EVGA Reveals that Bad Soldering Was the Culprit behind Bricked RTX 3090 Cards

[Uskompuf]

When Amazon Studios released the closed beta for their upcoming New World MMO users begun to report that the game was causing catastrophic failure of their RTX 3090 graphics cards. This issue appears to have disproportionately affected EVGA cards with initial analysis indicating that the cause could have been with the fan controller malfunctioning evident by the bad readings received from third-party monitoring tools. This behavior was determined to not be the cause of card failures with monitoring working correctly when using the EVGA Precision X1 software. EVGA found that the fan controller issue was related to noise on the communication bus which their software could filter out and have released a micro-controller update that fixes monitoring with third-party tools.

The cause of the dead cards has now been determined by EVGA to be the result of a rare soldering issue limited to a small batch of cards. EVGA performed X-ray analysis on a selection of broken cards returned to them and found that the soldering around the MOSFET circuits had "poor workmanship". EVGA also confirmed that this issue was only present on early production runs in 2020 and affects less than 1 percent of all the RTX 3090 cards manufactured by the company. Amazon Game Studios have also added a frame limiter to New World that will ensure this issue won't appear in new versions of the game.

All of the cards were earlier production run cards manufactured in 2020. Under an X-ray analysis, they appear to have "poor workmanship" on soldering around the card's MOSFET circuits that powered the impacted cards.

View at TechPowerUp Main Site

 

[Chaitanya]

One more addition in book for EVGA failures.

 

[Crackong]

Now I am disappointed.
Amazon didn't make the ultimate "GPU killer"
Crysis still sitting on the throne.

 

[ZeppMan217]

One more addition in book for EVGA failures.

At least they handled the problem better than Gigabyte.

 

[watzupken]

While it is bad that the bad soldering quality issue slipped through their QC checks, I think I have to give it to EVGA for taking quick actions to remediate/ remedy the issue. In addition, it does seems like they certainly looked into the issue and acknowledged the issue on their part.

At least they handled the problem better than Gigabyte.

This is precisely the point. I think while it is not desirable, but products with QC issue will most certainly happen at some point for any company. It is how they handle the issue that makes the difference. I prefer companies to be transparent, and take ownership of the issue, rather than one that says they take things seriously, but not doing anything at all until they get pressurised to do it.

 

[Naito]

This is precisely the point. I think while it is not desirable, but products with QC issue will most certainly happen at some point for any company. It is how they handle the issue that makes the difference. I prefer companies to be transparent, and take ownership of the issue, rather than one that says they take things seriously, but not doing anything at all until they get pressurised to do it.

Exactly. I'm avoiding Gigabyte until they can prove they'll own up and step up to their mistakes. EVGA should be praised for investigating and acknowledging the issue and not sweeping it under the rug. While there are extreme shortages, they now just have to make the RMA as smooth as possible for the owners.

 

[eidairaman1]

Exactly. I'm avoiding Gigabyte until they can prove they'll own up and step up to their mistakes. EVGA should be praised for investigating and acknowledging the issue and not sweeping it under the rug. While there are extreme shortages, they now just have to make the RMA as smooth as possible for the owners.

They had no other choice but to make an announcement.

One more addition in book for EVGA failures.

They may have a great PSU line thanks to Seasonic but their gpus are lacking

 

[Bwaze]

I wouldn't be surprised if some buyers weren't legible for dead card replacement - it was a mad scramble to get the cards, some of them were sold through strange channels, scalpers...

 

[john_]

I don't believe them. I think this is just damage control. Throwing the blame on bad soldering is a much better excuse for the company than admitting that the problem was bad choice of components or bad implementation or settings of them. Especially that "earlier batches" says what? That the problem was fixed latter. And if yes, how was it fixed? Pure luck or did they found out about it, but decided to say nothing about it? Maybe that Amazon game was in fact something good. All cards failed at the same time and it gone viral. If we had cards die here and there at random and in totally different times, many would had be having a hard time to convince the retailer or EVGA that they where not responsible for the death of their card.

 

[Chaitanya]

They had no other choice but to make an announcement.


They may have a great PSU line thanks to Seasonic but their gpus are lacking

Some of their PSU also have had complaints of failures/burning/explosions though not as spectacular or consistent as GB GP-P7/850GM ones.

 

[Jack1n]

The way EVGA handles warranty is a very big plus in my book. If retailers did not have to handle warranty for the entire period here in Israel, I would not consider any brand other than EVGA.

 

[laszlo]

since the industry has changed the solder to lead-free problems just increased especially at bga contacts of laptop gpu's...

the new lead-free solder doesn't handle so well the thermal expansion and contraction and have more internal flaws in the structure

 

[Vayra86]

Well done. Whether or not its true is irrelevant at this point if the buyers are being compensated, as much as we can have our doubts. Either way, the idea that something was wrong on the hardware side was obvious. Games don't blow up GPUs like this.

 

[Chomiq]

At least they actually investigated the issue and admitted their fault. Still would make me thing twice about buying EVGA again.

 

[Ferrum Master]

since the industry has changed the solder to lead-free problems just increased especially at bga contacts of laptop gpu's...

the new lead-free solder doesn't handle so well the thermal expansion and contraction and have more internal flaws in the structure

Nope, it is all about profiles. Lead free solder problems are thing of past now, the technology has matured.

Some just gimped 5C at one side and such problems emerge. It is a very large multilayer board, probably due to very high power design in mind, the VRM area is very copper dense and multiple vias poured thus needing more energy to properly diffuse solder. Well someone used a generic profile, that has worked fine previously on older cards and don't fix what ain't broken.

Other than that... it seems furmarking your newly acquired card ain't a bad idea... well at least while you have your warranty and not to have a revelation that you had a cold joint after the warranty ends.

 

[Bomby569]

EVGA strikes again. And when they do they do it big. At least this ones didn't explode, i guess it's a win.

 

[theGryphon]

I don't believe them. I think this is just damage control. Throwing the blame on bad soldering is a much better excuse for the company than admitting that the problem was bad choice of components or bad implementation or settings of them. Especially that "earlier batches" says what? That the problem was fixed latter. And if yes, how was it fixed? Pure luck or did they found out about it, but decided to say nothing about it? Maybe that Amazon game was in fact something good. All cards failed at the same time and it gone viral. If we had cards die here and there at random and in totally different times, many would had be having a hard time to convince the retailer or EVGA that they where not responsible for the death of their card.

Here is a likely scenario: Early production batches went through the soldering machine with a calibration setting that was not ideal. They were checked/tested with their internal QC criteria, some failed to be reworked, but some passed. Observing this, they recalibrated the soldering for next batches. Manufacturing involves numerous learning stages, so this is very natural.

What can be said is that 1) they should have a managerial call to fail that whole batch, observing high QC failures, 2) They should revise the QC testing criteria/environment.

 

[DeathtoGnomes]

This is what happens when outsourcing, its a shame video cards are not tested thoroughly as some power supplies are.

 

[ThrashZone]

Hi,
Just an unfortunate early adopters thing I remember 20 series deaths were happening too with many manufactures pretty much why I hold off for a while.
EVGA stepped up as any owner would hope.

I've had/ have many evga cards and none have died all are/ were really good oc'ers to so they are my go to gpu manufacture "maybe not directly from them" regardless of this situation or past.

 

[Chrispy_]

EVGA seems to always get praise for great customer service but at the same time they need it because this is far from the first time they've been in the spotlight for defective graphics cards.

"Sorry our products are shit but we'll replace it for you without a fuss" isn't as good as just having a higher-quality working product in the first place.

 

[Tardian]

Everyone no matter how careful makes mistakes. Good quality entities identify the problem as early as possible, recognize it is a problem, ascertain the cause of the problem, fix the problem, and then implement a permanent solution that ensures the problem does not reoccur. EVGA appears to have done a good job with this issue. It is impossible to QM every point of the solder.

 

[neatfeatguy]

EVGA wants to make an announcement about the issue with some of their RTX 3090 cards mechanically failing after launching New World.

Eventually the people will forget this all happened, so it'll just work itself out naturally.

 

[eidairaman1]

Some of their PSU also have had complaints of failures/burning/explosions though not as spectacular or consistent as GB GP-P7/850GM ones.

That is why I do research on who the OEM is using realhardtechx.com, that is how I just learned Corsair is cheaping out on the TX-M line by switching from Channel Well Technology (CWT) to Great Wall... This is why im so critical of them.

 

[laszlo]

Nope, it is all about profiles. Lead free solder problems are thing of past now, the technology has matured.

Some just gimped 5C at one side and such problems emerge. It is a very large multilayer board, probably due to very high power design in mind, the VRM area is very copper dense and multiple vias poured thus needing more energy to properly diffuse solder. Well someone used a generic profile, that has worked fine previously on older cards and don't fix what ain't broken.

Other than that... it seems furmarking your newly acquired card ain't a bad idea... well at least while you have your warranty and not to have a revelation that you had a cold joint after the warranty ends.

the "matured" version of lead-free solder , with the best components is v.expensive compared to lower versions not to mention than leaded one which is the cheapest...; lead-free ones comes in many versions depending on composition and producers tend to use the cheapest version of it hence the arising issues

in the end is like a lottery as we don't know what they use ....

 

[xkm1948]

Not surprised. The first batch 3090FTW3 had power balance issues along with many other issues. Not all was affected. EVGA hold a small replacement program and replaced quite a lot of those first batch cards. We had to sent in serial number along with many sensor readings. The first batch 3090 was made in two factories. One of then was new to GPU manufacturing and that is most likely the problem.