RX 7900 XTX reference at possible vapor chamber design problem + very high hot spot (110c)

[Falck]

@AusWolf , well, you could ask him if he did indeed buy faulty cards only knowingly for getting clicks.

I myself have a little more respect for his work. It would be weird if he now would do that, with his business. If exposed that would end his credibility?

@Veseleil , who/which is a reputable source in your opinion?

 

[AusWolf]

@AusWolf , well, you could ask him if he did indeed buy faulty cards only knowingly for getting clicks.

I myself have a little more respect for his work. It would be weird of he now would do that, with his business. If exposed that would end his credibility?

I was't me who said that - look at post #33.

Having respect for a man's work, and treating it as hard evidence are entirely different things.

 

[Dirt Chip]

The video falls into the "too long, didn't watch" category for me. Can you do a quick recap?

I've seen a post complaining about 110 °C hotspot with around 60 °C delta between hotspot and junction, but that was a severely overclocked card at a questionable news outlet. Knowing der8auer, I wouldn't be surprised if this was the case here as well. What I do find surprising is that we haven't heard anything from more credible sources (TPU, Hardware Unboxed, Gamer's Nexus).

There are several details missing at this point:

• What is "the huge delta" between hotspot and junction?
• Is the card running at default?
• How much does it downclock? Is it noticeable in performance (up to a level that a 3DMark stress test picks it up)?

My reference 6750 XT also runs at 105+ °C hotspot, but the card is rock solid, passes every stress test, and as AMD PR says: it's "within spec".

Yes, default. no OC whatsoever. He saw throttling in clocks, lower voltage and lower power consumption when temp went up to 110c.
He show big increase in temp together with increased RPM when mounted horizontally compere to vertically- extra 10-20c.
So that`s ~20c delta when vertically, ~30-40c delta when horizontally. And than it`s permanent even if going back to vertical- the temp stay as high as in the horizontally mount.
see 14:38 for the "flip test" -something obviously went wrong there with the card.

It`s not a gravity/bending issue, it was ruled out in a test with a GPU stand.
It`s not a mounting pressure issue as well (another test).
It repeated with at least 2 cards, one of them untouched.

Will be interesting to see, after holidays end, if any other channels will be able to replicate the findings.
If not, he might be wrong or on a biased sample size/units.
If replicated, AMD have grand issue to deal with.

 

[eidairaman1]

I can see the wolves circling. Let's keep this amicable and practical. I'm getting sick of partisan BS graphic card threads.

I wonder if this is the same with the 12v power connector thing with Nvidia, there's a supplier working on the cooler that messed up? After all, I doubt AMD actually makes the cooler, surely it's a third party part built to spec that they fit to the board?

Regardless, not good PR at all.

I vote for locking

der8auer 5 card testing show abnormally high delta of average to hot spot temp, plus permanent uplift in temp and fan RPM after mounting horizontally (see 14:38 for the "flip test"). Suggesting permanent vapor chamber failure.
Say a recall of all 7900XTX reference (including 3rd party) is not unreasonable move as a very high number (thousands so far) reporting this problem.
Might dwarf the 12VHPWR saga from NV side, if true.

Stuff your popcorn and enjoy the ride.

However a vc failure doesnt start a fire like a faulty adapter/psu.

der8auer 5 card testing show abnormally high delta of average to hot spot temp, plus permanent uplift in temp and fan RPM after mounting horizontally (see 14:38 for the "flip test"). Suggesting permanent vapor chamber failure.
Say a recall of all 7900XTX reference (including 3rd party) is not unreasonable move as a very high number (thousands so far) reporting this problem.
Might dwarf the 12VHPWR saga from NV side, if true.

Stuff your popcorn and enjoy the ride.

However a vc failure doesnt start a fire like a faulty adapter/psu.

Although easily avoided (just don`t buy the reference), it can damage badly AMD`s image as a quality alternative to NV.
A 'gift' with a sour taste from AMD to all those who get that shiny new 7900XTX for holidays- if will shown as a widespread issue.

Quality from nv? What about the notorious ram failure since the inception of the rtx badge?

 

[LFaWolf]

Thank you for posting this. I have been trying to get an 7900 xtx reference since launch and haven’t been successful. I was contemplating to get the 7900 xt if AMD does a price cut. Now I am going to stay away from them. The testing seems legit and scientific for me.

If I remember correctly, last gen reference cooler was either designed or manufactured by XFX. Not sure about this gen.

 

[AusWolf]

Yes, default. no OC whatsoever. He saw throttling in clocks, lower voltage and lower power consumption when temp went up to 110c.
He show big increase in temp together with increased RPM when mounted horizontally compere to vertically- extra 10-20c.
So that`s ~20c delta when vertically, ~30-40c delta when horizontally. And than it`s permanent even if going back to vertical- the temp stay as high as in the horizontally mount.
see 14:38 for the "flip test" -something obviously went wrong there with the card.

It`s not a gravity/bending issue, it was ruled out in a test with a GPU stand.
It`s not a mounting pressure issue as well (another test).
It repeated with at least 2 cards, one of them untouched.

Will be interesting to see, after holidays end, if any other channels will be able to replicate the findings.
If not, he might be wrong. If replicated, AMD have grand issue to deal with.

Interesting. Let's see if we hear any more about this from different sources in the near future.

 

[Veseleil]

@Veseleil , who/which is a reputable source in your opinion?

If you ignore the pronounced drama show and a lifetime long videos, GN is a quite reputable source from YT. I'm not saying Derbauer is a liar, but saying "AMD is in BIG trouble" judging by a handful of defective cards is a bad journalism imho.
There isn't many websites I follow atm, but I love TPU for various reasons. And when I'm in the market for a new hardware, I'm researching the whole damn internet, from weeks to months, as I don't trust anyone nowadays.

 

[delshay]

Thank you for posting this. I have been trying to get an 7900 xtx reference since launch and haven’t been successful. I was contemplating to get the 7900 xt if AMD does a price cut. Now I am going to stay away from them. The testing seems legit and scientific for me.

If I remember correctly, last gen reference cooler was either designed or manufactured by XFX. Not sure about this gen.

So what happened to CoolerMaster, they used to manufacture/design AMD coolers?

 

[AusWolf]

saying "AMD is in BIG trouble" judging by a handful of defective cards is a bad journalism imho.

This! Nothing more needs to be said. Thank you.

 

[Falck]

I was't me who said that - look at post #33.

Having respect for a man's work, and treating it as hard evidence are entirely different things.

Ah, sorry for accrediting you with that statement.

Well, I'm av interested party in this matter so there's some relief in his findings for me since I've tried re-torquing the card, re-mounting & re-pasting twice, undervoltingand and re-orienting, all with the same result, GPU avg temp 55-65°C with junction at 110°C.

So, with his investigation there's some more answers to my, and a lot of others, experience.

 

[AusWolf]

Ah, sorry for accrediting you with that statement.

Well, I'm av interested party in this matter so there's some relief in his findings for me since I've tried re-torquing the card, re-mounting & re-pasting twice, undervoltingand and re-orienting, all with the same result, GPU avg temp 55-65°C with junction at 110°C.

So, with his investigation there's some more answers to my, and a lot of others, experience.

You're experiencing the same issues? How much does the card clock down? Do you have 3DMark? If so, can you run a Time Spy (or TS Extreme) stress test?

 

[TheoneandonlyMrK]

@AusWolf , well I for one experiences the same problems. Massive delta in temps and throttling. My card is mountet in a ssupd meshlicious, so that orientation doesn't work either. I've even tried tilting the case to get the GPU vertical and that didn't either help.

Someone posted on reddit that if your DP cable has the pin 20 connected, that too could possible be a culprit somehow. I'll test it later.

@TheoneandonlyMrK , well, there's no law against drawing conclusions based on your own research, ref. your flat earth. The reason for me stating that is that in my experience there's a lot of people bashing other peoples hard work for no apparent reason. And then wehn they're proven right there's no retribution or humility from the people who are overly negative. It's to fucking easy being negative. Everyone forgets the bashing, and the troll keeps bashing others and others work. I'm sick of it.

I'm *not* saying we should all praise what der8auer does, ofc not. But have respect for the effort people put in this type of work. If it weren't for those people then most defiantly the companies would get away with a lot more shit than they already do.

Re troll.
(I am not trolling him or you, on his video comment section I don't exist at all but this is TPU a forum for debate, and That is all I'm doing)

Wtaf I said earlier I niether agree or disagree.

But I don't think his Research is extensive enough and absolutely not conclusive.

And as a test engineer who's scientific based proof backed testing has passed millions of pounds worth of analytical instruments.

Buying three broke ones then extrapolating to this degree is lamentable.

And yes I would tell der8aur to his face he's searching for clicks.


Also regarding your butt hurt for der8haur, I like the guy and his content, that doesn't mean I need to back every word he says despite buying a case with his name on it too.

And people actually working in statistical analyses earned they're qualification too with years but der8aur didn't, few have yet many think they are good at it well, I disagree with that thought too.

 

[Dirt Chip]

I vote for locking


However a vc failure doesnt start a fire like a faulty adapter/psu.


However a vc failure doesnt start a fire like a faulty adapter/psu.

Quality from nv? What about the notorious ram failure since the inception of the rtx badge?

If NV is bad quality, and AMD is not much different, they also bad quality. Not helping with sales.
It doesn't burn, just leave you with bad preforming and noisy high end, top of the line GPU.
It`s better, but still bad.
And all is very preliminary, might be turned like nothing.
I wish AMD I was just a small bad and specific sample, with even lower rate of failure of the NV adapter issue.

 

[TumbleGeorge]

If NV is bad quality, and AMD is not much different, they also bad quality. Not helping with sales.
It doesn't burn, just leave you with bad preforming and noisy high end, top of the line GPU.
It`s better, but still bad.
And all is very preliminary, might be turned like nothing.
I wish AMD I was just a small bad and specific sample, with even lower rate of failure of the NV adapter issue.

Probably part of problems are result of fabric clock settings and bad drivers(power in time management, curves) and will be fixed with time.

 

[der8auer]

Re troll.

Wtaf I said earlier I niether agree or disagree.

But I don't think his Research is extensive enough and absolutely not conclusive.

And as a test engineer who's scientific based proof backed testing has passed millions of pounds worth of analytical instruments.

Buying three broke ones the extrapolating to this degree is lamentable.

And yes I would tell der8aur to his face he's searching for clicks.

Absolutely not conclusive? Tell me more And I'm also an Engineer myself so you can talk to a colleague.

Maybe first you should listen more carefully. I first bought a random card 2 weeks ago which had this issue that's why I started investigating the entire topic. After 48 confirmed cases via email I bought 4 of these so I have a total of 5 cards for extensive testing. All of them behave the same way with the flip test. Ruling out all other possibilities like TIM and mounting pressure, let me know how this is "absolutely not conclusive"

 

[TheoneandonlyMrK]

Absolutely not conclusive? Tell me more And I'm also an Engineer myself so you can talk to a colleague.

Maybe first you should listen more carefully. I first bought a random card 2 weeks ago which had this issue that's why I started investigating the entire topic. After 48 confirmed cases via email I bought 4 of these so I have a total of 5 cards for extensive testing. All of them behave the same way with the flip test. Ruling out all other possibilities like TIM and mounting pressure, let me know how this is "absolutely not conclusive"

Batch, you ruled out a bad batch too?.
You know how many were released?.

Blind testing one in ten or some arbitrary amount might help instead of just testing broken one's.

And your statistical analyses qualies?.

Because your extrapolation Also seems extreme.


And finally wtaf, what in science was proved and not backed by further second and third person testing.

Consensus isn't usually reached by just one man, sorry.

 

[Selaya]

watched der8auer's video.

can already exactly picture what happened.
some fucking retarded imbecile made this disaster of a vapor chamber.
they put it on a card, jammed it on a test bench, ran a few test passes, saw nothing out of the ordinary.

the thing is, maybe one in a hundred people have their card in horizontal orientation (the way the vast majority of all actual users will use theirs) on their test bench.
evidently the guys who made this piece of crap didn't. they had theirs in vertical. it worked™ on vertical. they didn't realise it won't on horizontal. they didn't even bother to check/test/debug for it.

they just signed it off & shipped it like that.

merry (belated) christmas, amd.
maybe do some proper QC next year.

 

[der8auer]

Batch, you ruled out a bad batch too?.
You know how many were released?.

Blind testing one in ten or some arbitrary amount might help instead of just testing broken one's.

And your statistical analyses qualies?.

A "bad batch" would be even worse. If you compare that RTX40 gen was ~150k units for first batch. Obviously AMDs market share is lower so the amount of sold cards is also lower.
It's pretty safe to say that all cards in the current market are from the first batch as at the time shooting the video that card were out for 10 days.

But again, how is this not conclusive? If you are able to replicate a problem under the same condition every single time then it is conclusive. As an engineer you should know that. As an engineer you should also know that 10 is not a sufficient sample size to determine how many are affected.

 

[eidairaman1]

So i presume these were from all the same batch?

 

[Dirt Chip]

This thread already pop up one user with the same issue, I guess some having the same issue and are passive readers or will comment in the coming days.
Anyway, there's no denying that the tested units behave in unreasonable manner to the point of RMA.
How prevalence this issue is, and if it will become a true shitstorm for AMD, depends on the rate of affected units worldwide.
I sense it might be just the tip of the iceberg for this phenomena.
Hope not, we all deserve better competition and no consumer should get such experience with top of the line product.

 

[LFaWolf]

So what happened to CoolerMaster, they used to manufacture/design AMD coolers?

Cooler master designed and manufactured the AMD CPUs coolers, wraith or whatever it is called. I am not sure if they ever did gpu coolers for AMD

 

[Veseleil]

watched der8auer's video.

can already exactly picture what happened.
some fucking retarded imbecile made this disaster of a vapor chamber.
they put it on a card, jammed it on a test bench, ran a few test passes, saw nothing out of the ordinary.

the thing is, maybe one in a hundred people have their card in horizontal orientation (the way the vast majority of all actual users will use theirs) on their test bench.
evidently the guys who made this piece of crap didn't. they had theirs in vertical. it worked™ on vertical. they didn't realise it won't on horizontal. they didn't even bother to check/test/debug for it.

they just signed it off & shipped it like that.

merry (belated) christmas, amd.
maybe do some proper QC next year.

QC is a thing of the past. Hardware and software wise. They don't need it anymore, as early adopters are the actual testers. If people could stop running to buy every damn new piece of hardware/software the day after it's in the market, and wait for a year or so to let others finish the testing... I could go on, but some might find it offending.

 

[OneMoar]

its a design issue this issue can't just effect a handful of cards every single one of them with this cooler will be affected unless the ambient temps are really low

whats happening is that the liquid in the vapor chamber is boiling off, and either the volume calculations of the vapor chamber are WAY wrong. or the condensed liquid can't find a return path due to some internal restriction I would guess they have some internal chamber design that was designed by a intern ....


the result is hilariously disastrous you might as well not run a cooler at all at that point

Vapor Chamber Cooling Design Guide | Celsia

Design guide covering vapor chamber cooling for electronics: best uses, types, heat sink integration, calculators & much more.

celsiainc.com

this is way worse then the NV 12VHPWR if amd cares at all they are going to need to recall every single card for a revised cooler

 

[TheoneandonlyMrK]

A "bad batch" would be even worse. If you compare that RTX40 gen was ~150k units for first batch. Obviously AMDs market share is lower so the amount of sold cards is also lower.
It's pretty safe to say that all cards in the current market are from the first batch as at the time shooting the video that card were out for 10 days.

But again, how is this not conclusive? If you are able to replicate a problem under the same condition every single time then it is conclusive. As an engineer you should know that. As an engineer you should also know that 10 is not a sufficient sample size to determine how many are affected.

You replicated it five times, four of which were known broken.

Buy four new from across the batch and if you're right and they do it to then that would be along the path of proof.

And I would like to see your failing cards fixed with a waterblock, that would put a nail in the coolers coffin on your failing one's tbh.

The heatsinks could have been made in batches, perhaps one batch of those were made wrong.

You received 48 emails from people with this issue.

And that then becomes thousands?.

And what exactly do you have against the scientific method of second source confirmation, I think nothing .

I have at least watched your video now.

I don't doubt your finding, just your extrapolation and to a degree given your evidence, your conclusion.

Third party testing would reassure me of your findings.

But no I take no man's word as gospel, I doubt you do.

 

[OneMoar]

and PSA: The AMD fanboys can be quiet, sorry your best friends fucked up and robbed a bank
but they still gotta goto jail....

and to those thinking well it can't be that many cards

If one got down the assembly line like this there where thousands because they are all stamped and sealed and filled on one line