AMD says insufficient fluid in Radeon RX 7900 XTX vapor chamber causing overheating, offers...

[kiwigraeme]

The hurdles of running a business as big as this one. Stress. But division of labour helps out a bit.

OPM helps a lot to limit stress - compared to self-employed

Other Peoples Money - like poker until you have skin in the game

Same a Govt Departments - you can be stymied by some Ahole simply because they know they don't pay the lawyer fees .

Lots of Corp employees etc want to show great short term gains - get those kudos and bonuses - Even if they know it will all turn to ****. Same a bankers , traders know they are mostly immune to clawback - so why not go for bigger bonuses

 

[Bulllee]

Timing is everything but it's only the 6th of January 2023, a year ago was bad..very bad.

 

[waclark]

There you have it.

Hopefully the anti AMD bias from the media and influencers can find something else to over exaggerate about.

Over exaggerate? I hardly think so. You spend $1,000 on a GPU you don't expect it to heat throttle. Likewise, how are you going to feel when your $1,000 card dies prematurely because it's getting excessively hot? Also, how are they going to tell how many units are affected? This could be 1000s of cards and it's sure to hurt sales. I wouldn't touch an XTX right now if it was $500. The best outcome might be that AMD is forced to drop prices if the XTX doesn't sell well.

 

[punctualturn]

You spend $1,000 on a GPU you don't expect it to heat throttle.

Yet, none had died, can we say the same of the $2000+ 4090 that did die?

Likewise, how are you going to feel when your $1,000 card dies prematurely because it's getting excessively hot?

Unlike the flawed power plug that failed in protecting the now dead 4090s, none had die, since they throttle to a point that didnt kill themselves.

This could be 1000s of cards

Possibly and by the way, tell me one sector that hasn’t experienced a bad batch of components that was fixed with a recall or a simple rma/replacement?

. I wouldn't touch an XTX right now if it was $500.

Exaggerating much?

The best outcome might be that AMD is forced to drop prices if the XTX doesn't sell well.

Economy 101.

Hint, that also applies to nvidia.

 

[Mr Majestyk]

Maybe going forward they should also do a improved revision of the vapour chamber. Looks like it's not top notch cooler from derBauer's video, especially where the gpu support pillars are.

Was it confirmed they use water as the fluid. I hope it was water with say glycol at least.

 

[Nobina]

What I say is that the influencers conveniently never bother in pointing out their dirty tactics or how they always give nvidia a free pass.

So your Dear Leader Jensen can avenge you.

They did point it out, you just have some crazy expectations of what they should say. Nvidia does get marketed more because it's a more popular brand but you're going into a conspiracy theory when you say everyone is shoving Nvidia down everyone's throats as if they were paid shills.

And the picture you posted is a bad take at being funny, it's irrelevant to your or my posts.

 

[Axeia]

Only after enough people complained and it started blowing up on social media did AMD deign to even investigate, when this should have been caught on QC and immediately handled if it somehow made it to sale. It's very anti-consumer and I would never buy off AMD's website after this, and I'd be very, very cautious buying elsewhere too.

Sadly there's very few companies that operate this way nowadays. It depends on how big the percentage is if this should have been caught in QC.

You can't even fully trust reviews because how did this not show up from reviewers? AMD also has a history of sneaking bad drivers through, this was no different, yet "fans" want to discount issues while attacking another company.

This is different because drivers you can update and fix drivers and it's greatly overblown how bad their drivers are - I've used both NVIDIA and AMD for literal decades and I actually prefer AMDs software at the moment. No forced signup - lots of useful features in the software they provide.
Heck I don't even know even where there's bad drivers myth even comes from, it's been around for ages yet I have so few issues with either camp. The only weird one that took them way too long to fix was the mouse cursor corrupting itself in a multi-monitor setup and the black screens on Vega cards (but I haven't owned a Vega card myself).

Nvidia's GPUs are way overpriced but at least they work right. That's why they have the high premium as well as have increased market share. 20 years ago Nvidia barely existed if at all, while AMD and Intel were megacorps.

You're almost making NVIDIA come off as an underdog. For reference, here's an actual marketshare graph

So the last time AMD (or ATI at that point) had NVIDIA beat in markshare was 2005. For about one year ATI had NVIDIA beat because they had a better product. At any other point in time NVIDIA was doing better and knowing NVIDIA likely at higher profit margins.
So basically since AMD has owned ATI and thus the videocards are actually 'by AMD' (2006) NVIDIA has been the bigger company in the GPU market.

As for the 'at least they work' if the recent power connector fiasco wasn't bad enough and we're feeling nostalgic, let's go back to the NVIDIA 8000 series. Remember the bad solder with cards wrecking themselves and NVIDIA claiming it's due to no fault of theirs washing their hands clean of it leaving customers and partners out in the cold? It's when Apple stopped working with NVIDIA and hasn't worked with them since.

It's not some giant conspiracy, it's not mind control, it's just reality and the market rationally responding to the information and track record it has. Yes crypto inflated things, but that's an ancillary factor, that has now run its course anyway. In the long term a consumer has always been better off with an Intel/Nvidia build. AMD is not doing what it needs to change that, especially on the GPU front. Intel is beginning to stumble so there's an opportunity, Nvidia prices insanely so there's an opportunity, but AMD is not a mom and pop outfit and it keeps shooting itself in the foot. That company needs to look carefully at itself and start cleaning out clearly underperforming executives and departments. Lisa Su saved the company but I'd bet anything a lot of the rot is still there and she didn't get rid of them. New regimes usually do a wholesale restructure for this very reason.

I'll just re-quote that again "In the long term a consumer has always been better off with an Intel/Nvidia build."
Excuse me? Always? That's laughable.
When Intel was pushing more Ghz out of the prescott architecture my AMD 64 3000+ was EXCELLENT value for money. My god was that a good purchase. Probably my favourite CPU purchase alongside the Intel i5 2500K.
I have absolutely no regrets buying an AMD Ryzen 1600-AF either and later on swapping it out for a Ryzen 7 5800X. At both of those points in time AMD clearly had the better deals.
It's the same with the GPUs where depending on when you bought a card AMD sometimes (often times, especially in the lower price segment) had a better deal. Especially as you put it "in the long run" because AMD tends to stick enough memory on graphics card that they fare a lot better after the market has moved on a few generations.

I've been buying computer parts for decades and never regretted a purchase*. This does mean shopping for whatever is the best at the time so I have had many Intel, AMD and NVIDIA systems. It's a tough market at the moment though mostly because prices for everything are insane.
Intel and AMD are very competetive at the moment, AMD could do a lot better by having more affordable motherboard options. As for graphics card.. both companies have gone insane with this pricing, the cryptoboom has passed and they come out with scalper prices. NVIDIA rightfully got stuck with a load of RTX 4080 cards (and likely 4070 TI cards as the actual market prices of those seem to be stupidly high - which is funny because even at the MSRP it's hardly a bargain).
This would have been the perfect time for AMD to finally regain marketshare that has just been fading away, they'll be in the single digit percentages soon. Instead they chose to be as greedy as NVIDIA - sure they charged a little bit less but on the other hand they also offer less as the raytracing performance isn't even close to NVIDIAs.

So imo as someone with a pretty good track record of buying the right thing at the right time - don't buy any graphics cards at the moment unless you spot a nice bargain on the used market. There's finally signs that the market has had it with these insane prices and it's time for both NVIDIA and AMD to learn a lesson because they're as bad as each other when it comes to pricing.

*Except for when WoW had just come out and there was some weird bug with NVIDIA nForce2 chipsets disconnecting you from the WoW servers rather frequently. That was not a great time! That's the one time I've regretted a tech purchase. Other than that the nForce2 was a fine chipset.

 

[toooooot]

nVidia poorly designed a physical connector that inherently lead itself to not being fully seated properly. Now that we know this it's an easy fix, just try to push a little harder. Fact of the matter is that it still poorly designed and many people, unaware of the issue, may still not fully seat the connector. But the most annoying thing about the 12 pin connector is that no one asked for it, no one needed it and for many years you're likely going to have to use an adapter. I doubt AMD will ever use it and I think many PSUs will continue to ship without it.

It's not like jumping from the 4 pin Molex to PCIe8pin connector. We saw the first major jump in power needs going from the 6800gtx to the 7800GTX leading to the pci-e6 pin on the 8800gtx and ultra. That was 17 years ago. My biggest criticism of the 12 pin isn't that it's physically bad a connector, it is why? What problem does it solve?

I don't mind their new connector. What I hate about it is that your sleeved cable extensions become obsolete. I don't want Nvidia's ugly connector hanging from the card.
Cablemod made their own connector but it costs 30 bucks for just one connector.
At the very least Nvidia could position the connector on the other side of the card where it is not visible.

 

[HardReset]

You're almost making NVIDIA come off as an underdog. For reference, here's an actual marketshare graph

So the last time AMD (or ATI at that point) had NVIDIA beat in markshare was 2005. For about one year ATI had NVIDIA beat because they had a better product. At any other point in time NVIDIA was doing better and knowing NVIDIA likely at higher profit margins.

For example compared to Nvidia's GeForce 400 series, AMD was doing much better. Basically AMD stopped competing seriously when they realized Nvidia sells better even if AMD has superior products. Another thing is that Nvidia does not have integrated GPU for PC computers, and that's why Intel is clear leader on GPU market share and Nvidia has very hard time to even match AMD GPU market share.

As for discrete cards, making better product has almost no effect when looking at that graph.

 

[waclark]

Yet, none had died, can we say the same of the $2000+ 4090 that did die?

Running electronics at high temps for extended periods will have a negative impact on their longevity. The difference here is that you can solve the 4090 problem easily, with no recall. Just plug it in properly. And, to some extent, the GPUs did "die". They throttled performance significantly. A $1,000 card running well below it's capability is pretty much dead.

Unlike the flawed power plug that failed in protecting the now dead 4090s, none had die, since they throttle to a point that didnt kill themselves.

The plug is not flawed. The user did not connect properly. Perhaps you can run a test on your XTX and run it at 110 for a few weeks and see how it fares?

Possibly and by the way, tell me one sector that hasn’t experienced a bad batch of components that was fixed with a recall or a simple rma/replacement?

Never said there wasn't. But, relative to the 4090 plug of which there were less than 100 impacted, this is much larger. Thousands. Which is contrary to your comment about over exaggerating the issue.

Exaggerating much?

Nope, but I see you're doing a good job of comparing a user error, impacting dozens of 4090s to a major product recall impacting thousands of cards.

Economy 101.

Hint, that also applies to nvidia.

It definitely applies to Nvidia, however, if the AMD cards aren't selling that means people are buying Nvidia, they may not be forced into price drops if that's the case. Ultimately, I expect to see price drops within the next 90 days. How much and what products remains to be seen. I doubt the 4090 will come down much in price, until a 4090Ti comes out. 4080s and maybe 4070s will likely come down at least $50-150 if Nvidia is smart, and I'm not so sure they are.

 

[yRaz]

The plug is not flawed. The user did not connect properly. Perhaps you can run a test on your XTX and run it at 110 for a few weeks and see how it fares?

This, sir, is what is called "false equivalence"

The plug is flawed, but so is AMD's cooler. either acknowledge both of those facts or shut up

 

[Burty117]

As for discrete cards, making better product has almost no effect when looking at that graph.

I'm probably of the minority, I build a new PC for myself every 5-6 years and when I do, I buy the best I can.

I went with a 4090 because it is head and shoulders better than anything else out right now. It just is.

If AMD made a faster card than the 4090, I would have bought it, no questions asked.

Just pointing out that actually, a better product would have sold me.

 

[HardReset]

I'm probably of the minority, I build a new PC for myself every 5-6 years and when I do, I buy the best I can.

I went with a 4090 because it is head and shoulders better than anything else out right now. It just is.

If AMD made a faster card than the 4090, I would have bought it, no questions asked.

Just pointing out that actually, a better product would have sold me.

Market shares are units (all equal) shipped. Meaning one shpped GT730 is equal to one shipped RTX4090. Since top end cards sell very low amounts, they also don't affect market share much. Also good to remember "discrete cards" also include mobile GPUs.

 

[yRaz]

I'm probably of the minority, I build a new PC for myself every 5-6 years and when I do, I buy the best I can.

I went with a 4090 because it is head and shoulders better than anything else out right now. It just is.

If AMD made a faster card than the 4090, I would have bought it, no questions asked.

Just pointing out that actually, a better product would have sold me.

Frankly, I'm holding out for my "Money is no object" PC build because I want 8k120. If I was to do that today, I would have gone with the 4090 aswell. It's a fantastic product and even if I disagree with nVidia as a company, it's just a great product. We can sit here all day and beat cats with dogs arguing, but the 4090 is hands down a fantastic product. Currently have a 6700XT but I'm waiting for the RDNA3 refresh to upgrade because 99% of the stuff I do, I do in Linux. NVIDIA has made progress on the linux driver side of things recently but not to the point where I would invest in their ecosystem. But that said, if a 5090 could do 8k120 I would drop the money and skip AMD. there would be a lot of finangling that I'd have to do to make things work in and out of VMs, but that's a whole other tangent I don't feel like writing about.

 

[m3tavision]

Running electronics at high temps for extended periods will have a negative impact on their longevity. The difference here is that you can solve the 4090 problem easily, with no recall. Just plug it in properly. And, to some extent, the GPUs did "die". They throttled performance significantly. A $1,000 card running well below it's capability is pretty much dead.

The plug is not flawed. The user did not connect properly. Perhaps you can run a test on your XTX and run it at 110 for a few weeks and see how it fares?

Never said there wasn't. But, relative to the 4090 plug of which there were less than 100 impacted, this is much larger. Thousands. Which is contrary to your comment about over exaggerating the issue.

Nope, but I see you're doing a good job of comparing a user error, impacting dozens of 4090s to a major product recall impacting thousands of cards.

It definitely applies to Nvidia, however, if the AMD cards aren't selling that means people are buying Nvidia, they may not be forced into price drops if that's the case. Ultimately, I expect to see price drops within the next 90 days. How much and what products remains to be seen. I doubt the 4090 will come down much in price, until a 4090Ti comes out. 4080s and maybe 4070s will likely come down at least $50-150 if Nvidia is smart, and I'm not so sure they are.

You are aware that it's a production problem... from one AIB. It does not effect all cards.

There is nothing wrong with XTX, or XT like you are seemingly going on hysterically about.

 

[waclark]

You are aware that it's a production problem... from one AIB. It does not effect all cards.

There is nothing wrong with XTX, or XT like you are seemingly going on hysterically about.

First no one is going on hysterically about it. Your attempts at insulting me don't make your argument any stronger. Yes, it's a production problem and a serious one for AMD. They are going to have to RMA/Replace or Refund a lot of cards compared to less than 100 Nvidia 4090s. The point is that people calling out AMD are not "exaggerating" the scope of the issue contrary to what NeoMorpheus feels. I have yet to hear any info that would indicate how widespread the issue is. Is it a specific batch of cards, some cards in the batch multiple cards from multiple batches or what? Given that it is unclear as to how to determine if a specific card is affected I would be hesitant to buy an XTX right now not knowing if it would need to be returned.

 

[waclark]

This, sir, is what is called "false equivalence"

The plug is flawed, but so is AMD's cooler. either acknowledge both of those facts or shut up

The plug is not flawed. When properly connected is works just fine. I would also say the cooler isn't "flawed", however there is a manufacturing defect that is causing cards to overheat and throttle. As for shutting up, well, that's not going to happen and if you don't like it, tough.

 

[HardReset]

Given that it is unclear as to how to determine if a specific card is affected I would be hesitant to buy an XTX right now not knowing if it would need to be returned.

It has been known long time that only cards using AMD reference cooler design Could be affected. Buy card with custom cooling solution and that exact problem will not exist.

The plug is not flawed. When properly connected is works just fine. I would also say the cooler isn't "flawed", however there is a manufacturing defect that is causing cards to overheat and throttle. As for shutting up, well, that's not going to happen and if you don't like it, tough.

Plug is indeed flawed if you can connect it "properly", that is, it basically works but still burns. It's pretty hard to plug PCIe connector way card recognises it but connector is still so loose it will burn in matter of weeks.

I remember years back when cards used peripheral connector (there was no PCIe connector yet) as extra power, but card connector was too symmetrical. In other words, you very easily could plug connector wrong way. No matter if there was instructions, it was user error etc etc, that was flawed design. You shouldn't be able to plug connector wrong way and even if you do, it should not work at all.

 

[yRaz]

The plug is not flawed. When properly connected is works just fine. I would also say the cooler isn't "flawed", however there is a manufacturing defect that is causing cards to overheat and throttle. As for shutting up, well, that's not going to happen and if you don't like it, tough.

The point of a connector is to eliminate user error, it failed in that function. They also have 4 times the power going into something with the same area as an 8pin connector. That's a horrible idea for thermals

 

[Burty117]

Say you've never used the new 12VHPWR connector, without saying you've never used the 12VHPWR connector...

 

[yRaz]

Say you've never used the new 12VHPWR connector, without saying you've never used the 12VHPWR connector...

I have a 6700xt.

before all the burning connector jazz I thought putting up to 600 watts(12v 50 amps) through something that small was a bad idea and I still believe it's a bad idea even if the connector didn't have potential for user error. Just look at the adapters, you have 32 wires going into 12 to get the full 600 watts.

 

[m3tavision]

First no one is going on hysterically about it. Your attempts at insulting me don't make your argument any stronger. Yes, it's a production problem and a serious one for AMD. They are going to have to RMA/Replace or Refund a lot of cards compared to less than 100 Nvidia 4090s. The point is that people calling out AMD are not "exaggerating" the scope of the issue contrary to what NeoMorpheus feels. I have yet to hear any info that would indicate how widespread the issue is. Is it a specific batch of cards, some cards in the batch multiple cards from multiple batches or what? Given that it is unclear as to how to determine if a specific card is affected I would be hesitant to buy an XTX right now not knowing if it would need to be returned.

The plug is not flawed. When properly connected is works just fine. I would also say the cooler isn't "flawed", however there is a manufacturing defect that is causing cards to overheat and throttle. As for shutting up, well, that's not going to happen and if you don't like it, tough.

It's not a defect...
Why do you keep saying that, it is a production problem where certain coolers didn't get fluid. I am quite sure AMD sent people into that production facility to check for quality control...


Who knows, perhaps employees were paid to NOT fill certain cooler with liquid... as industrial espionage.. but either way it is a real simple fix. Fill them.