GPU Die Defects Found in PowerColor RX 9070 XT Graphics Card

[AleksandarK]

Surface defects on the GPU die of a PowerColor Radeon RX 9070 XT Hellhound have been linked to excessive thermal hotspots, according to analysis following Igor's Lab's investigation. Microscopic examination revealed 1,934 pits across the silicon die surface despite proper thermal interface material application. The defects affect over one percent of the chip's surface area, with the largest measuring 12.59 µm in depth and 212.36 µm in diameter—significantly exceeding industry tolerances. These imperfections impede heat transfer, causing localized temperatures to reach 113°C, exceeding the 110°C threshold for RDNA GPUs and triggering thermal throttling. The defects appear to stem from flaws in the backgrinding process that thins silicon wafers for packaging. Improper process control creates surface irregularities that compromise thermal conductivity and structural integrity.

The discovery raises questions about the effectiveness of automated optical inspection systems in detecting subtle surface abnormalities. TSMC, the maker of these chips, uses optical recognition tools to see if any defects occurred. AMD stated: "We are aware of the reported issue and believe this to be an isolated incident. We are working with our partners and internal teams to understand the issue." The company maintains this is an isolated case, though the findings suggest potential gaps in quality control protocols. Users experiencing thermal issues with affected RX 9070 XT units should initiate the RMA process for replacement while AMD and PowerColor investigate whether these defects exist beyond the identified sample. We hope no further issues are present, and so far, PowerColor hasn't received any customer complaints. We are on the lookout for further situation development. We reviewed PowerColor's Radeon RX 9070 Hellhound, and found no issues on our unit, so this indeed remains an isolated case so far.



View at TechPowerUp Main Site | Source

 

[rv8000]

Seems like a propaganda piece, what’s their sample size? Where’s the source link? (found it, why is it in tiny gray letters, in a minuscule font at the end of the page??)

We’re talking about 100ths of a mm ffs, the majority of surfaces are not flat, and the sole purpose of TIM is to account and or make up for that fact.

 

[HOkay]

Seems like a propaganda piece, what’s their sample size? Where’s the source link? (found it, why is it in tiny gray letters, in a minuscule font at the end of the page??)

We’re talking about 100ths of a mm ffs, the majority of surfaces are not flat, and the sole purpose of TIM is to account and or make up for that fact.

I'd agree, but if they're applying TIM & still getting 113C hotspot temps then they're obviously deep enough to be a problem.

 

[redeye]

OTOH, this what warranties are for, unexpected shortfalls…

 

[RejZoR]

Sounds like isolated issue. Not because I'd be AMD fanboy, but just because of nature of it. It's physical defect on the surface that can happen during transport or during assembly at PowerColor. Unlike for example missing ROP on RTX 5090 which is a pure fab processing error on very high level of laser cutting decision making, because if it's a random defect on a ROP, it's very likely it would artifact or corrupt data on more significant scale than just lacking some performance because it would be precisely be missing as whole perfect unit.

 

[rv8000]

I'd agree, but if they're applying TIM & still getting 113C hotspot temps then they're obviously deep enough to be a problem.

They didn’t even test mounting pressure and contact quality of the cooler. The article is hyper focused on nearly microscopic pitting without double checking typical issues for hot spots/high temps - cooler mounting.

Article seems slapped together to get views without thorough testing.

 

[Prima.Vera]

Are there more than 1 units discovered with this issue so far?

 

[ZoneDymo]

Do we even take Igor's lab seriously anymore after that whole metal fan thing from alpha cool?

 

[Denver]

If the defect led to extreme temperatures and throttling, it would be widely noticeable and reported if numerous cases existed.

 

[Sithaer]

While I agree that a single sample is like whatever but I would bet money that if this happened to an Nvidia GPU then the masses would be bringing their pitchforks already.

 

[Denver]

While I agree that a single sample is like whatever but I would bet money that if this happened to an Nvidia GPU then the masses would be bringing their pitchforks already.

How would you know if Nvidia removed the hotspot sensor?
Conspiracy successfully planted.

 

[Sithaer]

How would you know if Nvidia removed the hotspot sensor?
Conspiracy successfully planted.

I wouldn't know but some ppl do find out these issues one way or another, that wasn't my point. 'Stuff like this with barely any samples are only good to stir up tech drama imo'

 

[Veseleil]

Igor's Lab is not a trusted source to begin with IMO. Not as a whole bunch.
Steve is not the only drama queen apparently, many tech outlets exercise their unhinged love for this kind of a publicity. Shocking, disturbing, scam! They need to give their "precise lab testing equipment" a run for the money sometimes. I can understand that.
People that follow this space know very well that the industry (all of them) cost cutting trend had to show its disadvantages, one way or another. Faulty samples or even whole product series is nothing new, and we can only expect more of it in the future. Boo hoo.

 

[TSiAhmat]

How would you know if Nvidia removed the hotspot sensor?
Conspiracy successfully planted.

 

[JasBC]

Really do not get the tone of these comments. . .

 

[sudothelinuxwizard]

Seriously, why is this shit even newsworthy if it happens once? Phones blow up semi-regularly and no one bats an eye, so why should GPUs be any different?

 

[Jacky_BEL]

1934 defects on a single die? how about the other chips from that wafer? Surely it should have been noticed on inspection before leaving the fab.
I think a mishap during assembly is more likely.

BTW, 200 micrometers is huge in silicon surface finishing.
Looks like one of the photo's shows a crack developped at the edge of a pit.

 

[_roman_]

Seems like a propaganda piece

I read the german text on igorslab.

In my own words what I remember. I will not cross check now.

Igor was notified about a graphic card which got very hot.

IGOR clearly stated this is a special case. And not to make big ballon out of this one card he analysed.

It's the usual igor piece. I fast read it. I did not read in details about his thoughts about how waffers are made and such. If I read it, I'll read it again and than read it very carefully and than I'll write my thoughts there. It's the usual Igor about his viewpoint how things are made, quality process and such. and also criticism why such card went to the market. And criticism why it was not detected from powercolor.

The defects appear to stem from flaws in the backgrinding process that thins silicon wafers for packaging. Improper process control creates surface irregularities that compromise thermal conductivity and structural integrity.

The discovery raises questions about the effectiveness of automated optical inspection systems in detecting subtle surface abnormalities.

I see it from the viewpoint. A customer had problems with the graphic card and asked Igor for help.

edit: cleanup of text a bit

 

[mb194dc]

Seems like TSMC got a QA issue there and probably other chips out there with this issue.

 

[_roman_]

Do we even take Igor's lab seriously anymore after that whole metal fan thing from alpha cool?

There are not many pages which creates tech contents. Topics are difficult. Tech wise. Methodically difficult. Maybe it's in the nature of the topic itself.

Well Thermal paste topics = 50 / 50

Power supply units = cybernetics guy = 50 / 50 (I dislike all the praise for Corsair psu. I had personally a dead corsair PSU)

Cpu cases - as good as anyone else - why not

monitors = additional information's

newspieces = a lot of copy and paste most of the time

tech basics / electronics basics = read and criticise as I see it differently. I appreciate the motivation to write such articles. It's not very easy to do that.

Fan topic: Gamers nexus for example bought big machine. Which they hardly got in use yet. I think fan testing is complicated. Very complicated.

 

[freeagent]

TIM does not replace silicon though..

Seems like TSMC got a QA issue there and probably other chips out there with this issue.

This is along the lines of what I was thinking. Might be a bit rushed.

 

[Wirko]

Does anyone know how thick this Navi die is?

I mean, heat transfer can be significantly impeded around pits 200 um in diameter if the thickness is roughly the same 200 um or less, so heat from certain areas (above the center of any large pit) travels a significantly longer distance, sideways and up, through the silicon.

 

[Scrizz]

I think a mishap during assembly is more likely.

I agree with you on this. This looks more like some form physical damage ... maybe during cooler install or handling.

 

[Athena]

Microscopic examination revealed 1,934 pits across the silicon die surface despite proper thermal interface material application. The defects affect over one percent of the chip's surface area, with the largest measuring 12.59 µm in depth and 212.36 µm in diameter—significantly exceeding industry tolerances

I also think there is more to it than this article lets on

It could be mishandling of the chips at any point or some other reason, we just don't know enough at this point

 

[AusWolf]

If the defect led to extreme temperatures and throttling, it would be widely noticeable and reported if numerous cases existed.

Exactly my thoughts.