As "Oops of Twitch" pointed out, memory channel names layout do not match to VIK-on and that makes sense considering the fact that all memory chips at the bottom had ripped pads and two of them are G and H The confusion actually started while back when this card was new to me. After asking around for a while, i got mixed answers and i ended up using boardview as a guide. It looks like Boardview cannot be trusted or AMD is pulling our leg with making repairs more difficult. In which case, Thanks AMD Needless to say, Ill get the chips and put it on and we both learn something. Hopefully it will work. I'd rather get paid for a fix than given an expensive donor for parts any day.
Hey Tony.. good work to try out more options. But I think, if there was a separate jig/ device to test the GPU Core data and control lines, would have helped in such sticky scenarios! But I believe such devices will not be cheap anyways!
A lot of it seems to be the sheer size of these video cards, the PCB's inevitably flex more and you tend to find this damage to BGA components in the same location near the bottom of the card. The damage might even occur during first insertion into the slot if you're not extremely careful, if not the weight of the card and the sag that results when horizontally mounting seems to almost inevitably cause damage to BGA pads.
i wonder if the memory batch could be some kind of a snowflake that simply broke down with heating. is there any low melting aloy solder that isn't brittle as f like indium balls or something wierd like that so the memory could flex a bit without delaminating pads on either side? so you could place new memory chips one by one and test if the error propagate or if it just somehow start working but if not you could recover those chips without burning them
with the amount of corrosion on this card it was probably produced during the pandemic era and stored or mined near the Asian coast in humid salty conditions. I've seen hundreds of cards like this, even "new/old stock ones". If I was buying anything AMD I would stick to Saphire branded cards. But this is just my opinion.
I bought mine direct from AMD to avoid any issues. I know everyone doesn't have the cash for a new card, but buying direct is a great way to avoid this. I would have returned the GPU if i t showed up like this.
So... the chiplet wasn't the problem, after all?! Anyhow, chiplets are here to stay, whether we like it or not. Personally, I've no problem with them, so long as it means lower prices (at least to a certain extent) and/or larger stock quantities due to (much) improved yields. After all, yields start to get a little iffy once you move past 350 sq-mm or so, regardless of the fab or overall yields. That's one reason there's usually a massive price gap between mid-range and high-end GPU SKUs and the price-to-performance ratio plummets to the ground the higher you go in the GPU hierarchy.
@@northwestrepair What about soldering chips one by one and looking at how output of the test software changes? That might help deduce if it's a discrepancy between software and hardware channel numbers or not.
@@northwestrepair the question is if teh software is chiplet aware - otherwise if its a virtual mapping transparently to teh card thru the MCDs then youd need anotehr softeare that is chiplet aware - as that what could casue the faulty channels to shift depending on faulty chips / combo.
@@northwestrepairIf I may ask - Why don't you have a stock of various memories, they will come in handy sooner or later? P.S. Thanks for the videos, they're awesome 🙂👍
@@mareck6946 - There's nothing specific to be "aware" of, from the software's point of view. It doesn't care if the modules were made separately and then joined together or if they were all made at the same time from the same piece of silicon. It's like the difference between having a single 10 metre power cable or two 5-metre cables connected to each other. The connector might break, just like the cable itself might break, but the device being powered by that cable can't tell the difference. And it's not like a normal repairperson can repair the chiplet connections (or the connections inside a single chip - those would be even harder), so it doesn't make any difference for repairs, either. The most important step would be to put the GPU on a working card (or put a known working GPU on this card), to determine if the damage affected the GPU itself or just the card. If it's the GPU, you can't fix it with normal tools (regardless of it being monolithic or not). If it's the card, it might be fixable. But I think Tony has already said he doesn't have another card or another of these GPUs to test.
Low-quality memory chips whose pads rip off? That's new. Usually we get shown pads ripping off the board. As for the working card at the end, did you try each of the outputs? Maybe the one you use is good, but the one the customer uses is faulty.
Manufacturers have freedom to design their own cards with different layouts. Asus (for example) often makes two card models with different layouts for each GPU (a smaller one with just 1 or 2 fans, close to default clock, and a bigger one, with 3 fans, with higher overclocks). I think Sapphire also has multiple models for some GPUs. So you do need to get a boardview for the correct model, you can't just assume they're identical.
You're definitely onto something here. Memory chips can only be routed one way due to signal integrity issues, and VIK-ON and NWR's cores are rotated the same way. I think @northwestrepair has a problem with his memory layout, especially since VIK-ON's chart shows H channel right next to the PCIE hook, where impact damage is the most common.
I've been using AMD for a long time, so I was kind of surprised when you ripped AMD as hard as you did over their chiplet design. Never had issues with anything really so far, aside from general GPU tinkering from time to time because every GPU has its quirks. So I'm glad to see you giving this another shot, even though you have some really hardline opinions about the card. Not sure if it will matter too much to you, but I tend to only buy from Sapphire, except when it's a workstation card. But Sapphire makes those too I understand... if I am correct, AMD has all their own in-house cards manufactured by Sapphire; and Sapphire just makes their own branded models as well. The only AMD card I ever really had trouble with, technically wasn't even an 'AMD' card in the first place. It was an old ATi card, manufactured by Asus. Everything else has worked to spec, for the most part, with some tweaks or repastes needed due to shoddy paste application on the line. (A result of basically almost everyone doing it wrong for a long, long time, more so than any one manufacturers fault. Also cheap paste is cheap paste, it needs be replaced.) So again, glad to see you giving this one a second shot at least. And apparently even a third shot, heh... One last thing though occurs to me during this comment and watching at the same time... You mention the program might be glitched or something. I am remembering some other video, done by you? or someone else... where the order of the chips got reversed, and so things that were reporting as bad were not testing as bad like you are with checking the data lines. The resulting flipping around of the errors, perhaps, is because you did inadvertantly fix the original chips by happenstance, but another one was faulty still anyways so it jumped to showing that one in error now. But because you are focused on the wrong chips maybe? You keep getting sent back and forth between the 'issues'. I could of course be entirely wrong, but if it helps, cool.
@@northwestrepair definitely not normal for a 6700xt, its a 16 lane card, you should check it (rx 6600 is an 8 lane card on other hand), maybe the dataline capacitors are not in good shape? that would also explain the 'no detect' from the customer? who knows
6700 xt . what the life time of gpu maximum and the average and when the temprature about less than 65 celcius because i have gpu from 5 years and work correctly
@@sultanalharbi9182 Just be careful when you leave it in box somewhere for a long time, electronics dont like being stored for years (especially capacitors!), using it once every few months should reduce chances of developing problems greatly. It is never zero however...even on a new, out of the box card.
@@alexturnbackthearmy1907 this so true,i've recently took my zx48 from years of slumber and now i have lost 32kb of memorybuut,i can still play Jet Pac and,its still awe
I'm not sure if I would call _any_ brand a "top" brand, but PowerColor I would say is a cheaper one. Sapphire is decent, as are Zotac, and XFX or, to be more honest, they used to be widely considered decent. Things change, for the better or worse
@@GeorgeTsiros I had their Red Devil variant as my 2nd GPU (1st one was the reference card that had a vapor chamber defect). It was one of the limited edition ones that was #2300 out of 3000. It should have been better than the rest. I was really hoping they did something special but it was just another dumpster fire GPU
@@leonidastankiangaming Was it something with the board or the cooling? Did you try adjusting/fixing it? I have a vega56 (sapphire vega pulse) and it desperately needed cooling. The only waterblock out there was bykski's but even that was _not_ officially released. You will not find it listed on any official bykski site. It was badly made: it has a step on the coldplate, it is not flat.
@@GeorgeTsiros I did everything I could to reduce temps and power draw. I reduced the voltage, underclocked, and even adjusted the fan curves. The crazy thing is at idle the Red Devil GPU ended up running as hot as 60°c while using more than 100w doing absolutely nothing with fans blasting at over 2000 rpm. Under load, 105°c core temp with hotspot way exceeding that temp. it was using up to 600 w and couldn't go past 17.89 FPS in the Cyberpunk benchmark at 1080p without RT and without FSR. It gave the same results with Heaven Benchmark and Superposition Benchmark and MSI Kombuster. No luck. I swear Radeon 6000 Series and RDNA2 are more stable and reliable than Radeon 7000 Series and RDNA3.
Hi, I bought and RX 590 month ago for around 100$, it was working all fine until yesterday, while updating drivers, the screen went complete orange and pc crashed or something, then I started having either pink or green screens when I turn my PC on, and its like that since then, sometimes I see display for 1~5 minutes but then it crashes, the GPU was all fine before, its temprature was good, never crossed 75 in extreme usage, and I also didnt got any burned smell. Can you please tell me if you've encountered something similar in repairing gpus and if its repairable and if you have made a video on how you've fixed it, because I am from Pakistan so I can't send it overseas for repairs.
Wipe the driver and reinstall previous version. Perhaps windows took liberties and updated your driver itself. Try booting a linux bootable, manjaro is a good choice to test and make sure it's not windows related.
@@TheReaper-xl7qp Could be monitor cable, could be custom color settings in driver or it could be the card was flashed with a mining bios. If the card is dual bios, try the other, try replacing the hdmi/dp cable and try moving it to another port on the card.
I never used an AMD card, probably because i am not used to the alternative, but this videos teaches me reality and experience is often the better choice. Hopefully i dont need to replace my graphics card unless i need to.
If you're basing your decision about which card to buy on repair videos showing cards that suffered physical damage, you're doing it wrong (unless, of course, you're planning to hit your graphics cards with a hammer). There are plenty of channels (Gamers' Nexus, Hardware Unboxed, etc.) showing the actual strengths and weaknesses of each model. >95% of them will never be sent for repair (either they fail within warranty or people upgrade after a few years). Personally, I'm not a big fan of any GPU in the current generation, they're all overpriced and / or underpowered. The only ones that come close to being good value for money are Intel's cards (but their drivers are still very "green"), the 7900 GRE, the 7800 XT, and (in a few situations) the 7600 XT (enough memory but not enough bandwidth) and the 4070 Super (enough bandwidth but not enough memory). Even those are overpriced, but not as much as most other models. Hopefully, the next generation will be significantly better, or at least have sane prices for the real performance they deliver.
@@RFC3514 I am still a newcomer to the world of graphics card, but as of now, my reality is surrounded by people and contacts having experiences with NVIDIA graphics cards. Mine is an RTX 3090 EKWB. But i do agree, this generation of graphics cards is better avoided, where both sides have their respective problems.
@@noxxiisproject2075 - Oh, the main problem is the same on both sides: greed. This generation is barely any faster than the previous one, so they released drivers that do frame interpolation (i.e. create a "fake" frame between each two real 3D rendered frames) to pretend they're twice as fast. I wonder which of them will be the first to add _two_ interpolated frames between real ones, to say they're _three_ times as fast. At least AMD tends to make open standards that any company can use, while Nvidia tries to lock people into their products, like Apple.
@@RFC3514 Barely any faster? I've got a 7900xtx and a 6950xt in different pcs on either side of my desk. The 7900xtx is a solid 40-50% uplift in performance while using less power and costing less. Infact, the 6950xt was a refresh that cost 1250$ a year after the 6900xt (real previous gen) came out at 1000$. So it costs the same as the 2 year old card and is 50% faster. Comparing it to the mid gen refresh card in performance and comparing price against the clearance sale prices of previous gen is silly.
@@PineyJustice - Launch price of the 6950 XT was $1099, not $1250. And one year ago (when both were officially in production at the same time) it was $599 (official AMD price), while the 7900 XTX was $999 (also official AMD price). That was 66% more money for 40% (on a good day) better performance. Also, the 7900 XTX doesn't use less power. It uses less power _to render the same number of frames,_ but if you want that 40% increase in frame rate (i.e., under full load), it actually consumes slightly _more_ power (though not 40% more). It also consumes more power when idle. Drop down one "level" (i.e., 7800 vs. 6800) and the performance vs. price ratio becomes even more in favour of the previous gen. AMD's current cards aren't _quite_ as overpriced as Nvidia's, but they're still 20-30% more expensive than they should be, for the relative performance (or 20-30% slower than they should be for the price, if you prefer).
Shows a bunch of impossible to repair defects for a vast majority of users/owners... Comment section: "see you are just being biased. AMD makes awesome products!!!" Never change internet😂😂😂
@@emilbotev3656 - But everyone knows that when you drop a card with an AMD chip on it, that's AMD's fault. In fact, when you drop a card with an Nvidia chip on it, that's probably also AMD's fault.
Most cards he shows have defects that can't be repaired by the average user, the average user can't even check power lines,let alone fix many things he records regularly, such as when Nvidia pcb cracks under its own weight and he has to dremel layers and solder super thin wires, or heavy corrosion, or when he ends up with ripped pads whenever he has to reball the core and memory, etc. That's why they sent it to an specialized technician. The difference is he doesn't bitch and moan with Nvidia like he does with AMD.
Both issues can be fixed with an undervolt. Exactly what I did with my 3080Ti, dropped 50w of load, never exceed 75c anymore and coil whine almost gone
I have a sapphire nitro+ with +15pl and have around 78hotspot max most of the time then running to hot is due to GPU thermal paste pump out but mine has no coil wine my friends did until he swapped his psu to a 1000watt
As "Oops of Twitch" pointed out, memory channel names layout do not match to VIK-on and that makes sense considering the fact that all memory chips at the bottom had ripped pads and two of them are G and H
The confusion actually started while back when this card was new to me. After asking around for a while, i got mixed answers and i ended up using boardview as a guide. It looks like Boardview cannot be trusted or AMD is pulling our leg with making repairs more difficult. In which case, Thanks AMD
Needless to say, Ill get the chips and put it on and we both learn something. Hopefully it will work.
I'd rather get paid for a fix than given an expensive donor for parts any day.
Make a price-cut for those good-willing consumer :)
👍
Gpu repair expert confused, can we mortals ever trust these board manufacturers.
@@SergeiSugaroverdoseShuykov no, why cut price because AMD make a shit card?
@@Malc180s it has nothing to amd if you’d listen him you would know that
Yes, another video to remove your doubt.
Thanks for the follow up. It is nice to see the methodical approach and perhaps we all learn something about AMD cards in the process.
Awesome channel my friend 🙂 so much information, you rule!
your voice is basically therapeutic and the work is very interesting. good stuff lol
looking forward to future videos
It would be interesting to test the core on another PCB if possible, but I doubt you have another PCB to test this
I don't have any.
Thank you for the great videos.
we will be here, no worries
Hey Tony.. good work to try out more options. But I think, if there was a separate jig/ device to test the GPU Core data and control lines, would have helped in such sticky scenarios! But I believe such devices will not be cheap anyways!
am sure it wont but i dont know if one exist.
Love the video. Keep up the good work.
you never give up. do ya?
I try not to.
My own curiosity usually outweighs customers expectations
💪
You know what a chip is old when it grows out of it's pads.
Awesome video
Thanks bro 😁
. . . A lot of this solder - bally stuff was inherited from toy console computers pushed way back when seriousl issues were pretty well ignored. . .
A lot of it seems to be the sheer size of these video cards, the PCB's inevitably flex more and you tend to find this damage to BGA components in the same location near the bottom of the card. The damage might even occur during first insertion into the slot if you're not extremely careful, if not the weight of the card and the sag that results when horizontally mounting seems to almost inevitably cause damage to BGA pads.
i wonder if the memory batch could be some kind of a snowflake that simply broke down with heating. is there any low melting aloy solder that isn't brittle as f like indium balls or something wierd like that so the memory could flex a bit without delaminating pads on either side? so you could place new memory chips one by one and test if the error propagate or if it just somehow start working but if not you could recover those chips without burning them
I don't think so.
I already moved some of the chips around but it had no effect
with the amount of corrosion on this card it was probably produced during the pandemic era and stored or mined near the Asian coast in humid salty conditions. I've seen hundreds of cards like this, even "new/old stock ones". If I was buying anything AMD I would stick to Saphire branded cards. But this is just my opinion.
Sapphire or Powercolor generally yeah. Those two would be your EVGA and Asus of AMD even though Asus makes AMD cards.
I bought mine direct from AMD to avoid any issues. I know everyone doesn't have the cash for a new card, but buying direct is a great way to avoid this. I would have returned the GPU if i t showed up like this.
Msi is cheap crap, but the ones where they stick on a good cooler are decent. Don't expect good quality chips, though.@@Azureskies01
The 7900 XTX was introduced in December 2022 though.
@@BrunodeSouzaLino It was released in 2022 but was being made for months and months before then.
The saga continues
So... the chiplet wasn't the problem, after all?!
Anyhow, chiplets are here to stay, whether we like it or not. Personally, I've no problem with them, so long as it means lower prices (at least to a certain extent) and/or larger stock quantities due to (much) improved yields.
After all, yields start to get a little iffy once you move past 350 sq-mm or so, regardless of the fab or overall yields. That's one reason there's usually a massive price gap between mid-range and high-end GPU SKUs and the price-to-performance ratio plummets to the ground the higher you go in the GPU hierarchy.
I never said that.
It may still be.
I won't know until I get the memory chips.
But I think it won't make a difference.
@@northwestrepair What about soldering chips one by one and looking at how output of the test software changes? That might help deduce if it's a discrepancy between software and hardware channel numbers or not.
@@northwestrepair the question is if teh software is chiplet aware - otherwise if its a virtual mapping transparently to teh card thru the MCDs then youd need anotehr softeare that is chiplet aware - as that what could casue the faulty channels to shift depending on faulty chips / combo.
@@northwestrepairIf I may ask - Why don't you have a stock of various memories, they will come in handy sooner or later?
P.S. Thanks for the videos, they're awesome 🙂👍
@@mareck6946 - There's nothing specific to be "aware" of, from the software's point of view. It doesn't care if the modules were made separately and then joined together or if they were all made at the same time from the same piece of silicon.
It's like the difference between having a single 10 metre power cable or two 5-metre cables connected to each other. The connector might break, just like the cable itself might break, but the device being powered by that cable can't tell the difference.
And it's not like a normal repairperson can repair the chiplet connections (or the connections inside a single chip - those would be even harder), so it doesn't make any difference for repairs, either.
The most important step would be to put the GPU on a working card (or put a known working GPU on this card), to determine if the damage affected the GPU itself or just the card. If it's the GPU, you can't fix it with normal tools (regardless of it being monolithic or not). If it's the card, it might be fixable.
But I think Tony has already said he doesn't have another card or another of these GPUs to test.
This is the spirit
Too funny 😂😂😂. Brilliant work 😎😎😎
Good luck with that, it's a long shot though....
Ridiculous how fragile those pads are i had no idea, it's no wonder they get damaged from drops
Hynix and Samsung memory end up having pads ripped all the time. But it can be a good thing because in this case, PCB is saved.
it's a shame they use that aweful glue , they need to change to something else Thank you for all of the hardwork you do, and for the video🧔🤠😎
Low-quality memory chips whose pads rip off? That's new. Usually we get shown pads ripping off the board.
As for the working card at the end, did you try each of the outputs? Maybe the one you use is good, but the one the customer uses is faulty.
i know it that him will try to fix it im sure in 20-30day this dude will post a video and that gpu will be working 😊
Right good work 😅🎉
Thanks!
Oh yeeeah!
Here goes the creative research!
You never know unless you reverse-engineer it!
It does seem like those memory chips break pretty easy.
VIK-ON worked on a 7900XTX and his memory layout on his pictures looked different then your's ...confusing as heck
Manufacturers have freedom to design their own cards with different layouts. Asus (for example) often makes two card models with different layouts for each GPU (a smaller one with just 1 or 2 fans, close to default clock, and a bigger one, with 3 fans, with higher overclocks). I think Sapphire also has multiple models for some GPUs. So you do need to get a boardview for the correct model, you can't just assume they're identical.
I agree regarding fan placements, etc ....but not memory layout.@@RFC3514
Fan layouts are different yes.... but usually memory layout is not manufacturer specific @@RFC3514
You're definitely onto something here. Memory chips can only be routed one way due to signal integrity issues, and VIK-ON and NWR's cores are rotated the same way. I think @northwestrepair has a problem with his memory layout, especially since VIK-ON's chart shows H channel right next to the PCIE hook, where impact damage is the most common.
Thats interesting and confusing because my layout came from boardview file, not VIK-ON.
So i am not sure.
I've been using AMD for a long time, so I was kind of surprised when you ripped AMD as hard as you did over their chiplet design. Never had issues with anything really so far, aside from general GPU tinkering from time to time because every GPU has its quirks.
So I'm glad to see you giving this another shot, even though you have some really hardline opinions about the card.
Not sure if it will matter too much to you, but I tend to only buy from Sapphire, except when it's a workstation card. But Sapphire makes those too I understand... if I am correct, AMD has all their own in-house cards manufactured by Sapphire; and Sapphire just makes their own branded models as well.
The only AMD card I ever really had trouble with, technically wasn't even an 'AMD' card in the first place. It was an old ATi card, manufactured by Asus. Everything else has worked to spec, for the most part, with some tweaks or repastes needed due to shoddy paste application on the line. (A result of basically almost everyone doing it wrong for a long, long time, more so than any one manufacturers fault. Also cheap paste is cheap paste, it needs be replaced.)
So again, glad to see you giving this one a second shot at least. And apparently even a third shot, heh...
One last thing though occurs to me during this comment and watching at the same time...
You mention the program might be glitched or something. I am remembering some other video, done by you? or someone else... where the order of the chips got reversed, and so things that were reporting as bad were not testing as bad like you are with checking the data lines. The resulting flipping around of the errors, perhaps, is because you did inadvertantly fix the original chips by happenstance, but another one was faulty still anyways so it jumped to showing that one in error now. But because you are focused on the wrong chips maybe? You keep getting sent back and forth between the 'issues'.
I could of course be entirely wrong, but if it helps, cool.
Can you talk about upgrading Vram on cards if it's possible
The Furmark is reporting a 6700XT?
❤❤❤
Your bias against AMD is getting the better of you.
I've noticed this too
My bias against AMD is based on what I see.
I don't see much good. 7900xtx is the worst I've seen.
Bias formed from experience.
@@northwestrepair are all 7900xtx the same i have the msi trio classic version
I love my 7900xtx nitro to be fair. Best card I've ever owned. But I will take a guess that if it breaks I will hate it :)
🌠Raise now,you tarnished amd🌠meet a god who granted you life
"REVENANT" from gpu fixing
❤
How about changing the whole batch of vram
Why not buy a better card?
Hello do you want to share Mats mode fromnamd and nvidia?
I must have missed something but it says in your monitor software that it's a 6700 XT?
He's testing a different card
Try something vram heavy on the powercolor
i did. no change. He said it was no detect and also shorting something.
@@northwestrepair why is running only at PCIe x8 3.0?
@@Nucleoprotein AMD.
Normal for this card.
@@Nucleoproteinhis cpu is old
@@northwestrepair definitely not normal for a 6700xt, its a 16 lane card, you should check it (rx 6600 is an 8 lane card on other hand), maybe the dataline capacitors are not in good shape? that would also explain the 'no detect' from the customer? who knows
bing chilling
You had a glitch in the matrix not the software.
6700 xt . what the life time of gpu maximum and the average and when the temprature about less than 65 celcius because i have gpu from 5 years and work correctly
Pretty long actually, there is still plenty of 10 years old cards working just fine. It will be outdated long before card develops some problems.
@@alexturnbackthearmy1907 thank you so much
@@sultanalharbi9182 Just be careful when you leave it in box somewhere for a long time, electronics dont like being stored for years (especially capacitors!), using it once every few months should reduce chances of developing problems greatly. It is never zero however...even on a new, out of the box card.
@@alexturnbackthearmy1907 this so true,i've recently took my zx48 from years of slumber and now i have lost 32kb of memorybuut,i can still play Jet Pac and,its still awe
Isn't PowerColor one of the top brands for AMD GPUs? I've had 2 of their GPUs fail on me within 4 months. Both RX 7900 XTX to be specific.
no, it's not. it's an "emerging" brand for AMD.
if you want top brands: Sapphire, XFX, Asus, MSI, Gigabyte, etc.
I'm not sure if I would call _any_ brand a "top" brand, but PowerColor I would say is a cheaper one. Sapphire is decent, as are Zotac, and XFX or, to be more honest, they used to be widely considered decent. Things change, for the better or worse
@@GeorgeTsiros I had their Red Devil variant as my 2nd GPU (1st one was the reference card that had a vapor chamber defect). It was one of the limited edition ones that was #2300 out of 3000. It should have been better than the rest. I was really hoping they did something special but it was just another dumpster fire GPU
@@leonidastankiangaming Was it something with the board or the cooling? Did you try adjusting/fixing it? I have a vega56 (sapphire vega pulse) and it desperately needed cooling. The only waterblock out there was bykski's but even that was _not_ officially released. You will not find it listed on any official bykski site. It was badly made: it has a step on the coldplate, it is not flat.
@@GeorgeTsiros I did everything I could to reduce temps and power draw. I reduced the voltage, underclocked, and even adjusted the fan curves. The crazy thing is at idle the Red Devil GPU ended up running as hot as 60°c while using more than 100w doing absolutely nothing with fans blasting at over 2000 rpm. Under load, 105°c core temp with hotspot way exceeding that temp. it was using up to 600 w and couldn't go past 17.89 FPS in the Cyberpunk benchmark at 1080p without RT and without FSR. It gave the same results with Heaven Benchmark and Superposition Benchmark and MSI Kombuster. No luck. I swear Radeon 6000 Series and RDNA2 are more stable and reliable than Radeon 7000 Series and RDNA3.
Hi, I bought and RX 590 month ago for around 100$, it was working all fine until yesterday, while updating drivers, the screen went complete orange and pc crashed or something, then I started having either pink or green screens when I turn my PC on, and its like that since then, sometimes I see display for 1~5 minutes but then it crashes, the GPU was all fine before, its temprature was good, never crossed 75 in extreme usage, and I also didnt got any burned smell.
Can you please tell me if you've encountered something similar in repairing gpus and if its repairable and if you have made a video on how you've fixed it, because I am from Pakistan so I can't send it overseas for repairs.
Wipe the driver and reinstall previous version. Perhaps windows took liberties and updated your driver itself. Try booting a linux bootable, manjaro is a good choice to test and make sure it's not windows related.
@@PineyJustice its GPU fault, I formatted my hard drives, so there are no windows installed, yet I get green,pink screen even in bios.
@@TheReaper-xl7qp Could be monitor cable, could be custom color settings in driver or it could be the card was flashed with a mining bios. If the card is dual bios, try the other, try replacing the hdmi/dp cable and try moving it to another port on the card.
I never used an AMD card, probably because i am not used to the alternative, but this videos teaches me reality and experience is often the better choice. Hopefully i dont need to replace my graphics card unless i need to.
If you're basing your decision about which card to buy on repair videos showing cards that suffered physical damage, you're doing it wrong (unless, of course, you're planning to hit your graphics cards with a hammer).
There are plenty of channels (Gamers' Nexus, Hardware Unboxed, etc.) showing the actual strengths and weaknesses of each model. >95% of them will never be sent for repair (either they fail within warranty or people upgrade after a few years).
Personally, I'm not a big fan of any GPU in the current generation, they're all overpriced and / or underpowered. The only ones that come close to being good value for money are Intel's cards (but their drivers are still very "green"), the 7900 GRE, the 7800 XT, and (in a few situations) the 7600 XT (enough memory but not enough bandwidth) and the 4070 Super (enough bandwidth but not enough memory). Even those are overpriced, but not as much as most other models.
Hopefully, the next generation will be significantly better, or at least have sane prices for the real performance they deliver.
@@RFC3514 I am still a newcomer to the world of graphics card, but as of now, my reality is surrounded by people and contacts having experiences with NVIDIA graphics cards. Mine is an RTX 3090 EKWB. But i do agree, this generation of graphics cards is better avoided, where both sides have their respective problems.
@@noxxiisproject2075 - Oh, the main problem is the same on both sides: greed.
This generation is barely any faster than the previous one, so they released drivers that do frame interpolation (i.e. create a "fake" frame between each two real 3D rendered frames) to pretend they're twice as fast. I wonder which of them will be the first to add _two_ interpolated frames between real ones, to say they're _three_ times as fast.
At least AMD tends to make open standards that any company can use, while Nvidia tries to lock people into their products, like Apple.
@@RFC3514 Barely any faster? I've got a 7900xtx and a 6950xt in different pcs on either side of my desk. The 7900xtx is a solid 40-50% uplift in performance while using less power and costing less. Infact, the 6950xt was a refresh that cost 1250$ a year after the 6900xt (real previous gen) came out at 1000$. So it costs the same as the 2 year old card and is 50% faster. Comparing it to the mid gen refresh card in performance and comparing price against the clearance sale prices of previous gen is silly.
@@PineyJustice - Launch price of the 6950 XT was $1099, not $1250. And one year ago (when both were officially in production at the same time) it was $599 (official AMD price), while the 7900 XTX was $999 (also official AMD price). That was 66% more money for 40% (on a good day) better performance.
Also, the 7900 XTX doesn't use less power. It uses less power _to render the same number of frames,_ but if you want that 40% increase in frame rate (i.e., under full load), it actually consumes slightly _more_ power (though not 40% more). It also consumes more power when idle.
Drop down one "level" (i.e., 7800 vs. 6800) and the performance vs. price ratio becomes even more in favour of the previous gen.
AMD's current cards aren't _quite_ as overpriced as Nvidia's, but they're still 20-30% more expensive than they should be, for the relative performance (or 20-30% slower than they should be for the price, if you prefer).
Auto-like
maybe restart PC will help ?
yay first
AMD ❤❤❤
oten
I am also dying to know. By the time you die Elon will be able to reball you and youre good to go. :P
dead rx 7900 xtx ? .....or 6700 xt ?
i m confused....
6700 came in for warranty but it had nothing to do with it.
You are confused because you didnt watch the 7900xtx repair video earlier.
I think you need those chips to do the owner justice to really know and check memery rails for any faulty components
.
Shows a bunch of impossible to repair defects for a vast majority of users/owners...
Comment section: "see you are just being biased. AMD makes awesome products!!!"
Never change internet😂😂😂
Said "defects" actually being damage from dropping and overall shipping the card...
@@emilbotev3656 - But everyone knows that when you drop a card with an AMD chip on it, that's AMD's fault. In fact, when you drop a card with an Nvidia chip on it, that's probably also AMD's fault.
Most cards he shows have defects that can't be repaired by the average user, the average user can't even check power lines,let alone fix many things he records regularly, such as when Nvidia pcb cracks under its own weight and he has to dremel layers and solder super thin wires, or heavy corrosion, or when he ends up with ripped pads whenever he has to reball the core and memory, etc. That's why they sent it to an specialized technician. The difference is he doesn't bitch and moan with Nvidia like he does with AMD.
He is killing those memory chips. So not cool
I been hearing nothing but bad things from 7900xtx, that it runs way too hot and has coil whine
Both issues can be fixed with an undervolt. Exactly what I did with my 3080Ti, dropped 50w of load, never exceed 75c anymore and coil whine almost gone
I don't have either of those issues.
I think it depends on the model
my 7900xtx has neither of these issues
I have a sapphire nitro+ with +15pl and have around 78hotspot max most of the time then running to hot is due to GPU thermal paste pump out but mine has no coil wine my friends did until he swapped his psu to a 1000watt
The russian guy can repair torn pad on memory chip
It comes down to quality control, and China doesn't have any!.
No more blaming the chiplets alright?
can you upload rx6600 eagle boardview on gdrive? 🥲🙏
well,the calc for economy is paying offwhen pads gets ripped its time to buy a newas often the pads gets ripped more new has been sellin'and we buyin'
❤
😎👍