Will Zen 6 Work On AM5? EKWB Implosion? 4090 Connectors Still Melting? April Q&A [Part 1]
Вставка
- Опубліковано 31 тра 2024
- Thermal Grizzly: www.thermal-grizzly.com/en/kr...
Support us on Patreon: / hardwareunboxed
Join us on Floatplane: www.floatplane.com/channel/Ha...
Buy relevant products from Amazon, Newegg and others below:
AMD Ryzen 7 7800X3D: geni.us/FxBWD
AMD Ryzen 9 7950X3D: geni.us/BV9I9nH
AMD Ryzen 5 7600: geni.us/qGJVbZ7
AMD Ryzen 5 7600X: geni.us/Hn7m9
AMD Ryzen 7 7700: geni.us/6VP8wD
AMD Ryzen 7 7700X: geni.us/fAfH
AMD Ryzen 9 7900: geni.us/RRL3Fa
AMD Ryzen 9 7900X: geni.us/Rjlq
AMD Ryzen 9 7950X: geni.us/lTmCsO
Intel Core i3-13100: geni.us/DfDsan
Intel Core i5-13400: geni.us/lQZFpt
Intel Core i5-13500: geni.us/goRuq2
Intel Core i5-13600K: geni.us/lJgCa
Intel Core i7-13700K: geni.us/Gpuqh3
Intel Core i9-13900K: geni.us/AieGiL
00:00 - Welcome to Hardware Unboxed
00:34 - Ad Spot
01:14 - Our thoughts on the 12VHPWR melting issues?
09:34 - If AM4 is the greatest platform ever, what’s second?
13:40 - What does Zen 6 need to justify a change in platform?
19:06 - What do we think about the EKWB situation?
24:24 - Should developers prioritise XeSS over FSR?
26:27 - Will AI prevent performance improvements in new CPUs?
28:58 - Why aren’t people talking about Frame Gen latency?
Our Thoughts On The EK Situation, AMD’s Zen 6 on AM5? April Q&A [Part 1]
Disclaimer: Any pricing information shown or mentioned in this video was accurate at the time of video production, and may have since changed
Disclosure: As an Amazon Associate we earn from qualifying purchases. We may also earn a commission on some sales made through other store links
FOLLOW US IN THESE PLACES FOR UPDATES
Twitter - / hardwareunboxed
Facebook - / hardwareunboxed
Instagram - / hardwareunboxed
Outro music by David Vonk/DaJaVo - Наука та технологія
![Huge AMD Zen 5 Increase Soon? Should You Still Buy AM4? April Q&A [Part 2]](http://i.ytimg.com/vi/EWWSyA4UV3E/mqdefault.jpg)
![Huge AMD Zen 5 Increase Soon? Should You Still Buy AM4? April Q&A [Part 2]](/img/tr.png)
If you’re talking about Northridge Fix, he has a contract with multiple manufacturers to fix their melted 4090s. Thats why he gets so many. He has mentioned it before.
Yep
i THOUGHT i'd heard him say this, but i wasn't confident. surely there are those accursed NDAs in place now 🤬
Mines been fine for over a year now :o
The CableMod adapter melting issue was when I first heard Northridge mention it on a stream. He supposedly got all of the cards that CableMod had issues with to repair. Not sure what he's saying recently as I stopped watching him as I find him to be a bit of a blowhard personally. That 200 number is absurd.
The similarly named @northwestrepair is my guy for these sorts of videos now, and he even has some where he is fixing the aftermath of Northridge's bad work lmao.
He's literally using the melting thing as a marketing scheme for his repair shop and to ride NVIDIA hate like many content creators. If the melting issue was such a problem, the entire world would be talking about it every day instead of 1 guy. All of cable mod repairs go to him, and they've had a huge amount of melted adapters, the guy is basically living in a melted adapter bubble from an issue caused by 3rd parties. Every since the gamer nexus video, there's been 2 reports of melting beyond cable mod reports. And all the recent news in the last year have been from this one dude. When they've sold probably more than 500,000 units, even 200 out of that amount minus cablemod is questionable to user error. How do we even know if that box doesn't contain burnt adapters since 2022 when the 4090 launched?
To provide an alternative data point on the 12VHPWR connector melting issue, I work for a small computer store that has sold just over 450 RTX4090s since launch, and we have had a grand total of 3 RMAs with burnt connectors, two of which were the result of the customer using the since recalled CableMod adapters.
Our data suggests that the percentage of 4090 buyers globally who experience their 12VHPWR connectors melting is likely extremely small.
To be clear, we have received more RTX 4090s for RMA, but with different faults.
In general, would you say it's a GPU with above average, below average or plain average numbers of issues?
My brother has an AM4 board. he upgraded from a Phenom II x4 955 to a 2600k which had had for a short while. Then I helped him with a full Ryzen 7 2700x/GTX 1080ti upgrade at the end of 2018. He's now rolling with the 5800x3d and a RX 6900xt and loves it. He couldn't give 2 shts about Ray Tracing and Frame Generation, he just wants his games to run and not stutter.
Ray tracing sure. But frame gen is something that, if perfected, will help greatly with "games running". I first got acquainted with it in microsoft flight sim. The fps gain was staggering with little to no visual degradation whatsoever (and if there was, to me it was negligible). I really hope AMD figures out framegen and fsr as that seems to be the the future we're going for. Nobody cares about optimizing games anymore.
@@PCgamer238 They did - the nit picking they do over fsr over dlss requires them to zoom in with a microscope - dlss might be 'better' but it's hardly a 'this is potato and this is crystal' type of thing - it's more like 'this steak is medium not medium rare and I like medium rare' - both are steak - both are edible.
@@xThemadpoetBOTH ARE BLURRY
and oversharpened
@@nuffsaid7759 yes - I have nvidia on my laptop and amd on my desktop - neither are great - they are 'acceptable' to run native resolution with a better framerate if your card can't handle it - that's it - neither of them are 'omg I want to turn these on' tech. Just like ray tracing - yeah it's neat, but it sucks performance so hard why bother unless you are taking screenshots.
(Old person answer) - Socket 3 was the best! - started out with 25 and 33 MHz 486DX's. Scaled down to 16 MHz, scaled up to 100-133-150-160 (easy OC) MHz official processors from Intel, AMD, and many other vendors. There were Pentium overdrive chips that brought next gen performance to this socket. You also the platform support ISA, EISA, VLBUS, and PCI over time. You could have bought a high end 486 in late 1989 with this, and in 1995 upgraded to a 5x86-133 MHz via an adapter (or a Pentium Overdrive), and been good for a few more years after that. The platform started with DOS and ended up being able to run Windows 98..
Socket 3 was good but i recon socket 7 beat it... remember the good old cheap Cyrix 6x86, IBM, etc. AMD and Intel all ran from it as well.. Will never see those days again
Haha, I was hoping someone put in some old person answers. Not as old, but my mind went straight to Socket A. It was so good to see AMD compete with Intel again, and the overclocking was fantastic. Memories of the pencil trick.. overclocking was such a community vibe. Good times!
@@raif1029 Was also cool since they used it for both desktop and mobile so you could use multiplier unlocked mobile cpus in desktop mobos
@@BrockSamsonTV You’re right - Socket 5/7 were pretty amazing for longevity too! The platform even later gained USB and AGP
Socket 3 was also what immediately came to my mind.
northride fix mentions in his videos that hes working with factories, so its safe to assume that manufactores outsource their repairs to them which is how they get so many
Good info to know, makes more sense for sure.
Northridge is the repair company that does all of Cablemod's repairs, so all the repairs related to the faulty angled adapters were through there.
@@Hardwareunboxed Northridge has been doing 4090 RMA repairs for CableMod for a while
Still a garbage design name another pc conector that it alone employes workers just to fix
Also the cablemods adapter voids warranty.
Second-best platform ever is Socket 7 because you could use CPUs from Intel, AMD and Cyrix on it.
I was looking for this comment. Socket 7 was great. I remember feeling a little down when they started into the proprietary slot architectures.
IDT (Winchip) and Rise (MP6 and idragon) also made CPUs for Socket 7
HUB: "Name a company that doesn't make an AIO"
Me: "Easy. Noctua."
... or Mazda. I haven't watched far enough to hit that question but... McDonald's.
Every AIO is a Noctua AIO once you replace the fans, that's just less necessary these days.
@@kloroformd Do car radiators count as an AIO?
@@john_hunter_they are 100% custom loop
@@john_hunter_my vw doesn't have a radiator. Check mate.
AMD has a surprisingly clear rule. The AM socket determines the type of memory used - AM2 - DDR2, AM3 - DDR3 and so on, so AM6 is basically guaranteed to use DDR6 or an equivalent jump in memory technology
Wonder if we'll see another AM#+ socket if DDR6 or another future standard takes too long to get to market
Remember when AsRock used to release boards that supported two different generations of ram on one board? pepperidge farm remembers.
@@Collin_J AM#+ isn't coming back as long as Lisa Su is CEO.
@@Rose.Of.HizakiAsus had an experimental board that supports DDR4 and DDR5, I saw it in a CES video, was with Intel 12th gen launch I guess, but not sure if they launched or was just a prototype
All rules can be broken. That is no exception.
I’d nominate Socket 7 and Super 7 as the second best platform of all time. The Socket 7 supported CPU’s from many different manufacturers - Intel, AMD Cyrix, IDT. Each had multiple processor generations on Socket 7.
It expanded the card connectors from IDE to AGP and improvements on the bus clock speed.
First time AMD started moving the market, instead of just following Intel.
NGL, big respect to you guys for somewhat surviving this tech drought, I cant imagine having to think of making content when there is barely anything to make from
I jumped from 775 intel directly to AM4 (3800X -> 5800X -> 5800X3D. Now on AM5 (7800X3D) because i plan to use this for another 4 years. Intel is out of question until they sort out their new socket AND power consumption goes back to normal and not trying to sacrifice stability to stay competitive with AMD.
Gotta ask why u stayed a 775 socket so long
Same. I was donated an LGA775 motherboard with a Pentium D 930 in 2008, which I upgraded to a Core 2 Quad Q9650 a few years later. Still runs, but the motherboard can only handle 4GB of DDR2 RAM @800MHz (only has 2 DIMM slots). Built an X570 AM4 system in 2019 with a 3800X. Upgraded that to a 5950X a couple years later when it became available at MSRP. Still a very capable system with a 4070 GPU and good amount of fast RAM.
Just a reminder cuz I dont think a lot of people still grasp this - you dont have to buy the super top end Intel SKU's, and you can undervolt and lower clocks to massively improve power consumption, thermals and also have perfect stability. Intel's still a whole process node behind AMD at the moment, so will still be behind on power efficiency as a whole, but I dont think a lot of people understand that out-the-box TDP isn't some inherent characteristic of the chips, and this goes for GPU's as well.
I went from an i5 6500k to an r5 2600, to a 5600x, to AM5. The upgrade paths offered by AMD at a lower price point were what attracted me, the 7800x3d was an easy sell, I live 80 miles from the KC microcenter, so I just snagged it in a bundle when we went down there to go to a ramen shop.
@@maynardburgerThe issue for myself is that I don't buy parts to undervolting, downclock, overclock, etc.. I buy a part to put in my computer and play games and 90% of people are the same way. I CAN do it, but you always have to test and it affects different games, etc.. It's simply a bunch of work with no real benefit unless you have extremely high power rates, but then why get high powered gear. If I can just hit auto-undervolt, I simply won't bother. And in that case, the difference is minimal.
X58/LGA 1366 was the second best plattform imo. The xeons you could get for cheap, DDR3, and the chinese boards that popped up later down the line really made it appealing and viable for many years.
i really liked my 920 (mainly for x264 reasons), but i far more love the 1700X/1950X i got for $150/450 new a few Black Fridays ago (for encoding and distributed computing reasons). followed by AMD's generational improvements over Intel's fat, hot booty as of late 😂
2011 - 4 generation of xeons/hedt
Right now X99 is king.
Machinist X99 + Xeon 16XX v3 (which are unlocked). 100$ for an 8 core overclocked to 4.5GHz, 32GB+ of RAM with Ryzen 3700X performance
I used 1366 for years until I upgraded to AM4.
i built my first PC around this platform, and later on make a killing off buying old x58 systems that I could refurbish. Westmere and Gulftown processors, FTW
Side story. I bought one of these systems for $80us. It was a gem with an i7-920, Asus P6T Deluxe v2, a nice 1GB 4870, Antec 1200 case, and the old Corsair HX1000w. After we exchanged the funds, the sellers were gloating because they got the system for $30 from a yard sale. I then told them the system was potentially worth $500, after I made $150 worth of upgrades to the system. You should have seen how wide-eyed they got. And when it was all said and done, I made about $250 off this system.
Card RMA is a big fkin hassle for most people because people buy them online, some online shops do the dirty and claim you have to RMA it with the vendor directly, the vendor charges you shipping, take weeks to months then shithead companies like gigabyte slap orange stickers with arrows pointing to crap thats not even related to the problem and refuse to RMA it, then after some more weeks, send it back at your expense with inflated shipping cost.
at least it is here in NZ.
milage varies between brands and shops, but the gist of it is they will worm the hell out of it anyway they can and make it take forever.
as a prior worker in the cellphone industry, those Orange Arrows are usually legit. also, sorry to be the asshole, but NZ is a bit of an edge case for shipping costs. if you want better shipping rates, lobby your politicians for better UPU prices. i know the pain of having to ship a smol item (SCAM4-1000B) to another smol country, Lithuania. was over $17 for the smallest and lightest of packages (but the minimum allowed for a customs label)
Over here in the EU in the vast majority of cases the RMA/warranty claims are done through the retailer who is required by law to offer at least a two year warranty and who then deals with the vendor. I had to do this once and it was an extremely smooth experience as I just reported the issue, returned the product and got a replacement a few days later.
I'm guessing most people with burned out connectors in the EU will handle it this way meaning that repair shops won't see those cards.
Are you talking about a certain retailer with P and B in their name, if anybody does this in NZ they are breaking the law.
@@electricindigoball1244 Same as in NZ, retailer is required to support the RMA process, if people buy from overseas then that's on their heads!
@@electricindigoball1244
To be exact warranty is extra service. At least in Finland retailer has the responsibility to fix or compensate for 2-4 years even if the warranty is void. Assuming you didn't do something to break The product.
One deals with the manufacturer only when it's extended warranty (like 10 year cases that some PSU's had)
DDR6 is expected in 2026, but DDR5 was available in 2020 then it took another 2 years for AM5 to come out. Wouldn't surprise me if we see Zen6 in 2026 as the final CPU on the AM5 platform and then Zen 7 comes out on AM6 in 2028.
Northridge Fix has claimed in his videos that he’s working directly with the faulty video card manufacturers that are outsourcing the repair work to his shop that’s why he’s receiving so many. Obviously it wouldn’t make sense for someone who just purchased a fairly new video card that’s still under warranty to pay out of pocket and send it to a third party repair shop. My RTX 4090 Founders Edition has been working flawlessly so far, 🤞
Thanks Steve! Thanks Tim!
AMD's cpu substrate has been 40x40mm for over 20 years now since Athlon64 launched in Sept 2003, For context Intel was using Pentium4 at that time.
You can still use a Socket754 heatsink on an AM5 board and vice versa.
I love how I now have multiple cpu waterblocks that fit AMD processors. I keep upgrading because waterblocks keep getting better. Let's me handmedown blocks to my older builds.
i'm using a Ninja 4 on my AM5 (which looks like it hit TPU in 2015) 😍 however, this might be an end for * this * Scythe unless i magically master some fan curves to hit a 105 watt target. dat power/heat density is INTENSE
I’ve had a moddiy 12vhpwr cable on my 4090 since shortly after launch and have had no problems. Bought their right angle cable as soon as it popped up and again zero issues.
My 4090 connector melted 5 days ago after 18 months of use. The 12 vhpwr melted on the gpu and psu side.
What did you do with it? Warranty? Repair?
Oooh, you are together! In the same room 😮😂
LGA775 on paper was good but in reality you still had to buy new motherboards for newer CPUs.
LGA 2011-3 in my opinion is a contender for third place given how affordable it made getting a decently powerful computer for those without a good used market
You could get double socket boards, and the engineering sample CPUs were cheap as well. There are different grades of samples, so the ones close to production have all the features like temperature monitoring and throttling.
Energy consuming monsters though. They're terrible for gaming too.
The upgrades weren't there and microcode updates were. Now I have to bypass Windows requirements to OC enough to keep it viable. Perhaps now it's a decent platform but if you've had it since release it's not very exciting.
@@yensteel terrible for gaming? not really. those high core count ones hold their head above water with their large amounts of L3 cache. The i7-5960x on the other hand is surprisingly competitive when it's overclocked. Unlike the E5-1660 v3 or 1680 v3 that people bring up, the 5960x will easily handle 32GB 4x8GB of B-Die RAM at 3200MT/s with the CL pushed down to 14 or lower.
The dual-socket boards are not much to write home about unless you're a CPU miner or you needs a lot of cores for cheap. lga1366 and lga2011 has dual-socket boards too. These E5-2600 CPU are so damned cheap now that ES and QS chips are not even part of any discussion I've seen about this platform.
@@m8x425 Thanks a lot for your correction. I'll double check on the sources then rely on personal experience.
Looking online, the L3 cache of my 2 cpus were 30mb, compared to 8mb. Are there any good comparisons between it and devil's canyon?
Just curious but how do you guys source founders edition cards from Australia? I'm a Melbourne person as well and have always wanted one :(
didn't NVIDIA said it was user error therefore voiding waranty, or 1yr has passed and no waranty anymore.
My 4090 suprim liquid has 3 year warranty. It's the use of Cablemod's adapter that voids warranty prematurely. User error of plugging it in on mass scale wouldn't void the warranty. Although I am not sure how would vendors prove any of this.
That's claim of user error is bullcrap though. Fully seated cables are still melting.
OMG, the socket A was epic !!!
The original video tested what can cause the issue in minutes, but there are other factors that can cause it over a prolonged time. There are reports of them letting go 6 or 8 months down the line. The safety factor is only 1.1 on the connector, corrosion, dust, hair, anything gets in there and it only needs to cause over a 10% degradation in contact to start causing issues. Remember the pcie 8 pin can handle about 300w but when the 500w 295x2 came out, we saw some reports of melted connectors with 2 8 pin connectors. A brand new out of the box high end connector will be ok sure, but a few plug cycles, dust, corrosion, anything and that safety margin starts evaporating fast. Even a small amount of stress on a fully plugged connector can affect resistance, resistance is heat and as heat builds resistance goes up further. A situation that is fine may be fine until it isn't when it goes just over the edge and snowballs.
I have two 4090s overclocked to 600-700w for over a year with no issues.
What kind of connector do you use? I’ve been using the adaptor that came with the Gigabyte 4090 and had no problems for 2 years as well. Though I can’t help but notice that people who had melting connectors had 3rd party adapters.
@@towertooth24157 cheap Amazon adapters,one is a 90° one too.
My 4090 FE never goes beyond 450 watts, and got it during launch. Its almost 2 years old and my 12vhpwr plug is fine -- no symptoms of melting. 450 watts at a 4090 is 100% load -- the melting problem is definitely when you go above 450 watts.
Ever since I learned that putting it on 70% power limit means 33% less watts for 5% performance loss, I have been running mine like that. I don't even see 400 watts so its relatively safe imo.
If I recall correctly, CableMod was using a particular shop to repair all the cards that had melted connectors while using their adapters, which was a lot. That's one instance of a single shop reporting a lot of issues with melted connectors.
Yes Northridge fix. He has youtube channel.
Yeah I posted about it above. Won't paste it here as it might get flagged as spam
hah! that was exactly my progression. As the PC guy in the family, I tend to buy pretty heavily into platforms and stay there so all the builds have interchangeable parts for eventual troubleshooting. LGA775 - LGA1155- AM4.
We have sold around 220 PC's with RTX 4090's with only 1 melted connector. Obviously no issue with anything lower than a 4090
BUT, what is the power limit of those 4090's you sell? I bet the farm that none are over 530W.
Maybe you haven't experienced it but there are reports of 4080's also melting (Northridge Fix showed one he was repairing in a video). The 4090 is obviously more common but it has happened to the occasional 4080 as well.
AMD socket A was also a banger IMO! Those Athlons were a lot of fun to overclock!
glad to hear you guys coming forwards to make the statement that "its a flawed connector"... because this is indeed true. however to put a finer point on how situations are always going to arise its simply like this: sometimes there is a bug in software, and the card ramps up 100% and does not shut down. so in those circumstances it's a bit like a random stress test that the user is not aware of. then combine that with the eh the higher max power requirements of the 4090 sku + also the persistent issues around the plastics thermal cycling and slowly creeping / walking the plugs to get worked out at a very slow rate. so slow, in fact, that it's much more dangerous than any rapid pop out. since the ultra slow thermal creeping 'walking' of the plug ensures 100% that it will then encounter that razor thin danger melting zone. which gamers nexus found, but was so much more challenging to artificially reproduce. poorly proven design, poorly understood problem fault. and even poorer understandings from the management, the standards bodies, and the users themselves. good job it's not occurring too often. but here we have it: this particlular design should never have been promoted, encouraged, nor released. never ever. and while tweaking might help, then again fundamentally we cannot change the design very much from its original specs in order to maintain product compatibility. and i have been against this connector ever since they added the 4 sense pins. the original 12pin proprietary connecter seemed fine, as does the corsair psu new type 6 micro cables ones. but they had to mess about with it....
a card running 100% for forever should not have any hardware problem. That connector should be able to handle whatever the card or do user do.
Has anyone ever thought these people use crappy power strips no quality equipment.
People are cheap and nobody is talking about how the machine is plugged to the wall is sounds like a power surge vs bad connector 💯
@@tommypearson9260 if it was a power surge, your psu would shut down or be dead, but not melting a conector inside the PC... and this conectors melted even with the original nvidia adapter, dedicated 12vhp from the psu manufacturer and so on
Socket 7 is the second best imo. There were chips from the Pentium 75mHz all the way through to the K6-III at 450mHz, Six times the speed. Intel, AMD, IBM, Cyrix and a couple of other companies all made processors for that socket.
Buying a CPU right now for it's AI capabilities is like buying 20 series in 2018 for ray tracing.
I bought two 4090’s when they were launched and both have had zero problems. One in an itx case air cooled and the other in an EK water block.
Same here bought two but in my case 50% of them went kaput.
@@ivaylotsankov7292 that’s not good. Did you use a third party cable?
The amount of repair outsourcing is immense, I think Dell was one of the few computer hardware brands that didn't outsource much.
EKWB have a huge amount of Enterprise business, and they're not doomed but they will have a lot of brand recovery for their consumer side.
Additionally, according to Eddy from Alphacool the market is ever increasing in recent years, be it enterprise or consumer.
This is new to me. I have a Gigabyte 4090 and have had it for a while. I have replaced the cable since the day I got it since I did hear about this problem a while back. I have two different cables that can be used for it, one Corsair cable made for the wattage and one that actually has a right-angle built into it so that there is no strain put on the cable or card. Everything has been working for me perfectly and nothing running hot, or at least hotter than expected, even under load.
If cpus hvae AI accelerator, I hope it could be used to dynamically optimize what the cpu is being used for. For example, if games that heavily saturate 3D cache is being utilized, it could try to dynamically relocate every process to the vcache cores in CPUs with two CCD like 7950x3d. Maybe even AI-assisted branch predictors and voltage regulators (Basically automatic curve optimizer?).
No issues with my Zotac 4090 AMP Extreme Airo that I've had since release so well over a year and it's been running 24/7 (idle and load) with many hundreds of hours gaming and it's been flawless so far. I did remove the card a few months ago to upgrade my NVMe as I needed a larger size so I had to disconnect the connector and reinsert it but still no issues, checking the connector temperature it's pretty cool even under full load. This is being powered using a 3x 8 pin cablemod adapter extension lead as it's connected to a AX1600i PSU. Still don't like the connector and it needs changing, but people also need to make sure the connector is inserted perfectly.
For 2nd best CPU socket I have to go with Socket 7. Supported CPUs from 4 different manufacturers, split rail voltage all while being backwards compatible with Socket 5. AMD later introduced Super Socket 7 with 100MHz FSB support.
There were also more than two chipset manufacturers around.
17:21 Yes, 4 memory channels on the 12 & 16-core parts please! 🙏🏻
A thing to 24:24
Would it be possible to try XeSS on Intel vs DLSS vs FSR vs XeSS on other GPUs?
I remember Gamer Nexus. Recommended highly that the short cable going into the card. Should be plugged in and checked BEFORE the card was installed into the computer. As it was easy to not insert that cable properly when the card was installed
As a technician who works at a computer shop in New Zealand, we actually don’t get a lot of burnt 4090 surprisingly, couple of them through out this year but definitely won’t say that’s a lot. Wonder what’s going on with other people saying this online.
It's user error, it's obvious that a country with overall higher IQ on 4090 owners is going to experiment less 4090 burnt connector accidents.
Krisfix, which is in Germany, also mentioned that the amount of burnt 4090 connectors he gets is extremely low, and northwest repair does not have many burnt connector videos in his channel either.
Steve Gamers Nexus covered this angle adapter issue extremely thoroughly.
LGA775 for sure. I personally went from a single-core Pentium 4 / GeForce 6800 to a Pentium Dual-core / HD 5770 to a Core 2 Quad / GTX 660 from 2004 to 2012. Great platform.
Ive got both a 3 into one "Normal" cable and a 12vhb cable that Seasonic provided with my PSU. Cant decide if I should use the 12vhp. :~( Perhaps matey is just bloody good at repairing 4090s?
I have launch 4090 fe never burned. Only plugged 12vhpwr connector from Corsair one time in its life.
I hope for 50 series they go back to the other 8 pin connections.
I think three cables in the front look cool too 🤷♂️
LGA 1366 is on my list of greatest platforms. Just had so much going for it... first gen i7's, support for 6 core,12 thread xeons. PCIe lanes for days.... triple channel DDR3... didn't get much at all in the way of new cpus, but still had a heck of a lot of longevity.
As an early riser in California, I really appreciate the upload times
Socket 7 or supersocket 7 was second best or if not the best
Those were the days. It was the last time AMD and Intel used the same socket/MB.
Second best is LGA 775, you can use 771 xeons with a sticker and some micro code editing on your bios file. It had support forever, multiple DDR version support there where motherboards from the main stream that gave you both DDR2 and DDR3 slots, the price was extremly healthy and cost efficient. The motherboards themselves where often also open, meaning you can buy the second tier asus board flash the maximus bios and get all the overclocking features, since they have a smiliair vrm you wouldnt actually be far off the ultra expensive versions.
x79 is a platform I'd submit for consideration. my 3930k was honestly good right up until 2019 when the memory controller finally died. I bought a 12core ivy bridge xeon (for $43 no less) and still run it as a home lab.
I rolled with x79 for nearly 10 years. When the E5-1680 v2 became somewhat affordable, i got one to replace my 4930k, and the 4930k replaced my 3930k. I rolled with the 3930k for nearly 6 years. (I never did the NVMe BIOS mod, but I didn't need to either.) 4.6GHz was the sweep spot for the 1680 v2 I have and it would run flawlessly with 2400MT/s cl11 RAM.
Before the worldwide flu, I got hooked up with an x99 Rampage V Extreme for a good price. After using x99 in my other system for a while, I'd say its a step ahead in every way in comparison to x79. AVX2, native M.2 slot, and also having the option to overclock the Ring Bus really helps this platform. The only downside is Broadwell-E processors stink for overclocking, and you'd need a J-Batch 5960x to do any real overclocking on top of the 5960x eating power when overclocked.
Would you be able to post Q&A audio to your podcast feed? I would love to listen to these while on the go!
If memory serves, GN talked about 4090s and the Northridge shop in the past. I think Cable Mods(?) was sending a lot of the cards there at the time. Not sure if there is anything weird going on now.
Could gpus use the spring pin connection like fans aios cases have started using?
27:42 there are big benefits to running AI workloads locally instead of in the cloud. The two biggest ones are:
1. you aren't being dependant on the AI company maintaining their service (you're also immune to subscription price hikes if the local AI software you're using isn't subscription based).
2. you can use AI processing without sharing your or your company's data with the AI company.
Made it into the q&a boys, from here on out its going to be smooth sailing
Is that a thing?
Northridge Fix is like one of two places that I know of that will fix this type of problem. Is a great measure of the size the problem is.
Here's a question for you: I know it makes sense to wait for Zen 5, but if we were to assume that Zen 5 is expensive, is there still value in Zen 4? At the moment, I see that I can get a 7800X3D, 7900, or 7900X3D all for the same price ($500 CAD, or $370 US). My use-case is that I play single-player games at 4k, no ray-tracing (6950XT), but also run 3x 4k displays with youtube, Twitch, and OBS running at the same time. Is the 7900X3D the best match?
I think Socket 7 is the second-best socket only because it supported so many different flavors of processors from multiple company's... The good old VIA chipsets, etc... Starting to show my age now... :P
One of EKs biggest mistakes IMO was not moving to additive manufacturing, and moving the manufacturing in house. I use EK for most of my water cooling, but they missed some large market signals 6+ years ago.
I ran my 2500K up until mid 2019. Which I replaced with a AMD R5 3600.
1155 is definitely a contender, but I would go back much further to the 486 DX2 66. Probably not as much longevity as 1155, but I really loved that chip, and I ran it for about 5 years.
I've also had 6 different 4090's since launch (Gaming Trio, 2 Suprim X, Tuf, Strix) and my current one i've had for about 7 months in the Gaming OC SKU from Gigabyte. I've used one MSI A1000G pcie5 PSU the entire time and have used TONS of different native cables from angled cables from cablemod all the way to sketchy random 12vhpwr cables from Amazon. I've spent alot of time running them under full load as well, even pushing them with the power slider cranked to 600w running timespy benchmarks and I have never once ran into a single issue. Maybe I just jinxed myself. LOL
Also, love the Daily Bugle legos in the back I want it!
Sounds like you made sure they were plugged in al the way.
@@Apollo-Computers I always just gently push left to right until I hear a click, then do it a few more times for safe measure. To pull it out I just gently wiggle it left to right and it comes right out.
Representing the 2600k here. Bought it used 6-7 years ago, still doing some fair gaming, although with a few chuggs here and there lol. Considering jumping to 7800x3d before summer, but i might aswell wait a bit more now to see zen 5 first. whats a few months on a decade : p
The repairers seem to get the RMA'd cards from the companies they were returned to I believe...example cablemod covering melted connectors making cust whole and sending the replaced cards for repair to sell, or for future replacements
I got my MSI 4090 in November 2022 and so far original connector is fine
Love to see more PCIe lanes with the next socket. For someone that uses old parts after a gaming PC upgrade for their NAS, really hard to add an HBA + 10Gbe NIC and possibly a GPU. I won't be able to do that with my current 5900x rig, just not enough x8 slots :(
They could for example, increase the memory channel to 3, which would allow them to reasonably increase the max number of core without having the cpu to be largely memory limited. I'm guessing that they would also make it so that only the top tier chipset have all three channel and lower tier may have one maybe, but the socket still would need to be changed as the num of pin increase. It would also kinda move the custom desktop segment more towards prosumer market.
My guess on the number of burnt connector GPUs being in a third party shop is that cards sent in for an RMA are being replaced with new cards, and the bad cards are being sold bulk by the case or even by the pallet. These shops then fix the ones they can and put them out for sale on ebay or whatever. The ones they can't fic may be parted out for fans or whatever.
I love that look of Steve at 14:02, you have on in the shipping or already in the benchmark rig don't you?
Is there something done to solve the issue with the melting connectors? I did not see this being a problem with the previous style of connectors. So this „improvement“ clearly was not good.
Good Price to performance on new Plattform with DDR6?
Thats a bit gimmicki if we look back to 12th gen and DDR5 + the really bad scalling.
I'm using the amazon special cables too on my 4090 and having no issues..
40% lift would probably get people to jump early but it would still stink if it could be on the same socket. Agree with Tim, it would have to coincide with a memory change and to a less degree a PCIE change
Why dont you put a laser thermal tester onto the connectors and cables when under full, medium and low loads on your cards and let us know the results ?
Steve from Gamers Nexus already did plenty of testing like this on these connectors
Could you guys make a video comparing gaming performance on the 7800x3d vs 3900x with a 7900xt? I'm contemplating upgrading but not 100% sure if it would be worth the upgrade
Regarding the 4090 Connector, neither the Original connector that came with my unit nor CABLEMOD version resolved problems I had (red led on the GFX and black screens). I had to buy a third one from SEASONIC ( I have a SEASONIC prime 1000W) and that one finally fixed my issues. I think it was the PSU OEM's that mitigated the melt power problems with time though the current connector format is still not robust IMHO.
The robust solution is not to buy that crap. God help you if your job requires you buy these garbage GPUs that start fires.
@@nimrodery Not a single one has started a fire.
@@Safetytrousers and yet I still don't hire people holding torches while grinning
I still have a Dell Optiplex 780 DT sporting a Core2 Quad Q9650 and GTX 750 Ti, 16GB RAM, copper CPU cooler and an SSD with Windows 10. It boots up faster than my son's Ryzen 9 7900X, though I will concede that his rig does everything else just a tad better...thanks for the great content!
One extra variable though is that this repair shop has put themselves "out there" and recognized as a place that fixes this issue.
As an electronics engineer myself, the whole issue with the pins beats me. We use similar connectors from Molex (the brand) and Phoenix with much higher continuous currents without any problems.
So they must use either very low quality variants or combine pins that er not really well compatible with each other.
The channel's contacts say there is not an undue problem anymore.
Sure, but are they in applications where the connector is almost guaranteed to have a lot of lateral stress on it?
I always thought the issue was where they put the connector on the card, not the connector type itself.
@@wallywest2360 no, I just put down this comment for no reason? Obviously it's a very similar and comparable situation.
@@p_mouse8676 I see, your vague statement of "we use similar connectors" was obviously the same exact thing. Got it.
@@wallywest2360 there's absolutely nothing vague about what I said before.
Had an i7 920 on X58 in 2009, upgraded to a xeon 5675 for about 40USD in 2016 and only changed for an upgrade bundle couple years ago: 5600X, 32GB 3600 DDR4 with B450 tomahawk max 2 and 860 evo NVMe 400€. I'm still pondering if I should upgrade my 1080 gaming X for a 7900GRE since it has started to show limits recently.
People should do some research on Northridge fix before posting quotes from them. From reviews of the shop, i wouldnt take anything other than a broken phone screen there. So many customers talking about how dishonest the shop owner is. Ripping people off, lying about fixing things, lieing about pricing, lying about time frames. To me it sounds like someone just wanted some free advertisement for their shop.
Alex from Northridgefix works for cable mod so alot along with others come without being rma ed. That said I'm in Australia and have repaired 53 in last year, some local some oseas no particular brand. The problem is real, cracked / bent pcbs also real big problem, n those are normally refused rma. Northwestrepair in USA also does a few actually EDIT over 18 months
Dark Power 13 1000w is very good and has been rock solid.
I'd contend Intel's HEDT platforms are up there with AM4. X58 was my jam. Lots of cores, lots of overclocking, a couple generations of guaranteed CPU upgrades. Lots of IO. Was a pretty solid platform.
I can imagine an AM5+ if we end up with PCIe 6.0, for example, but like with AM2+/AM3, the product will work on both platforms.
My ROG 4090 OC LC is still live and kicking with the adapter ROG did provide with my GPU.
And I use Thor 1200w PSU, and I 90% of the time set my Power Target to 30% for my 4090 just in case... I have no issues what so ever with the connector, YET.
But I think cheap PSU's, poor connection to the GPU by the user and bad connector design is the issue.
I don't get why they made the 12VHPWR connector so that the sense pins are connected before the power pins as you plug it in. It should be the other way around, so that the sense pins only make a connection after the power pins are firmly connected.
FXs were and still are good CPUs.
I'm still running a box with a FX-8300 and it's ok.
Hey guys, I'm thinking about upgrading to AM5, but I don't know if I should wait and see if there will be a B750 chipset. What do you think it needs to bring to be justifiable?
Don't upgrade, if you have AM4 just stick with it until AM6 or even 7
think im gonna upgrade to an am5 build next month no doubt been debating intel or am5 but i think i want am5 i never had an intel rig but am5 the way to go
My first PC that I bought entirely myself, was an LGA775 with a Core 2 Duo E6850 with a Radeon HD 4870 512MB and 4GB of RAM
It ran amazingly on my 720p screen at the time. But it got killed by a lightning strike. "Everyone" had an LGA 775 CPU in my social group, even though none of us had the same CPU generation 😅 Ranging from Pentium 4 to Core 2 Extreme
I had to rma my 7900xtx through asrock, when I was approved to send it in to them they send you an info sheet of what damage they won't cover under warranty and if it was burned in any way it wouldn't have been covered. So I think the manufacturers are denying the rma's. People that would buy a 4090 are likely to watch tech channels and they want to get it fixed by a youtuber so that particular shop gets a skewed sample size.
I just got my RMA 4090 back last week from PCCG plus a refund for the cable mod adapter ( both purchased same order ). I would like to know if it was the cable mod adapter that caused why was I not contacted and told to stop using it? It took about 5 weeks for it to be returned. Anyway, its still in the box and I am not sure what to do now.
there where multiple warnings in the media from cablemod, to imidiatly stop using the conector
@@georgwarhead2801 what media, on some internet but thats it? So many didnt know about it and surely it was up to the retailers to contact their customers
BTW the repair Time on nvidia connector 20 mins n 4 bux for connector.
12VHPWR Adapter that came with my Asus tuf 4070 ti
super was so tight and hard to plug in and out to get it to seat properly I can see how some guys leave it loose. I'm sure it only happen to me and Nvidia didn't make thousands like mine. Went with the Seasonic cable, no problem fit perfectly.
i have a 4080 and a 4080 super in 2 different rigs, no melting issues, maybe that is just down to the fact they draw less power but it isnt an insignificant ammount of power at roughly 320 watts each.
I am currently fixing a RAM issue on a system i've built about a year ago with 13700k and a 4090. The connector isn't burnt... yet... but despite it being properly pushed in and locked, after a period of heavy gaming, like 3+ hours with Cyberpunk 2077, it starts to smell a bit of a hot plastic. I've inspected all the pins 1 by one both on the GPU and adapter side, they all look good, but the fact is that if i replace that GPU with my 6900XT, the smell in that system is gone and it's my rig that starts to smell after hours of gaming. I am also in the camp that the contact surface simply isn't enough for that amount of power draw. There is nothing wrong with the PCIe connectors an it's perfectly fine to go back to those.
New psu?
After two hrs of heavy load with my 3090 I start to smell plastic as well. It's coming from my new psu. It's faded away.
I distinctly remember using my Q6600 from when it released in '07 until it couldn't run Dragon Age Inquisition in '14, LGA775 is truly the 2nd best. But as HUB said, it may even been the GOAT, if I haven't seen how well AM4 improved when I upgraded from 3700x to 5800x3d. If my LGA775 lasted for 7 years, I don't see upgrading my AM4 until DDR6 compatible parts are available and cheap to buy, which may have my AM4 last for maybe 8/9 years total.