables? Cheers and congrats on finding the root cause!
I seriously want the MBA cards to be all faulty to the point it forces AMD into an embarrassing recall, would serve AMD right for the horrendous marketing around the whole RDNA3 launch. That smugness of Scott Herkelman should cost him his reputation within RTG. The amount of trash talking and posturing was ridiculous. Time for RTG marketing to be entirely gutted.
I just bought a AMD built 6950 XT and it was running up to 110c while under load and would never come back down until I let it idle. I then found your video through JayzTwoCents, and so I tried laying my case on it's side to get the GPU vertical, and same load it never went above 92c, and hovered more around 87c if I opened the glass panel that was now the top. This issue could be bigger than just on the 7900's or I was just extremely unlucky with my GPU. At least I now know I can run the card safely in the vertical position but what a pain in the a$$...
I also bought the AMD 6950 XT (reference card) and had the same problem. Because of this video and your comment i tried laying my case on the side and this helped a lot. Hotspot temp going down from 110c to 92c (under load). Also less of a difference between GPU edge temp en hotspot temp. Delta is now 20c difference instead of 40c (GPU temp 70ish hotstop now 90ish instead of 110c). Thanks for the tip!!
@@jorisgrootens5028 Glad it helped, still kind of a huge annoyance. Since I like my tower on my desk and laying it down took up too much room and didn't look great anyway, I ended up moving everything into a Hyte Y40 since the case included the vertical GPU mount and riser cable. Smooth sailing since.
@@slvrlancer You guys should RMA. Mine was staying below 100c hotspot, delta between hotspot and edge around 20-23c, under heavy load it oscillated at 77c edge and 97c hotspot with slight variations.
AMD had such a good chance to win people over from nvidia this gen and honestly was with a lot of people i spoke to who were interested in upgrading. They really need to act very fast on this or else its all for nothing.
Its over, you get one chance and they blew it. They blew it with overpricing lackluster cards and settling to be a parasite living off Nvidias Monopoly rather than actually compete, and then if that wasnt bad enough it turned out not only were the cards horrible performance for the asking price but they were also objectively bad tech by being inherently defective. Between RDNA1 architecture/driver issues, Vega…. And now RDNA3, its time to come out and admit that atleast the Radeon portion of AMD is just straight up bad and getting an amd card is a coinflip on of its gonna be good or bad. They had a chance to shake off the issues of the past and actually let people know radeon being unreliable was a just a stigma and instead they reinforced it after talking shit about Nvidias flaws only to be even worse. Thats wild.
Random German? Germans are masters of engineering. This is systematic, analitic testing of product. This is science, wery well done. AMD screwd up with this one.
@@zorandamljanovic6171 correction, the company that designed, tested, and produced the vapor chambers for AMD screwed up. Your comment is like blaming AMD if something was wrong with the chips they bought from TSMC
@@mjkittredge True, but it does say Radeon on those carda. Not all of them wore Radeon, but at least one was. So... AMD shoud test there stuff before they release them. Same goes for nVidia.
Love watching your content! Not only are these deep dives incredibly valuable to the PC community, they are also super interesting from a science and engineering perspective. Cheers!
Happy New Year Roman and thank you not only for this video which is an amazing investigation into what is causing these hot spot issues but all hard over this last year. Keep up the great work!
This is so spot on... I bought an XT and XTX and after tons of testing deemed the XTX suitable. The whole time I had my case laying down with the GPU vertical. I just finished my build and stood my PC up and now I'm hitting 108 junction temps from my 95 max yesterday. A delta of 40c (sorry not using kelvin haha.) total. I may have to return the card at this point... Thanks for the video I'm glad we have a solution it seems!
I really enjoy your whole approach here. Rather than just deciding what the problem is without sufficient testing (Nvidia power connector videos!) you’ve identified the problem with methodical testing and engineering, presenting facts and very readable graphs and data. Excellent work 😊
@@Pisscan lol maybe? I actually like J2C. He makes no apologies that he is just a guy with an opinion, but he did go a bit crazy on the whole Nvidia adaptor situation 😅
Amazing investigative work as usual. Thanks for focusing on facts rather than being an alarmist like a lot of youtubers who simply broadcast misinformation.
I need some help, can you tell the measurement of the screws you use for increase the pressure mout of the card? I wanna try that solution for my RX6950XT, because I think the center of the car are not doing a well contact
Happy New year. This was really great research into the cause of the hot spots... Thank you for the video. Re: Kelvin... I find it odd that the DIN standard requires reporting differences in Kelvin, but in reality since one kelvin is the same magnitude as one degree on the Celsius scale, it doesn't really change anything. So if something has a 10 Kelvin difference, it would be the same magnitude (10) in Celsius degrees.
It's a clarity difference. By reporting delta in Kelvin, there is no risk that anyone reading mistakes the delta value for the actual value at the measurement point.
@@ElijsDima Ah... very interesting. But, after some thinking on it, I don't see the real necessity. We use delta values for hundreds of different units in science all the time, and we don't create unique units for deltas. So, for example the delta for 20kg and 10kg has kg for units. We don't create a unit called dkg to make sure people know the difference. Thank you for pointing out the reason it is used. (It still seems very odd to me). I wonder what would happen if by conventions we started adding d to the units for deltas? dkg, dK, dsecs heheh
@@marcfruchtman9473 Kelvin isn't a uniquely created unit for this single purpose. Kelvin is the standard. There may be some rationale behind the fact that the delta is a calculated/derived figure rather than a sampled figure. If the measurement apparatus uses a particular unit, then that's what you report, and after that you do the unit conversions for any further calculations. So if the software report uses C then you record in C, then do conversion to K.
I don't know why people would complain about using Kelvin, it's directly related to Celsius. 0 Celsius (freezing point of water) is 273 Kelvin. 100C (boiling point of water) is 373K. If you did the delta in Celsius then a delta of 0C would be 273K and that just doesn't make sense.
Very informative....one question, Didi you try swapping coolers to a different AMD vap chamber. This might be the way to find out if the wrong internal pressures are varied in the vapor chamber
Swapping coolers (from good working card) should have been second thing to do after checking thermal paste and mounting pressure!!! Obviously, he did not think of that.
Yes, I was waiting for this the whole video, this would have restricted the problem to the cooler straight from the start. Though the video would have been shorter.. 😅
@@ElvirBegovic If the card is cooled down properly in one position and overheats in the other position then it's rather obvious that it's a cooler problem because silicon can't behave this way. Replacing a bad cooler with a good cooler would fix the issue, we can be sure of it without swapping anything, so what's the point?
Thankyou Roman for your detailed look into this! I do think that there is a chance that this wont 'blow-up' the same way that the NVIDIA issue did , purely because people have given AMD a pass because they dislike NVIDIA so much, so lets hope to push this and get a response from AMD :)
AMD sadly always get given a pass, you only have to look at the public response to the 7900XT which thankfully has been called out by GN, HUB and the vast majority of reviewers, even those who have given AMD passes in the past.
@@NBWDOUGHBOY Nvidia pricing is criminal so it’s justified. AMD unfortunately are following Nvidia’s lead despite having inferior products. Corporate greed is out of control.
@@Squashed8Ball Yep. You are 1000% right. Both companies are wrong! They are both cutting corners and charging customers 2 times what a Flagship GPU used to cost. It sucks.
Great video, very informative. I purchased a reference 7900xt (Sapphire) and can't get my hot spot to go above 82 degrees, so I'm thinking it's working as intended on my end. However, I will continue to monitor temps to see if anything changes.
AIO liquid coolers flow direction is based on outlet on the pump. It doesnt matter if it goes left or right, the end result will still be the same. Vapour chambers are based on thermal dynamics. Liquid contained in the vapour chamber is basically boiled to the point of evaporation and together with the heat travels up the heatpipe to an area where it is cooler or actively cooled by a fan and that cooler area or fan helps the vapour condense back into a liquid form that trickles back to the base plate of the vapour chamber where it goes through the same process again. (This is all coming from the top of my head so I think i am mostly correct for the vapour chamber bit but there are other videos and websites that can explain the whole process in more detail)
The problem may be that once the vapor condenses in the colder part of the heatpipe, it's unable to go back to the hot part of heatpipe to start the new vaporization process. Due to the curvature of pipes, I can see that there may be some more favorable starting orientations. It's either that, or simply just bad chiplet design, as people reported the same issue even with a water block cooling solutions.
So I should mount it vertically? Or return it? My card hovers between 97c/100c sometimes 102c(Hotspot temp) when the fans are ramping up. The fans are going well over 2000 rpm most of the time. Thing sounds like a jet engine. I'm not happy :( The card itself is never over 60c so it's always 40 plus delta under use.
der8auer is such a great engineer, love all his engineering series. He should set-up a consultancy for QC and design enhancement for these manufacturers.
For those wondering about Kelvin and Celsius : In thermodynamics, when you're doing complicated maths, you don't want negatives values, and you was a little numbers as possible to simplify the equation and allow you to find the solution to the problem as easy as possible. This is exactly why Kelvin was invented, removing all potential for negative values from equations (at least all those that are related to temperatures), therefore all the thermodynamics equations are based on the Kelvin scale, no the Celsius scale (even though, as you all know, the diference of temperature between two Celsius degrees is the same as the difference between two Kelvin values). If you wanted to do your calculations on the Celsius scale, using all the proven thermodynamics equations, you'd have to add 273.15 degrees to all of your temperature values (25C is 273.15K + 25 = 298.15K). It's jarring and just plain not necessary. It's easier to do all the math using Kelvin and then convert it to Celsius once you got your result. For deltas though, it's barely matters if you say that you have a delta of 25C or 25K as those are the exact same temperature difference
Awesome work Roman 💪🥳🎊 and a great find indeed! Let’s be real every business or corporation stuffs up……. the thing is how they respond and rectify the matter which is important! We’ve seen this from Nvidia, Intel, AMD, Gigabyte, MSI, Asus ….. looking forward to what the next chapter of this saga holds in store 🍿🍿
Well. At least in some Cases, yes... it could explain some strange Numbers. But who knows really. Whishfull Thinking as well i think. Lets wait and see what happens.
I am a bit surprised this is not common knowledge within such a community that kelvin and celcius are interchangeable if using deltas. The only difference is the minimum temperature: kelvin has no negative temperature. It only has a negativ sign if you compare a higher temperature to a lower temperature so the delta of those to temperatures would be -x Kelvin.
You could replace the vapor chamber with one from a good card and measure the temperatures again :) if temps are normal, you are sure it is the vapor chamber
He doesn't have to. He has used the Sherlock Holmes method. when you have eliminated all the other factors, whatever remains, however improbable, must be the cause.
Why? This man does this for a living. There is no need to do that. The problem is identified. Don't give AMD cover. They messed up and they need to fix it ASAP.
From the start of the video I had the feeling it had to be that the vapor chamber was drying out. I was hoping you would test it horizontal in the opposite orientation, ie with fans pointing up instead of down. Maybe that would allow the liquid to reach the hotspot, and reestablish a proper flow. The capilary effect may need some help from gravity. (Also it would serve as a test for other potential issues, such as pockets of hot air not escaping by convection when the card was horizontal.)
But if only some cards experience this there has to be a variable in the manufacturing process? Could it be that the composition or amount of the fluid is wrong in some cards?
@@wombatillo Likely the lower pressure in the vapor chaber is not low enough, causing the boiling point to be too high, Then again why would orientating it differently cause it to work better? aybe the heatpipes/vapor chamber are leaking? Bad stuff here regardless of the problem. I was really hoping AMD would gain some traction in the gpu market as we all benefit from close competition.
7900 xt owner here no problems with cooler, I think your unaware the 7900xt is a smaller different model cooler. This seems limited to the 7900 xtx model.
Random comment here - when the donut of death appeared (furmark) my laptop fans actually span up a bit. Confused the hell out of me haha. Thank you for this reporting Roman, and for getting those gamers good cards. Very decent of you.
Fantastic work, agree 100%. I would only add that this should be a (very bad) production issue, not a design flaw. I have a direct from AMD 7900 XTX and mine is perfectly OK, I never saw hotspot temps above 90C and that's running stress tests and many benchmarks with the card overclocked including +12% PL. And I have it on the horizontal position. So I guess I got lucky, still, if AMD has to issue a blanket recall I would take it because even those of us with a good unit will lament it when it's time to resell the card as used, unless AMD steps up and does whatever it takes to fix this.
Could you try flipping the card fans-up? That would be to have the liquid go down with gravity again (if in a liquid state at all) and/or also kind of shake it between position change (by letting your mount for example fall on the table from small height cable permitting), to try to dislodge if it's stuck in the wick/surface tension? If any of that does anything, I don't know what it means but it'll be hilarious:)
It's unlikely that force will reset the function because it's a vacuum with mostly water vapor instead of liquid water. There's likely an angle that would cause it to work again but there's little to gain from finding it.
Thanks for great work you have done. Have you tried to swap the vapor chamber between Good and Bad cards? just to be sure that there are no other variables? (e.g. chip, vram etc)
Really good question, I'm going to mess around with my xtx tonight in horizontal vs vert. I haven't NOTICED an issue (the things a beast) but maybe its there.
Celsius and Kelvin values are the same when measuring differences (deltas), water has 100 degrees of difference between its melting and boiling points in both scales. The only difference is when measuring absolute temperatures, 0 degrees in Celsius is 273.15 degrees in Kelvin, 100 degrees in Celsius is 373.15 in Kelvin, still 100 degrees of difference between both temperatures in either scale.
oh this reminded me of an article i read about space ship design and how some heat pipes depend on gravity to function properly and some depend on zero-g. not that this is the case here since the solution isn't just orientation, but it was interesting to me that until then i just assumed a heat pipe was a simple device that was just relying on conduction and didn't know there was a liquid inside with phase change. i thought oh heat pipes are more interesting than i imagined 😀
Ice giant CPU hsf heat pipes only work in certain orientation, if used in other orientations other than the ones they say it can work in, it will not work. They explicitly say what positions it can be mounted in. Relying on gravity as well as heating and cooling. They can also stop function if too much heat is provided in those positions aka dryout. It is fairly interesting technology with limitations.
Could you possibly run your tests on the affected cards in your possesion with a water cooler in place of the vapor chamber design as an additional confirmation?
@@LegionGamingTV there are other issues as well. the ram speed issue where you can get the promised 3 ghz speeds from sensors but a 50% decrease in framerate. So all the stats look better on paper but there is a slow down somewhere else. Jayztwocents believes there is an issue with the memory manager as well. too many issues with this card but atleast till now no cards have self combusted but they sure arent gonna get record lifespans with these cards.
Curious if you are going to put on a water block to solidify the results of the testing. It would eliminate the vapor chamber as a factor and see what the actual true temps of the card would be under nominal cooling. Great process and video!
I would be interested to see you test the card with the fans facing up, that way gravity would be possibly pulling the condensed liquid back to the hot side of the vapor chamber
Wouldn't make much of a difference if any most likely. Capillary forces are doing the vast majority of the work in vapor chambers. Otherwise they would never work even when not defective like this
Could the issue be that AMD sourced the vapor chamber from more than one source? If these were all made at the same source the same way wouldn't they all have the same problem? Or could it be a bad batch?
It could just be the default profile is flying a little close to the sun on low fan speeds compared to power draw and there is nothing wrong with the chamber as such*. Just some are that little tiny bit more on the drying out side within the production tolerances - the 'known good' card I kinda expect will suffer the same fate in a more airflow restrictive case or hotter environment. *Obviously a bit more headroom in the design would be good, but this failure mode is to be expected of any heatpipe/vapour chamber that gets insufficient cooling to actually condense the working fluid fast enough.
Same manufacturer and factory can still mean different machine used. All parts have tolerances and in some cases tolerances can be ok, but in sum add to get something out of specification. It's a common thing that happens. Often you will hear of such a problem in the automobile branches with recalls of new cars because of an "unspecified problem". Years ago Intel had such a problem with mainboards, all MB of a specific type blew up because a component (it was a specific capacitor) from a supplier was inside the specified tolerances, but Intel miscalculated the worst cases of allowed tolerances and that led to destroyed mainboards. I worked at the time in an IT company and we had to change hundreds of boards with exploded capacitors and systems smelled like rotten eggs because of that.
Happy New Year, Roman! Thank you for your excellent work investigating this issue, your previous video on this topic received quite a bit of critique but you’ve turned things around with this video. (Even in literal sense with the card’s orientation) My AMD 7900XTX is thankfully not affected by this issue (for now) but I’m a little concerned that taking the card out of my case could result in the card exhibiting this behaviour. At the moment AMD seems to be in denial when it comes to these issues going as far as to have stated that 110 degrees junction is ‘within spec’ so I wonder if it would come to a total recall. The next question would be whether this is an issue with the design of the vapour chamber or a QC issue where the design tolerance is simply too tight where this issue can quite easily occur. One question if I may, did you check the serial numbers for these cards to see if they appear to be a part of the same batch?
Definitely the design. Looks like they moved the “in spec” goal post to accommodate a design flaw that was too late to change. Don’t forget they delayed the release by a month. How it got to this point is beyond belief. They would have definitely encountered the problem during testing, moving the cards around, in and out of test rigs. This is just one guy. They’ve got a team of engineers.
Check the batch numbers to see if the failing cards are all from the same batches or a few batches. You probably won't get info on assembly lines but I could imagine there is a problem with the equipment on some assembly line that is causing them to get the wrong amount of coolant in the vapor chamber. Maybe a leaky valve or something like that, causing the fill señores to measure one thing while some of the coolant evaporates before the chamber is sealed. Or the wicking material is not being sintered properly, causing the coolant not to move around the chamber properly.
AMD just let these units go to customers without any testing. For the money involved, it's unforgivable. We need AMD to deliver. Without them in the market, Nvidia can charge what ever they like. EKWB will do very well from AMD's massive F up.
thanks so much for the time and sharing the info... im waiting for my XFX 7900xtx to come in but contemplating returning after hearing about these higher temps... would you recommend skipping out on the 7900xtx until AMD confirms and addresses this formally?
Random question: at around 10:12, when doing this, would it make sense to physically push together the pcb and the heatsink together, just to ensure any gap created from the gravity loading is gone? Also, would it make sense to flip the card around so that the fans are face up so it hugs the pcb better? Would both these scenarios reduce the hotspot temperature? Edit: My bad it seems that mounting pressure isn't the issue at all
Sapphire MBA 7900 XT here I'm getting 79c hotspot and 92c junction temps with a 10 min furmark test. Does this junction temperature seem high or is it within range of what is expeceted?
I've tried stressing my XT out to see if i can produce the high temp, but I run mine Vert...... I can't even get the GPU junction over 75 degree's Even running multiple stress tests, or Cyberpunk 2077 on Psycho, still don't get to 80 degree's
Great work Roman. You keep upping the bar to your fellow podcasters. Very thorough scientific method that leaves no stones unturned. Thank you for your awesome detective work once again.
It's a mini refrigeration system simple as that. Liquid cools, whatever it's supposed to cool. Picks up heat ( removes it ) than flashes over to a superheated vapor. As it is traveling through the heak sink and fan's blowing, it's suppose to remove that heat and flash it back into a liquid. It just goes in a circle. Maybe the amount put into that card was not enough or too much. That will definitely cause a problem. It can get more technical, but this is how it's designed to work. If I went to your home and let out a half a pound or added a half of pound in your A/C system. It will not work right on a hot day, if work right at all.
Рік тому+9
Very good video and informative. I have an XFX Merc 7900 XTX mounted horizontally and did not see any temperature problems. Cooler runs at 2000rpm, and hot spot temp is around 80 degC. Regarding the delta... "scientific" saying you can use any unit as a delta. Degrees Celsius, Kelvin, mK, or even invent your own unit (as long you give enough information about your unit). The convention you cited is a convention for a use case or field. Being a convention or not, scientific or not, it is just more accessible for people to correlate a delta variation with the unit you used for the measurements. It would make life more difficult if I start using meters for length and inches for a delta. Using K doesn't complicate much because de relation to deg Celsius is 1 to 1, but why make life more difficult? Not everybody knows it.
Why be inaccurate just to please people who complain? Kelvin is by definition a delta, Celcius is a derivative of kelvin (with a fixed offset).
Рік тому+4
Great research.... did you try to mount horizontally, but upside-down. I.e. with the cooler upwards on top of the PCB? This should be ideal for the VC, as gravity would, in theory, help the liquid to reach the hot parts.
And I want to thank you for making your videos in english. Knowledge isn't limited to one country or one language. So thank you for taking the time to film two videos.
I wonder what would happen in a horizontal inverted position (back plate facing down). This would only be relevant in cases with convertible MB trays though.
It's certainly a scenario that should have been tested before coming to the absolute ubequivocal conclusion that it must be the vapor chamber design (which ironically doesn't get tested or verified). The conclussion has merit certainly but it's the testing regime here was by no means methodologically or intuitively exhaustive, but I guess that's the high bar for this platform, admittedly better than most. However to the OPs observation; hypothetically, horizontal mounting of the card in the manner shown to cause the high temps aka fans facing down, gravity could be affecting the fan blade body or the mount to the motors or other mechanical interface, pulling them away from the optimal position for airflow. This also plausibly explains the caveat when starting vertical mount, switching horizontal and back to vertical as once shifted out of orientation the fan blade bodies may have enough thrust to maintain this height/gap especially once the RPMs drive higher. This would have been very easy to test by simply mounting a card fan up, horizontally, and repeating the above test, though I would do so with a card that was first demonstrating lower deltas in a vertical position first just to avoid unnecessary variables. Btw while this line of inquiry may not bear fruit, it is not without historical merit, there have been oem cpu coolers long ago who had similar problems where the fan blade hub would get jostled in shipping causing poor cooling and overheating cpus, simply applying a bit of pressure to snap it back down in place would remedy, in fact I seem to recall an AMD cpu from the 486/586 days of the early 90s being the first time I heard of such, though I'm getting old enough to probably mixing up years or decades let alone models and brands... 🤷♂️
Please explain how the temperature deltas are shown in the graphs. If I do a conversion then (50C = 325.15K) and (50K = -223.15C) so are the scales wrong on your graphs or am I misunderstanding how they are represented? Sorry if I'm being an idiot. lol
A temperature difference of 50K is exactly the same as temperature difference of 50C. Kelvin and Celsius units are the same, only the C scale has a zero offset by +273.15K, taken as the freezing point of water at atmospheric pressure. So a C scale is in refference to freezing water point, while K scale is absolute temperature, where 0K is the absolute zero.
Big thumbs up for all your hard work. You have single-handedly helped out the gaming community in great measure. Hopefully, now that the issue is known it can be fixed.
I would love for you to test the new Sapphire Nitro+ card. They have been using vapor chambers for a very long time so I would think their card would perform much better.
Great diagnostic video! I too want to see a professional autopsy of the cooler and a water block test of the affected cards. Probably in reverse order, though :P If the water block test reveals the same problem, then that turns the focus back to the board and away from the cooler, which would be really trippy. However, I don't think that is the case. I think you're on the right track with the cooler being sus.
I noticed 96C hotspot temperatures on my stock 6750XT, then I overclocked it and it went to 109C while gaming so I quickly undervolted it and now it doesn't go above 86C on the most demanding scenarios while usually sitting at around 74C
I got the same card havent opened the box yet. And after watching these videos thinking returning it. But I like the look of the card so much. How much undervolt did you used?
@smuludgn not much, I'm running it at 1130 mV, which is 94% I've heard people take it as low as 1100 but mine wasn't stable there. I don't know if it's a problem that qualifies for a return, specially since there aren't, or I haven't heard, reports about cards burning for running at those temperatures. Seems like it's the normal thing for these AMD cards to have a hotspot on the mid 90s while AMD has said that they can a max hotspot temperature of 110C. I've had mine since july so 10 months now gaming every night and not a single problem. By the way, mine is the Sapphire Pulse version which has just 2 fans and no RGB, there are other models that come with 3 fans, maybe those have lower temperatures.
I think I want to start including those GPU power boards into builds, I like seeing metrics whenever I can for fault finding or thermal testing, plus, they look sick and quite cyberpunk-ish
Wow, I would have never guessed something like _that_ could go wrong. For people who don't want to RMA, would it be best to recommend them to keep using the cards but vertical-only for the time being?
If it's a defective vapor chamber, you're going to have to RMA it and replace a part anyways. Unless you're going to dissassemble it and go with third party cooling solutions, I don't see a reason not to send it back if AMD asked for a recall.
Have you tried testing with the GPU horizontal with the cooler on top rather than underneath? Some cases are orientated this way. May reduce the problem further
Much much MUCH respect for you & the Grizzly lab. Thank you for putting your time into this and giving such a well done and well put explanation on what could potentially be the issue. It's easy to point the finger at the manufacturers for their errors especially after this last round of GPU releases, but let's hope AMD takes the high road here for their own sake and makes this right. I know their R&D department has to be sweating, but this stuff happens.
It might be a quick process in the factory, but then you add days or weeks of user time to box it, send it in, wait for it to go through the queue at AMD, have it shipped back, etc. Not a great way to improve your brand plus extremely expensive for AMD. So much for first quarter profit this year.
@@AwesomeBlackDude yeah anyone who bought a $1300 AMD Gpu on release day deserves this hahahaha. Has no one been around the last decade??? AMD are famous for fucking shit up and releasing unfinished products.
Delta temperatures in Celsius and Kelvin are the same. Engineers and scientists use these two units interchangeably without even thinking about it when talking about deltas.
So if coolers are the problem, then AMD might not be totally screwed since board partners can easily remedy it by designing a better cooling solution. Situation could've been worse if we found out that the problem is with the chip design itself.
@@iancurrie8844 Yes, but the problem seems to be with some reference coolers units only, not all of them, so it's not a fundamental design flaw. I don't think AMD will issue a recall to every MBA card. Maybe they will only take back the cards that suffers from this issue and refund the customers, because it will probably take a long time to address the issue on the manufacturing side. Designing a new cooler now is completely out of question. It would be cheaper to AMD to just stop making them and rely only on board partners for this generation.
AMD still doesn't know to design a heatsink.
AMD is amazing at disappointment.
Thats Why you buy AIB cards 😆
Lol.
experiment pliss build water block maybe change ?
Both AMD and I expect CoolerMaster, need a good slap. It's kinda baffling how they thought to get away with this for the second time and no one in a multi million dollar R&D department noticed this. Reference 6900XT showed the same behaviour but only when it was mounted vertically, where the vast majority would have their card mounted horozontally so the issue didn't seem so big.
Ah, GN had explained the phenomenon of "dry out" at some point where if a heat pipe is overwhelmed it just poops its pants and the liquid never condenses back leading to poor thermal conductivity
this would be more of a defective batch or three of coolers.
Sounds like exactly this :)
dry out from geometry dash
@@white_man1352 Geometry Dash referense😮
Who are You, Who are so Wise in the Ways of Science?
Imagine if Roman and Steve had the resources of these massive manufacturers. The level of quality would be unmatched
Id say its more a matter of resource management.
Im sure the engineers behind these products are excellent. They're just working with their hands tied behind their backs by the bean counters.
i feel liek this shoudl be a thing they should be paid to be part of the quality control team but being on the public side
AMD reward this man for his diligent troubleshooting.
They are too busy making exaggerated performance charts to pretend they are so power efficient. Meanwhile, my 4090 sips power while annihilating everything AMD has ever made.
@@mattgreenfield8038 Your 4090 must be the only one that "sips power" because they didnt put those fancy high power connectors on them for a joke. Nvidia does good power design typically and these cards are built for a lot of power and have rather high power draw requirements. Jensen would likely not even exaggerate this much. The 4090 is a powerful beast that needs to be well fed.
@Boanerges57 if you had a 4090, you would know they never hit over 400 watts even when you max the power limit to 600 watts, unless you are running something extremely demanding at 4k. Look at the difference in quality between the 2 reference brands. I got a Gigabyte Gaming OC 4090, but the 4090FE is so much more well built than these AMD reference cards. Another domination by Nvidia. Why would anyone buy a 7900XTX instead of a 4090. Makes no sense. If you can't afford it, just get a 4080. This 7900XTX is worse than Vega.
AMD WHAT?? reward this man?? they knew this already and they have the numbers for damage control if and when the public catch's wind of this issue.
@Okay Yeah damage control? How much damage control can they do for this disaster, which is this RDNA3 generation. This is now worse than Vega. I'm so happy I didn't wait for these dumpster fire 7900 series and bought my 4090 on day 1. No problems. No undervolting. No black magic needed to get the 4090 to demolish these cheap RDNA3 cards. What a disaster for AMD. They deserve it for mocking the 12vhpwr that was caused by accidental user error. Look at the build quality difference between a 4080FE or a 4090FE... compared to these AMD reference cards. OMG, lol 😆
You are a legend. You did AMDs job for them and gave so many of us a legit answer. Thank you.
Its insane to believe that AMD behind closed doors doesn't already know about this issue . They are scrambling and panicking to figure out a solution before they say anything
THE THEROY OF GRAVITY overheats AMD cards such a legend thanks
shhhhh amd do not want people to know about this
They want people to think that 110c is normal
no he didn't. He THINKS it's the vapor chamber. He didn't take that apart and discover that it 100% is yet.
@@mjkittredge He pretty much tracked it down to the vapour chamber, so yes it definitly is the vapour chamber.
Wow, the amount of work it took and put into this video seems intense. So much testing. Thank you Sir.
Thanks Roman! Your work, and background as an engineer totally gives us more faith in what and how you do things, than any manufacturer!
I don't follow Reddit or Twitter just UA-cam for this info. None of the other channels are talking about this detail because of the holiday season, but thank you for your in depth analysis and news spotlight on this topic. This brings me more understanding of what JayTwoCents was possibly seeing with the down clocking under load. I was hopeful for team red this year, but looks like I'm sitting on the sidelines till one of these GPU manufacturers gets their act together.
It's a shame this problem is pretty much irreversible as is. Great diagnosis concluding with the vapor chamber. AMD is going to have to go back to square one. I always wait a certain amount of time before purchasing new tech, as with past launches it saves grief.
They don't call it the bleeding edge because it's sunshine and unicorn farts.
thats why only intel and nvidia :) you forgot that amd got out new cards but no drivers working for em..
@Tigas And you forgot that NVIDIA tried to burn down you house. People that spam "buy NVIDIA" everywhere are special type of scumbags.
@@tigriukasinlove Don't be dumb, you should never just buy into a monopoly. Did AMD fuck this up this time? They sure as hell did. However, if you want Nvidia to continue raising prices to the moon and Intel to go back to its 14n++++++, then by all means be a corporate shill and believe they have your best interest at heart.
I love how ONE person is doing a better job than a full QA department. Or they reported it and management decided to ignore it. Huge kudos to you for exposing those scams.
Yea....QA is like TY for figuring that out for us. He should be working for one of these companies, but I don't blame him for doing his own thing...seems he's doing OK. It's hard to believe that they would ignore that, cost more money in long run.
AMD doesn't have a QA department.
@@Stiitchjones he literally runs his own company based on cooling
@@LiveType Hmmmm, makes you think why aib cards custom models actually had that delay launch, huh? Amd clearly knew this was a thing. Shame on them. At least im glad that first with cpus and now with gpus we get to see the true face of that company. Not that i ever had any doubts, but yeah.
1) very limited amount of cards at launch
2) instant rumors about new revision on the way
=> means that AMD knew there is a problem with vapor chamber for a long time, maybe before launch
That really tells a lot about character of people in AMD management.
I mean everybody makes mistakes, the difference is how you face them.
10:00 when the acrylic stand didn't work, my thought is that the liquid in the vapor chamber is unable to wick up to the underside where the GPU chip is in contact. My hypothesis is that if it's mounted horizontally, but upside down, the temperatures would be normal.
AMD: It's normal. Just turn your PC upside down. That'll fix it.
Nice to see unbiased and scientific approach to figuring out problems. Nice job.
This is not a scientific approach, more like a technician approach.
@@zoltankramer3383 scientific approach would involve magnets
It's a UA-cam approach. No offense.
That is why amd is not sending him chip anymore 😅
@@zihechen3111 AMD might send a lawsuit to cease and desist. Its the same issue like nVidia, there is no problem but there are faulty cards out there all the time. Asking for people to send their broken cards in will result in... broken results. Also there might be plenty of other reasons why the hotspot temp is that much higher.. and even if it is higher maybe it is totally fine and it is meant to go that high. Publicly quickly rushing some hot takes video out for clicks is not the way to go imo, just ask AMD about it first atleast.
You could try to change the vaporchamber from a known "good" card to one of the bad cards.
Yes agree.
What’s the point? He already knows what’s wrong.
yeah not that much of a point now . this is 100 percent the problem and not a theory at this point. And it almost sounds like "working coolers" are gonna be rare finds.
Working coolers may also stop working as was mentioned in the video. Worked in one position, was changed to another orientation and it stopped working, but when changed back to the original orientation it still didn't work correctly. Seems like a very poor design where the liquid can be moved to a part of the card that breaks the phase change loop.
@@ComputerGeeks-R-Us right, but not permanently broken right? I think once the whole thing cools the phase change loop works when vertical on cold start again? I already see some hints in reddit that ppl are returning XTX cards that are ok with hotspot in fear that they will develop this issue in the future ... this is very damaging to RDNA3 ... this also shows (to nvidia as well) why AIBs have value and should be able to make decent profits
This right here is my favorite kind of video, testing different cases to pinpoint the root cause. Great job
YEAH...In the Medical field should be the same them many cures will be discovered and people fixed !
Bought a Power Color Red Devil recently, ran fine on the provided cooler (which is not reference design as far as I can tell).
I then slapped on Alpha Cool's Core cooler and got horrible temps, 105-110c hotspot. Noticed a lot of people on the AC forums having the same issue.
I tried remounting, repasting with 3 different pastes. Nothing worked.
Then i noticed the card was boostin to 3Ghz and 600W without me even touching the overclock settings.... !
I can only assume the card detected a much, much better cooler and then boosted like mad.
I set the clock back to stock clocks manually and it's been running at ~55c core and ~80c hotspot ever since.
thank you for doing the work on this. hard to believe their in-house testing did not find these results prior to product launch.
this is what really pisses me off.. impossible to miss
@@fredredd410 Not if you test for a measly 30 seconds.
Your not wrong though, impossible to miss if they bothered to do it properly.
they don't test every single card. and if its a low percentage it would be easy to test a bunch of them and miss the few that have this issue. it also seems like they don't 100% consider it an issue as they aren't allowing RMA for this issue. (yet)
It's AMD, they probably didn't even do in house testing.
You're nuts if you think they did not know this. as i said elsewhere amd released a half baked product for sale to the public. they were hoping that it would go unoticed just like their weird clocks speeds all over the place and their low fps from higher clock speeds. amd entire staffing are shills even the ones that knew releasing the product was wrong but did nothing but sit back and allow it to happen. HInt Hint Lesie Zue LOL.. now they will go into damage control mode and say it's only a few batchs here and there.
Your due diligence is impressive and appreciated Roman! Cant wait for the Vapor Chamber cut aways and maybe some watercooled testing... that would seal this diagnosis for sure and put AMD in no position to ignore this.
yes, I am full of admiration, but in my opinion he could also check how the card behaves when it is placed on its back with the cooling up, it can tell us what exactly is the problem with this chamber. it would be good if the card was tested in all positions, it would give a lot of valuable information
..I think you underestimate the ability of humans to live in denial.
@@konbiex4684 Yeah, but those would be unrealistic mounting options since average user will likely mount one of the two positions he tested. That aside, for the science of it, it would be another way of testing this vapor chamber theory. Yes.
You can't really measure the pressure in the chamber by cutting it open, unless you maybe do it under vacuum.
Measuring the amount of liquid in there tends to be difficult as well, as the water tends to be just stuck in the chamber wall coating and doesn't want to come out. You have to boil it off and re-condense it and that's experimentally pretty shaky to do it cleanly not losing or gaining any water, considering air has a fair amount of moisture in it.
So only if there was an issue with the wall coating you're gonna likely see it when you cut it open, but not the two more usual and expected heatpipe or vapour chamber faults.
@@SianaGearz indeed, but now after first tear down he knows where the fill port is before its sealed and covered. It would take quiet a bit of prep, but if he thinks it is needed, I believe Roman could create a proper test to see what the volume and pressure are. Would be another great vid!
I can’t quite get over the fact that no one picked this up from any of the manufacturers. Was there any kind of quality assurance testing?? Surely when you make a card you don’t just test it on a test bench. You test it in a normal case that any home user would use?! Mind boggling that this flaw actually went out to market!!!
I had a similar problem many years ago. My AMD R9 390 was mounted vertically (with the IO facing upwards) in my Silverstone FT02 case and it ran very hot and never came close to advertised clocks. While troubleshooting I turned the case on its side and ran a benchmark. It ran much cooler and at its advertised speeds. I think the heatpipes or vapour chamber did not work very well when mounted like that.
Ended up having to change cases to something with a normal orientation and the problem went away.
Problem here is case independent. He had it in an open test bench and it still shits the bed.
@@gionieves2646 yeah but I think he is saying he switched cases so that he can mount the cpu cooler “correctly”
@@noobslayer815 *GPU🤣
Would have been interesting to see the "bad" vaporchamber swapped onto a known good card.
Great investigative work. I was guessing the whole way though.
I second this
yeah that was like the first thing I thought...??
It might be impossible if PCB is different.
He should have contacted AMD first and ask about this phenomenon. Making some hot takes clickbait videos is not the way to go imo. Also if the card is working fine otherwise I dont even see a problem here.. for years hardware has been made to work at 100c, it just sucks if its a laptop and it burns your jewels.
@@TheTanelChannel How is it "working fine" if it's aggressively thermal throttling and can't reach it's intended power target?
And here we are 6 months later from this video and the first I've heard of the problem is from me researching which GPU to buy. So it seems literally nothing is being done about it.
This is huge. Shame gamer nexus has taken Xmas off. Maybe HUB will catch on to this story.
Well done Der8auer
4x 1000$ and you already had a XTX so 5K for one story. Amazing dedication.
Well it is not like it is 5k$ of bricks :D Just require some cash flow :P
@@rkan2 He was chopping them up with a metal grinder? Yeah, if you want to buy one go ahead
@@Irishgamer01 Where did he chop up the cards?? Ehh
you still can slap water block on them ... cards don't seem to have "problems"
@@oinodin This right there. Those Cards are 100% fine still. Not everyone can use them but for ppl who would buy Waterblocks anyways this would be basically a none Issue. Just sell them for like... i dont know... 150 Euro less if the Cooler is broken permanantly and GG. It just needs to be clear that the Card is working fine and thats that. If Roman buys Waterblocks and mounts them then he could probably sell them for more even. I dont know about the Card which he slightly modifyed but even that one should be fine if he tells the Person what he did and he sells it slightly cheaper.
Always blown away by your hard work, dedication and excellence in coming up with videos like this. If I was AMD or NVIDIA I'd keep you close and include you in a private pool of testers before openly releasing to the wider global market. Your skill set along with your passion for cutting edge hardware makes you more valuable than an employee with the same skill set but with NO heart. It is unimaginable to think of the potential loss of AMD if this was truly a design flaw. What a headache to start off 2023. Thank you Roman and happy new year. Keep 'em coming.
Dear Der8auer -- Happy New Year! -- Quick question what's the part you use that acts as both a PCIe 180 degree breakout board that also lets you read the power draw from the PCIe power cables?
Cheers and congrats on finding the root cause!
I seriously want the MBA cards to be all faulty to the point it forces AMD into an embarrassing recall, would serve AMD right for the horrendous marketing around the whole RDNA3 launch. That smugness of Scott Herkelman should cost him his reputation within RTG. The amount of trash talking and posturing was ridiculous. Time for RTG marketing to be entirely gutted.
Very impressive work. Let's hope AMD respond correctly.
They already responded with a tone-deaf response.
lol, no hope, nvidia and amd gpus teams are run by accountants and marketing guys
@@RicardoReview Basically yes.
Lol, this issue was well documented a decade ago, I'll never touch an AMD because of all the issues they have making an actually good product.
@@rohithmekala2608 Responded to this video that was posted just hours ago?
I just bought a AMD built 6950 XT and it was running up to 110c while under load and would never come back down until I let it idle. I then found your video through JayzTwoCents, and so I tried laying my case on it's side to get the GPU vertical, and same load it never went above 92c, and hovered more around 87c if I opened the glass panel that was now the top. This issue could be bigger than just on the 7900's or I was just extremely unlucky with my GPU. At least I now know I can run the card safely in the vertical position but what a pain in the a$$...
Dang, I just ordered 6950XT. We should consider vertical mount bracket? 😂
I also bought the AMD 6950 XT (reference card) and had the same problem. Because of this video and your comment i tried laying my case on the side and this helped a lot. Hotspot temp going down from 110c to 92c (under load). Also less of a difference between GPU edge temp en hotspot temp. Delta is now 20c difference instead of 40c (GPU temp 70ish hotstop now 90ish instead of 110c). Thanks for the tip!!
@@jorisgrootens5028 Glad it helped, still kind of a huge annoyance. Since I like my tower on my desk and laying it down took up too much room and didn't look great anyway, I ended up moving everything into a Hyte Y40 since the case included the vertical GPU mount and riser cable. Smooth sailing since.
@@slvrlancer You guys should RMA. Mine was staying below 100c hotspot, delta between hotspot and edge around 20-23c, under heavy load it oscillated at 77c edge and 97c hotspot with slight variations.
Considering AMD is denying that there even is a problem at all, I'd be interested to see if they do anything at all.
Nothing wrong with the card exactly. Just faulty coolers.
They will do some software upgrade that throttles the card.
@@ThezRude but the cooler is a part of the card... 🧐
@@ThezRude Cope harder AMD fanboy.
@@ThezRude just remove the cooler, whats left will be 100% in working order 😂😂😂
AMD had such a good chance to win people over from nvidia this gen and honestly was with a lot of people i spoke to who were interested in upgrading.
They really need to act very fast on this or else its all for nothing.
No, if we have learned anything ever is that people will always buy Nvidea
Its over, you get one chance and they blew it. They blew it with overpricing lackluster cards and settling to be a parasite living off Nvidias Monopoly rather than actually compete, and then if that wasnt bad enough it turned out not only were the cards horrible performance for the asking price but they were also objectively bad tech by being inherently defective. Between RDNA1 architecture/driver issues, Vega…. And now RDNA3, its time to come out and admit that atleast the Radeon portion of AMD is just straight up bad and getting an amd card is a coinflip on of its gonna be good or bad. They had a chance to shake off the issues of the past and actually let people know radeon being unreliable was a just a stigma and instead they reinforced it after talking shit about Nvidias flaws only to be even worse. Thats wild.
unfortunately not in productivity, saww 7900xtx blender performance and decided on a 3080 instead
They messed up as soon as they announced their 7900xtx wasnt $699 IMO
@@Angel7black I agree the pricing is also a problem but it all trickles down and they put in a lot more competition this gen.
its crazy to me that some random german was able to diagnose the issue before the entire team at AMD could figure it out.
he's theorizing about it. Until he replaces the vapor chamber or takes it apart it's just speculation
You assume AMD didn't know anything about this.... learn from history, tabaco, 3M, Leaded fuel....☹
Random German? Germans are masters of engineering. This is systematic, analitic testing of product. This is science, wery well done. AMD screwd up with this one.
@@zorandamljanovic6171 correction, the company that designed, tested, and produced the vapor chambers for AMD screwed up. Your comment is like blaming AMD if something was wrong with the chips they bought from TSMC
@@mjkittredge True, but it does say Radeon on those carda. Not all of them wore Radeon, but at least one was. So... AMD shoud test there stuff before they release them. Same goes for nVidia.
Love watching your content! Not only are these deep dives incredibly valuable to the PC community, they are also super interesting from a science and engineering perspective. Cheers!
Happy New Year Roman and thank you not only for this video which is an amazing investigation into what is causing these hot spot issues but all hard over this last year. Keep up the great work!
This is so spot on... I bought an XT and XTX and after tons of testing deemed the XTX suitable. The whole time I had my case laying down with the GPU vertical. I just finished my build and stood my PC up and now I'm hitting 108 junction temps from my 95 max yesterday. A delta of 40c (sorry not using kelvin haha.) total. I may have to return the card at this point... Thanks for the video I'm glad we have a solution it seems!
Some engineers at AMD are looking for a change of underpants right now, heads are definitely going to roll.
They probably be at Intel or Nvidia by the end of this week
🤣😂🤣😂
Should start at the top, they really sold this piece of crap at “$1000”, realistically $1100-$1300
I think CM does the cooler, not AMD but definitely a QC failure.
@@Angel7black Ignorant comment
@@puciohenzap891 they probably tested the cards only in a vertical mount ( normal orientation for there massive testbenchs)
Truly heroic effort. This man deserves a medal. Just used hydronaut on my 5800x3d install. It’s not a medal but I hope it helps😄
Liquid medal :)
American education...
@@jondonnelly3 love a good t1000 pun🙂👍
@@dr.derekrobinson1920 ??? I’m from Australia.
@@dr.derekrobinson1920 Evidently it's your education that's lacking lmao
Great video as always but please get rid of all static watermarks on your videos for your fans with OLEDs!
I really enjoy your whole approach here. Rather than just deciding what the problem is without sufficient testing (Nvidia power connector videos!) you’ve identified the problem with methodical testing and engineering, presenting facts and very readable graphs and data. Excellent work 😊
Shade at jayztwocents 💀
@@Pisscan lol maybe? I actually like J2C. He makes no apologies that he is just a guy with an opinion, but he did go a bit crazy on the whole Nvidia adaptor situation 😅
Amazing investigative work as usual. Thanks for focusing on facts rather than being an alarmist like a lot of youtubers who simply broadcast misinformation.
I need some help, can you tell the measurement of the screws you use for increase the pressure mout of the card? I wanna try that solution for my RX6950XT, because I think the center of the car are not doing a well contact
Did you try changing to another cooler, remove the vapor chamber from the equation, to rule out silicon problem?
if it was a silicon problem, the orientation would not change anything
@@georgwarhead2801 yeah I know, but I'm sure there are those that will try and connect it again to A0 silicon revision ...
Happy New year. This was really great research into the cause of the hot spots... Thank you for the video.
Re: Kelvin... I find it odd that the DIN standard requires reporting differences in Kelvin, but in reality since one kelvin is the same magnitude as one degree on the Celsius scale, it doesn't really change anything. So if something has a 10 Kelvin difference, it would be the same magnitude (10) in Celsius degrees.
It's a clarity difference. By reporting delta in Kelvin, there is no risk that anyone reading mistakes the delta value for the actual value at the measurement point.
@@ElijsDima Ah... very interesting. But, after some thinking on it, I don't see the real necessity. We use delta values for hundreds of different units in science all the time, and we don't create unique units for deltas. So, for example the delta for 20kg and 10kg has kg for units. We don't create a unit called dkg to make sure people know the difference. Thank you for pointing out the reason it is used. (It still seems very odd to me). I wonder what would happen if by conventions we started adding d to the units for deltas? dkg, dK, dsecs heheh
@@marcfruchtman9473 Kelvin isn't a uniquely created unit for this single purpose. Kelvin is the standard. There may be some rationale behind the fact that the delta is a calculated/derived figure rather than a sampled figure. If the measurement apparatus uses a particular unit, then that's what you report, and after that you do the unit conversions for any further calculations. So if the software report uses C then you record in C, then do conversion to K.
@@marcfruchtman9473 They do, it's called the delta symbol. Δkg
Kelvin is never used by marketing teams because it’s a scale that starts from 0 like any real chart should and that’s why I love it.
I don't know why people would complain about using Kelvin, it's directly related to Celsius. 0 Celsius (freezing point of water) is 273 Kelvin. 100C (boiling point of water) is 373K.
If you did the delta in Celsius then a delta of 0C would be 273K and that just doesn't make sense.
Very informative....one question, Didi you try swapping coolers to a different AMD vap chamber. This might be the way to find out if the wrong internal pressures are varied in the vapor chamber
Swapping coolers (from good working card) should have been second thing to do after checking thermal paste and mounting pressure!!! Obviously, he did not think of that.
Yes, I was waiting for this the whole video, this would have restricted the problem to the cooler straight from the start. Though the video would have been shorter.. 😅
This is an excellent idea
@@ElvirBegovic If the card is cooled down properly in one position and overheats in the other position then it's rather obvious that it's a cooler problem because silicon can't behave this way. Replacing a bad cooler with a good cooler would fix the issue, we can be sure of it without swapping anything, so what's the point?
@@arekb5951 point is that he spent half an hour/day fixing mounting points with a ""precision tool"" and a handheld grinder just to fix nothing!
Thankyou Roman for your detailed look into this! I do think that there is a chance that this wont 'blow-up' the same way that the NVIDIA issue did , purely because people have given AMD a pass because they dislike NVIDIA so much, so lets hope to push this and get a response from AMD :)
AMD sadly always get given a pass, you only have to look at the public response to the 7900XT which thankfully has been called out by GN, HUB and the vast majority of reviewers, even those who have given AMD passes in the past.
Aw, poor NVidia, ever the victims of mean old AMD fanboys lmao
Good Point. Ppl cape for AMD so hard because they look at Nvidia as a Villain and AMD as a Hero. Real sad to see.
@@NBWDOUGHBOY Nvidia pricing is criminal so it’s justified. AMD unfortunately are following Nvidia’s lead despite having inferior products. Corporate greed is out of control.
@@Squashed8Ball Yep. You are 1000% right. Both companies are wrong! They are both cutting corners and charging customers 2 times what a Flagship GPU used to cost. It sucks.
Great video, very informative. I purchased a reference 7900xt (Sapphire) and can't get my hot spot to go above 82 degrees, so I'm thinking it's working as intended on my end. However, I will continue to monitor temps to see if anything changes.
I have that one on my shopping cart waiting to confirm if the XTs are affected or not. :D I hope you are enjoying yours.
Please let me know if yours are still good. I have the exact same card.
You know what this reminds me of? It reminds me of the flow direction in AIO Liquid Coolers. I wonder how they designed the vapor chamber.
L🤫L 🌬️☝️
AIO liquid coolers flow direction is based on outlet on the pump. It doesnt matter if it goes left or right, the end result will still be the same.
Vapour chambers are based on thermal dynamics. Liquid contained in the vapour chamber is basically boiled to the point of evaporation and together with the heat travels up the heatpipe to an area where it is cooler or actively cooled by a fan and that cooler area or fan helps the vapour condense back into a liquid form that trickles back to the base plate of the vapour chamber where it goes through the same process again.
(This is all coming from the top of my head so I think i am mostly correct for the vapour chamber bit but there are other videos and websites that can explain the whole process in more detail)
The problem may be that once the vapor condenses in the colder part of the heatpipe, it's unable to go back to the hot part of heatpipe to start the new vaporization process. Due to the curvature of pipes, I can see that there may be some more favorable starting orientations. It's either that, or simply just bad chiplet design, as people reported the same issue even with a water block cooling solutions.
Could you please attach an AIO cooler to one of the affected 7900 XTX's to see how differents the temperatures are?
Yes
Agree
It would likely be far better. But would be helpful to verify.
👍
So I should mount it vertically? Or return it? My card hovers between 97c/100c sometimes 102c(Hotspot temp) when the fans are ramping up. The fans are going well over 2000 rpm most of the time. Thing sounds like a jet engine. I'm not happy :(
The card itself is never over 60c so it's always 40 plus delta under use.
big thanks for this video..I feel the community owes you one. This and steves video about the 12v cables are really underappreciated.
How is this underappreciated? Video has been out 5 hours... This is not Tik tok.
please feel free to make a donation on my behalf
In the end, it was user error not plugging the 12v cable in all the way. Egg on the face of GN and Steve
@@gutterg0d fuck tictok. and, I think you missed my point.
@@ssaini5028 That is the conclusion in the video, but still vendor error for not adding the plug locks to the plugs., think before you speak/type.
der8auer is such a great engineer, love all his engineering series. He should set-up a consultancy for QC and design enhancement for these manufacturers.
Jimmy Neutron was created based on him
That won't solve anything, 99% sure that AMD knew about these issues but decided that "it's fine" because they couldn't release the card any later.
For those wondering about Kelvin and Celsius :
In thermodynamics, when you're doing complicated maths, you don't want negatives values, and you was a little numbers as possible to simplify the equation and allow you to find the solution to the problem as easy as possible.
This is exactly why Kelvin was invented, removing all potential for negative values from equations (at least all those that are related to temperatures), therefore all the thermodynamics equations are based on the Kelvin scale, no the Celsius scale (even though, as you all know, the diference of temperature between two Celsius degrees is the same as the difference between two Kelvin values). If you wanted to do your calculations on the Celsius scale, using all the proven thermodynamics equations, you'd have to add 273.15 degrees to all of your temperature values (25C is 273.15K + 25 = 298.15K). It's jarring and just plain not necessary. It's easier to do all the math using Kelvin and then convert it to Celsius once you got your result.
For deltas though, it's barely matters if you say that you have a delta of 25C or 25K as those are the exact same temperature difference
Awesome work Roman 💪🥳🎊 and a great find indeed! Let’s be real every business or corporation stuffs up……. the thing is how they respond and rectify the matter which is important! We’ve seen this from Nvidia, Intel, AMD, Gigabyte, MSI, Asus ….. looking forward to what the next chapter of this saga holds in store 🍿🍿
Every time I have to talk to someone in customer service I die a little inside.
Mad. This kind of hard throttling could also explain the bad benchmarking results that the xtx card seems to gt.
Review samples are cherry picked and tested throughoutly before-hand, none of the reviews show big delta between edge and hotspot temp
@@krizby87 They are tested but not cherry picked.
Well. At least in some Cases, yes... it could explain some strange Numbers. But who knows really. Whishfull Thinking as well i think. Lets wait and see what happens.
I am a bit surprised this is not common knowledge within such a community that kelvin and celcius are interchangeable if using deltas.
The only difference is the minimum temperature: kelvin has no negative temperature. It only has a negativ sign if you compare a higher temperature to a lower temperature so the delta of those to temperatures would be -x Kelvin.
You could replace the vapor chamber with one from a good card and measure the temperatures again :) if temps are normal, you are sure it is the vapor chamber
He doesn't have to. He has used the Sherlock Holmes method. when you have eliminated all the other factors, whatever remains, however improbable, must be the cause.
I think I would love to see this
@@FabledGentleman thats bs, it's an arrogant method of diagnosis which assumes you have thought of every possible factor.
This is a good idea.
Why? This man does this for a living. There is no need to do that. The problem is identified. Don't give AMD cover. They messed up and they need to fix it ASAP.
From the start of the video I had the feeling it had to be that the vapor chamber was drying out. I was hoping you would test it horizontal in the opposite orientation, ie with fans pointing up instead of down. Maybe that would allow the liquid to reach the hotspot, and reestablish a proper flow. The capilary effect may need some help from gravity. (Also it would serve as a test for other potential issues, such as pockets of hot air not escaping by convection when the card was horizontal.)
But if only some cards experience this there has to be a variable in the manufacturing process? Could it be that the composition or amount of the fluid is wrong in some cards?
@@wombatillo Likely the lower pressure in the vapor chaber is not low enough, causing the boiling point to be too high, Then again why would orientating it differently cause it to work better? aybe the heatpipes/vapor chamber are leaking? Bad stuff here regardless of the problem. I was really hoping AMD would gain some traction in the gpu market as we all benefit from close competition.
could be an issue with the wick, if its missing in an area liquid won't move as its supposed to.
I also wanted to see if testing the other horizontal orientation would have an affect.
7900 xt owner here no problems with cooler, I think your unaware the 7900xt is a smaller different model cooler. This seems limited to the 7900 xtx model.
Thank you for your comment🙏 i was getting very nervous as i have a reference 7900xt on the way
Random comment here - when the donut of death appeared (furmark) my laptop fans actually span up a bit. Confused the hell out of me haha. Thank you for this reporting Roman, and for getting those gamers good cards. Very decent of you.
Fantastic work, agree 100%. I would only add that this should be a (very bad) production issue, not a design flaw. I have a direct from AMD 7900 XTX and mine is perfectly OK, I never saw hotspot temps above 90C and that's running stress tests and many benchmarks with the card overclocked including +12% PL. And I have it on the horizontal position. So I guess I got lucky, still, if AMD has to issue a blanket recall I would take it because even those of us with a good unit will lament it when it's time to resell the card as used, unless AMD steps up and does whatever it takes to fix this.
Could you try flipping the card fans-up? That would be to have the liquid go down with gravity again (if in a liquid state at all) and/or also kind of shake it between position change (by letting your mount for example fall on the table from small height cable permitting), to try to dislodge if it's stuck in the wick/surface tension? If any of that does anything, I don't know what it means but it'll be hilarious:)
It's unlikely that force will reset the function because it's a vacuum with mostly water vapor instead of liquid water. There's likely an angle that would cause it to work again but there's little to gain from finding it.
Thanks for great work you have done. Have you tried to swap the vapor chamber between Good and Bad cards? just to be sure that there are no other variables? (e.g. chip, vram etc)
Have you tested cards that DO NOT have the hotspot issue to see if there are any physical differences in the cooler dimensions?
Really good question, I'm going to mess around with my xtx tonight in horizontal vs vert. I haven't NOTICED an issue (the things a beast) but maybe its there.
Celsius and Kelvin values are the same when measuring differences (deltas), water has 100 degrees of difference between its melting and boiling points in both scales. The only difference is when measuring absolute temperatures, 0 degrees in Celsius is 273.15 degrees in Kelvin, 100 degrees in Celsius is 373.15 in Kelvin, still 100 degrees of difference between both temperatures in either scale.
oh this reminded me of an article i read about space ship design and how some heat pipes depend on gravity to function properly and some depend on zero-g. not that this is the case here since the solution isn't just orientation, but it was interesting to me that until then i just assumed a heat pipe was a simple device that was just relying on conduction and didn't know there was a liquid inside with phase change. i thought oh heat pipes are more interesting than i imagined 😀
Ice giant CPU hsf heat pipes only work in certain orientation, if used in other orientations other than the ones they say it can work in, it will not work. They explicitly say what positions it can be mounted in. Relying on gravity as well as heating and cooling. They can also stop function if too much heat is provided in those positions aka dryout. It is fairly interesting technology with limitations.
Could you possibly run your tests on the affected cards in your possesion with a water cooler in place of the vapor chamber design as an additional confirmation?
If you watercool your 7900XTX you will be fine.. he already confirmed it was the heat sink and putting a water cooler on it won’t change anything.
My thoughts as well! That would be absolute proof!
Apples/Oranges
LC would give a different result over AC regardless
@@Nightdare but will it have the same results with mounting differences from horizontal to vertical.
@@LegionGamingTV there are other issues as well. the ram speed issue where you can get the promised 3 ghz speeds from sensors but a 50% decrease in framerate. So all the stats look better on paper but there is a slow down somewhere else. Jayztwocents believes there is an issue with the memory manager as well. too many issues with this card but atleast till now no cards have self combusted but they sure arent gonna get record lifespans with these cards.
Curious if you are going to put on a water block to solidify the results of the testing. It would eliminate the vapor chamber as a factor and see what the actual true temps of the card would be under nominal cooling. Great process and video!
I was thinking the same thing. Test it again with water cooling to see how the performance improves.
Mine is water blocked still hits up to 112°C
I would be interested to see you test the card with the fans facing up, that way gravity would be possibly pulling the condensed liquid back to the hot side of the vapor chamber
waste of time
Wouldn't make much of a difference if any most likely. Capillary forces are doing the vast majority of the work in vapor chambers. Otherwise they would never work even when not defective like this
Nobody mounts their GPU that way. Total waste of time.
Could the issue be that AMD sourced the vapor chamber from more than one source? If these were all made at the same source the same way wouldn't they all have the same problem? Or could it be a bad batch?
It could just be the default profile is flying a little close to the sun on low fan speeds compared to power draw and there is nothing wrong with the chamber as such*. Just some are that little tiny bit more on the drying out side within the production tolerances - the 'known good' card I kinda expect will suffer the same fate in a more airflow restrictive case or hotter environment.
*Obviously a bit more headroom in the design would be good, but this failure mode is to be expected of any heatpipe/vapour chamber that gets insufficient cooling to actually condense the working fluid fast enough.
Same manufacturer and factory can still mean different machine used. All parts have tolerances and in some cases tolerances can be ok, but in sum add to get something out of specification. It's a common thing that happens. Often you will hear of such a problem in the automobile branches with recalls of new cars because of an "unspecified problem".
Years ago Intel had such a problem with mainboards, all MB of a specific type blew up because a component (it was a specific capacitor) from a supplier was inside the specified tolerances, but Intel miscalculated the worst cases of allowed tolerances and that led to destroyed mainboards. I worked at the time in an IT company and we had to change hundreds of boards with exploded capacitors and systems smelled like rotten eggs because of that.
@der8auer EN did you try putting the card horizontal, with the GPU facing up?
Happy New Year, Roman!
Thank you for your excellent work investigating this issue, your previous video on this topic received quite a bit of critique but you’ve turned things around with this video. (Even in literal sense with the card’s orientation)
My AMD 7900XTX is thankfully not affected by this issue (for now) but I’m a little concerned that taking the card out of my case could result in the card exhibiting this behaviour.
At the moment AMD seems to be in denial when it comes to these issues going as far as to have stated that 110 degrees junction is ‘within spec’ so I wonder if it would come to a total recall.
The next question would be whether this is an issue with the design of the vapour chamber or a QC issue where the design tolerance is simply too tight where this issue can quite easily occur.
One question if I may, did you check the serial numbers for these cards to see if they appear to be a part of the same batch?
Definitely the design. Looks like they moved the “in spec” goal post to accommodate a design flaw that was too late to change. Don’t forget they delayed the release by a month. How it got to this point is beyond belief. They would have definitely encountered the problem during testing, moving the cards around, in and out of test rigs. This is just one guy. They’ve got a team of engineers.
Check the batch numbers to see if the failing cards are all from the same batches or a few batches. You probably won't get info on assembly lines but I could imagine there is a problem with the equipment on some assembly line that is causing them to get the wrong amount of coolant in the vapor chamber. Maybe a leaky valve or something like that, causing the fill señores to measure one thing while some of the coolant evaporates before the chamber is sealed. Or the wicking material is not being sintered properly, causing the coolant not to move around the chamber properly.
Exactly it a manufacturing defects but people will blame amd anyway
@@legros731 It iis AMDs fault in the end. They chose the design and cooler supplier with shitty QA to maximize profits. Stop simping.
@@legros731 it's still their fault because they released the card with this QC issue
@@legros731 Of course, they're the ones selling these for $1000.
@@legros731 AMD and Nvidia are not our friends.
My Sapphire 7900 XTX reaches 112C Hotspot. If i play Elden ring on 4K and the GPU is under 100% Load it crashes.
You should return it.
AMD just let these units go to customers without any testing. For the money involved, it's unforgivable. We need AMD to deliver. Without them in the market, Nvidia can charge what ever they like.
EKWB will do very well from AMD's massive F up.
Replace *any* with *enough*
thanks so much for the time and sharing the info... im waiting for my XFX 7900xtx to come in but contemplating returning after hearing about these higher temps... would you recommend skipping out on the 7900xtx until AMD confirms and addresses this formally?
Return it !!!
Random question: at around 10:12, when doing this, would it make sense to physically push together the pcb and the heatsink together, just to ensure any gap created from the gravity loading is gone? Also, would it make sense to flip the card around so that the fans are face up so it hugs the pcb better? Would both these scenarios reduce the hotspot temperature?
Edit: My bad it seems that mounting pressure isn't the issue at all
Genuinely thank you for your work!! Great channel a nice more technical approach than others. Happy New year dood !
Did you test the card in all 6 orientations once it had gone to 110C? I wondered if a different orientation would ‘fix’ the issue.
He tested that. It would not go back to normal when moved to the working position when on.
Sapphire MBA 7900 XT here
I'm getting 79c hotspot and 92c junction temps with a 10 min furmark test. Does this junction temperature seem high or is it within range of what is expeceted?
Thankyou,
Can see you put alot of work into these vids for us and it's appreciated!
Does this also affect the 7900xt? Thx
probably not, since the XT uses a different cooler as the XTX, i just never heard of a XT having this problem so take that with a grain of salt
Can't say for myself, but I'm pretty sure it does. Let's hope the 7800 and below skus are fine.
I've tried stressing my XT out to see if i can produce the high temp, but I run mine Vert...... I can't even get the GPU junction over 75 degree's
Even running multiple stress tests, or Cyberpunk 2077 on Psycho, still don't get to 80 degree's
@@Xsorus The vertical mount is the reason why, if you mount it horizontally your temperatures may rise.
@@trr4gfreddrtgf Its quite possible, but I have a hyte y60 case so that's not going to be an issue
Great work Roman. You keep upping the bar to your fellow podcasters. Very thorough scientific method that leaves no stones unturned. Thank you for your awesome detective work once again.
.... this might explain why AMD did not send him a review sample:-)
It's a mini refrigeration system simple as that. Liquid cools, whatever it's supposed to cool. Picks up heat ( removes it ) than flashes over to a superheated vapor. As it is traveling through the heak sink and fan's blowing, it's suppose to remove that heat and flash it back into a liquid. It just goes in a circle. Maybe the amount put into that card was not enough or too much. That will definitely cause a problem. It can get more technical, but this is how it's designed to work. If I went to your home and let out a half a pound or added a half of pound in your A/C system. It will not work right on a hot day, if work right at all.
Very good video and informative. I have an XFX Merc 7900 XTX mounted horizontally and did not see any temperature problems. Cooler runs at 2000rpm, and hot spot temp is around 80 degC. Regarding the delta... "scientific" saying you can use any unit as a delta. Degrees Celsius, Kelvin, mK, or even invent your own unit (as long you give enough information about your unit). The convention you cited is a convention for a use case or field. Being a convention or not, scientific or not, it is just more accessible for people to correlate a delta variation with the unit you used for the measurements. It would make life more difficult if I start using meters for length and inches for a delta. Using K doesn't complicate much because de relation to deg Celsius is 1 to 1, but why make life more difficult? Not everybody knows it.
Why be inaccurate just to please people who complain? Kelvin is by definition a delta, Celcius is a derivative of kelvin (with a fixed offset).
Great research.... did you try to mount horizontally, but upside-down. I.e. with the cooler upwards on top of the PCB?
This should be ideal for the VC, as gravity would, in theory, help the liquid to reach the hot parts.
And I want to thank you for making your videos in english. Knowledge isn't limited to one country or one language. So thank you for taking the time to film two videos.
I wonder what would happen in a horizontal inverted position (back plate facing down). This would only be relevant in cases with convertible MB trays though.
It's certainly a scenario that should have been tested before coming to the absolute ubequivocal conclusion that it must be the vapor chamber design (which ironically doesn't get tested or verified). The conclussion has merit certainly but it's the testing regime here was by no means methodologically or intuitively exhaustive, but I guess that's the high bar for this platform, admittedly better than most. However to the OPs observation; hypothetically, horizontal mounting of the card in the manner shown to cause the high temps aka fans facing down, gravity could be affecting the fan blade body or the mount to the motors or other mechanical interface, pulling them away from the optimal position for airflow. This also plausibly explains the caveat when starting vertical mount, switching horizontal and back to vertical as once shifted out of orientation the fan blade bodies may have enough thrust to maintain this height/gap especially once the RPMs drive higher. This would have been very easy to test by simply mounting a card fan up, horizontally, and repeating the above test, though I would do so with a card that was first demonstrating lower deltas in a vertical position first just to avoid unnecessary variables. Btw while this line of inquiry may not bear fruit, it is not without historical merit, there have been oem cpu coolers long ago who had similar problems where the fan blade hub would get jostled in shipping causing poor cooling and overheating cpus, simply applying a bit of pressure to snap it back down in place would remedy, in fact I seem to recall an AMD cpu from the 486/586 days of the early 90s being the first time I heard of such, though I'm getting old enough to probably mixing up years or decades let alone models and brands... 🤷♂️
Please explain how the temperature deltas are shown in the graphs.
If I do a conversion then (50C = 325.15K) and (50K = -223.15C) so are the scales wrong on your graphs or am I misunderstanding how they are represented? Sorry if I'm being an idiot. lol
A temperature difference of 50K is exactly the same as temperature difference of 50C. Kelvin and Celsius units are the same, only the C scale has a zero offset by +273.15K, taken as the freezing point of water at atmospheric pressure. So a C scale is in refference to freezing water point, while K scale is absolute temperature, where 0K is the absolute zero.
Is the power monitor tool you are using custom made or is it commercially available? The one that does your 2x8 pin u turn with the LCD display.
Big thumbs up for all your hard work. You have single-handedly helped out the gaming community in great measure. Hopefully, now that the issue is known it can be fixed.
So... either AMD recall everything, or sent out replacement cooler and tell customer to either replace it yourself else go pound sand?
Which ever solution is cheaper is the one they most likely will take. 🤷
AMD acting like they have the over 80% GPU market share
🤣😂🤣😂
@@meekmeads Doesn't matter. They're still going to take the cheapest option. 🤷
So AMD wasn't the saviour every amd fan boys were hoping. Who would have thunk?
I would love for you to test the new Sapphire Nitro+ card. They have been using vapor chambers for a very long time so I would think their card would perform much better.
They probably designed the reference card :/
Not Sapphire, but my XFX Merc XTX is running great, even with a high overclock and silent fan curve. 👍
Great diagnostic video! I too want to see a professional autopsy of the cooler and a water block test of the affected cards. Probably in reverse order, though :P If the water block test reveals the same problem, then that turns the focus back to the board and away from the cooler, which would be really trippy. However, I don't think that is the case. I think you're on the right track with the cooler being sus.
I noticed 96C hotspot temperatures on my stock 6750XT, then I overclocked it and it went to 109C while gaming so I quickly undervolted it and now it doesn't go above 86C on the most demanding scenarios while usually sitting at around 74C
I got the same card havent opened the box yet. And after watching these videos thinking returning it. But I like the look of the card so much. How much undervolt did you used?
@smuludgn not much, I'm running it at 1130 mV, which is 94%
I've heard people take it as low as 1100 but mine wasn't stable there.
I don't know if it's a problem that qualifies for a return, specially since there aren't, or I haven't heard, reports about cards burning for running at those temperatures.
Seems like it's the normal thing for these AMD cards to have a hotspot on the mid 90s while AMD has said that they can a max hotspot temperature of 110C.
I've had mine since july so 10 months now gaming every night and not a single problem.
By the way, mine is the Sapphire Pulse version which has just 2 fans and no RGB, there are other models that come with 3 fans, maybe those have lower temperatures.
I think I want to start including those GPU power boards into builds, I like seeing metrics whenever I can for fault finding or thermal testing, plus, they look sick and quite cyberpunk-ish
Wow, I would have never guessed something like _that_ could go wrong. For people who don't want to RMA, would it be best to recommend them to keep using the cards but vertical-only for the time being?
If it's a defective vapor chamber, you're going to have to RMA it and replace a part anyways.
Unless you're going to dissassemble it and go with third party cooling solutions, I don't see a reason not to send it back if AMD asked for a recall.
Have you tried testing with the GPU horizontal with the cooler on top rather than underneath? Some cases are orientated this way. May reduce the problem further
Much much MUCH respect for you & the Grizzly lab. Thank you for putting your time into this and giving such a well done and well put explanation on what could potentially be the issue. It's easy to point the finger at the manufacturers for their errors especially after this last round of GPU releases, but let's hope AMD takes the high road here for their own sake and makes this right. I know their R&D department has to be sweating, but this stuff happens.
Wow, wasn't expecting this finding! Great job! Sucks for amd though. Hope a cooler replacement would be a quick RMA process.
It might be a quick process in the factory, but then you add days or weeks of user time to box it, send it in, wait for it to go through the queue at AMD, have it shipped back, etc. Not a great way to improve your brand plus extremely expensive for AMD. So much for first quarter profit this year.
Just to stay away from these new defaulty products especially in a recession cycle.
@@AwesomeBlackDude yeah anyone who bought a $1300 AMD Gpu on release day deserves this hahahaha.
Has no one been around the last decade??? AMD are famous for fucking shit up and releasing unfinished products.
Delta temperatures in Celsius and Kelvin are the same. Engineers and scientists use these two units interchangeably without even thinking about it when talking about deltas.
true. i dont understand the "outrage" of some people.
its also the legal temperature in the EU and many other countries. so people should know it.
So if coolers are the problem, then AMD might not be totally screwed since board partners can easily remedy it by designing a better cooling solution. Situation could've been worse if we found out that the problem is with the chip design itself.
Third party boards and coolers work very well indeed. The issue is just with AMD’s design.
@@iancurrie8844 Yes, but the problem seems to be with some reference coolers units only, not all of them, so it's not a fundamental design flaw.
I don't think AMD will issue a recall to every MBA card.
Maybe they will only take back the cards that suffers from this issue and refund the customers, because it will probably take a long time to address the issue on the manufacturing side.
Designing a new cooler now is completely out of question. It would be cheaper to AMD to just stop making them and rely only on board partners for this generation.