News Third-Party GPU Makers Respond to Nvidia RTX 30-Series Crash to Desktop Issues

kal326

Distinguished
Dec 31, 2007
1,229
108
20,120
Well this will be some interesting instant karma for all those scalpers potentially sitting on problematic cards. Hard to charge way more than retail and have no easy way guaranteed support to the potential secondary market buyer.
 

bigdragon

Distinguished
Oct 19, 2011
1,107
547
20,160
I remember having a mobile Nvidia GPU that had the "solder bump" issue a long time ago. Getting that fixed was an absolute nightmare because Dell didn't recognize crashing during gaming to be a problem. I expect something similar is going to play out here. Vendors will dial back to the boosts to improve stability, but the design will forever be rushed and flawed. Nvidia should have delayed the Ampere launch.
 
  • Like
Reactions: Avro Arrow
Who actually has a dang card to verify any of this LMAO!!!!
Apparently TH Labs does, let's just hope that they tell us what's going on sooner rather than later. You never know...
i hope all the cards taken by scalpers have this issue.
So do I but IIRC, the scalpers were mostly grabbing FE models which don't have these issues.
Well this will be some interesting instant karma for all those scalpers potentially sitting on problematic cards. Hard to charge way more than retail and have no easy way guaranteed support to the potential secondary market buyer.
It would also be instant karma for those morons who are willing to pay triple for the card and feed said scalpers.
I remember having a mobile Nvidia GPU that had the "solder bump" issue a long time ago. Getting that fixed was an absolute nightmare because Dell didn't recognize crashing during gaming to be a problem. I expect something similar is going to play out here. Vendors will dial back to the boosts to improve stability, but the design will forever be rushed and flawed. Nvidia should have delayed the Ampere launch.
I agree. I don't understand what nVidia was thinking because the RTX 3080 is really a great product. A rushed launch shouldn't have been necessary and an EXTREMELY-rushed launch like this makes nVidia look terrified of ATi's RDNA2 architecture. I can't remember the last time I've seen nVidia behave this way, not even when the HD 7970 was poised to beat everything that they had. Something's up and I don't know what it is.
 
Last edited:

newood

Distinguished
Mar 30, 2009
12
2
18,515
I just want to point out that these companies have to follow Nvidia's power delivery specs. So if it turns out to be a capacitor issue its on them and not the third-party GPU maker. The other obvious thing is the third-party GPU maker and buyers are the ones that are going to suffer the most.
 
With the little time that has passed since the cards were available, jumping to conclusions that this was all a cap issue was the first mistake everyone involved had. But I guess you have to start from somewhere when trying to fix a problem right?

There could be lots of things that can crash a new pc component, hardware is just one of them, but could be related to software too (drivers, windows installation, windows updates, other apps, some games, under certain conditions, etc.).

Maybe is something related to both things, hardware and software.

Perhaps, since most reviewers always start benchmarking with a fresh install of windows they never encounter this issue. And some people that just take old GPU out and install the new one have more chances to crash.
 
I agree. I don't understand what nVidia was thinking because the RTX 3080 is really a great product. A rushed launch shouldn't have been necessary and an EXTREMELY-rushed launch like this makes nVidia look terrified of ATi's RDNA2 architecture. I can't remember the last time I've seen nVidia behave this way, not even when the HD 7970 was poised to beat everything that they had. Something's up and I don't know what it is.
I'm curious to know how NVIDIA would be scared of a product they have no access to and unless it was the pointy haired bosses sniffing WCCFTech news posts all day, would have no reason to speculate what kind of product AMD would have other than to take AMD's official word... which is about as good as anyone else's. Granted RDNA1 put up a good show, but would it be enough to make NVIDIA scared pantless over the next generation?

If anything, I'm more willing to believe NVIDIA wanted at least something good to come out of this year to make the investors happy. Looks like that's starting to backfire.

With the little time that has passed since the cards were available, jumping to conclusions that this was all a cap issue was the first mistake everyone involved had. But I guess you have to start from somewhere when trying to fix a problem right?

There could be lots of things that can crash a new pc component, hardware is just one of them, but could be related to software too (drivers, windows installation, windows updates, other apps, some games, under certain conditions, etc.).

Maybe is something related to both things, hardware and software.

Perhaps, since most reviewers always start benchmarking with a fresh install of windows they never encounter this issue. And some people that just take old GPU out and install the new one have more chances to crash.
You could believe this is case of "people only write negative reviews" or bad actors are trying to amplify the image they want what they have beef with. However if a lot of people do have issues, regardless of what they tried, then it should be given a look. Especially when it happens under the same certain conditions of the hardware itself.
 
  • Like
Reactions: Phaaze88

OMGPWNTIME

Distinguished
May 1, 2009
56
5
18,535
Something that's confused me about all of this, these companies all claim that none of the retail cards have the faulty cap layout, yet aren't many consumers the ones reporting the crashing issue?
 
Something that's confused me about all of this, these companies all claim that none of the retail cards have the faulty cap layout, yet aren't many consumers the ones reporting the crashing issue?
The problem here is that it's not really faulty if it still works fine within the advertised specs. From what I read, stability issues happen way beyond the boost speeds advertised to you. e.g, the 3080's boost clock is advertised at 1740 MHz. If you're crashing at say 2000 MHz but stable at 1950MHz, for all intents and purposes, the card is working as advertised.
 
  • Like
Reactions: Phaaze88
You could believe this is case of "people only write negative reviews" or bad actors are trying to amplify the image they want what they have beef with. However if a lot of people do have issues, regardless of what they tried, then it should be given a look. Especially when it happens under the same certain conditions of the hardware itself.

I don't believe or not believe anything, Im not nvidia nor AMD fanboy.
Also define a "lot of people"?. Do you have the exact number of people having this issues?
There has been issues, but how many people is a "lot"? 10, 55, 100, 1587, 17000 buyers?, How many card were sold till today?. What percentage of those cards are having issues? There are still a lot of question that need to be answer before jumping to conclusions.
This is not a normal launch, we are under a global pandemic and the stock of cards available were really small. Some countries didn't got any card at all.
Of course this needs to be investigated (and my post states that), but just because something seems to be the "miracle" answer, it might as well not be exactly that.


When AMD launched thier Ryzen 3xxx CPU "lots of people" (see what I did there :) ) complained about 3 things:
  • voltage was going crazy, and crazy high.
  • cpu were not hitting the right boost clocks.
  • fan was very loud and going up and down all the time..
We then found out that there werent many voltage issues, and some real issues were on the software, user or motherboard side (Yes, early BIOS was a bit messy but AMD fixed most of the stuff on the first week, it just took another week or two for motherboard maker to release the BIOS).

The CPU not hitting the advertised boost clock, I never encounter a single issue with my CPU, even with the horrible stock fan my early Ryzen 5 3600 (I got it as sson as it was available in my country) hit the right 4.2GHz single core and around 4.05GHz all cores depending on the workload.

The stock cooler was not broken, it was working as intended. When a CPU change its boost state as zen 2 does its normal for the fan to do what this cooler does.

Anyway, back to nvidia RTX 3xxx.

If this is a hardware issue, then it may be because of how (I think) nvidia rushed the launch to stay in front of AMD (I may be wrong but we will never know). And that may have caused that the usual QA testing (specially) of the AIB models design, which nvidia requires and demand, been rushed too.

I guess only time will tell, if we ever found out.
 
Last edited:
THIS, is why you DON'T buy hardware when it first comes out. If you are an early adopter, and there are major issues, then it really shouldn't come as THAT big of a surprise. It certainly wouldn't be the first time issues came up shortly after launch. New hardware, wait a month or two. Every veteran builder and gamer with an ounce of common sense knows that by now. Besides which, if you wait, it's usually less expensive by then as well AND you'll probably end up with better hardware as the binning process is refined.
 
THIS, is why you DON'T buy hardware when it first comes out. If you are an early adopter, and there are major issues, then it really shouldn't come as THAT big of a surprise. It certainly wouldn't be the first time issues came up shortly after launch. New hardware, wait a month or two. Every veteran builder and gamer with an ounce of common sense knows that by now. Besides which, if you wait, it's usually less expensive by then as well AND you'll probably end up with better hardware as the binning process is refined.
True, but some cases are also true in reverse - I bought a Radeon RX480 8Gb (reference model from Sapphire) the month it came out - I paid €270 at the time (it was the recommended retail price in my country). Three months later, due to bitcoin mining, it was selling for almost double and stayed there for more than a year. Its successor, the RX580, was pretty much the same.
I still own that card. For almost 2 years I could have resold it for more than I bought it. I was sorely tempted at times, but with nothing to replace it I relented. After 18 months I replaced the cooler to make it more silent and less hot, and all my attempts at overclocking it led to instability and crashes.
The only thing I regret about buying it early is that there's just no way to overclock it in a meaningful way. But, having owned it for almost 4 years now, I really can't say it was a bad investment.
Pretty much like the HD7770 1 Gb and HD4850 I had before it.
 
Bitcoin mining is an outlier. Nothing like that ever happened before, and it is unlikely anything like that will ever happen again, especially since there is so much in the way or mining specific hardware now if it ever becomes a hot thing again. Even so, it doesn't change the fact that had there been architectural problems with the card, which is what I am talking about, not whether it's resale value might appreciate or depreciate, then it wouldn't have had any value as a mining investment anyhow.

I am talking about fundamental engineering and architectural issues that are missed until products are in the wild, and then fixed in revisional hardware, so that buying the same hardware after the bugs get ironed out makes a lot more sense than getting in early and at best dealing with a replacement or at worst getting stuck with a box of lemons.
 

nofanneeded

Respectable
Sep 29, 2019
1,541
251
2,090
Nvidia statement is not helping at all because they kept us in the gray area .. they did not point at the problem at all and by saying that it might not be capacitors problem they will make people more and more afraid from buying the product.

People will start to fear that the problem is in the GPU itself ...
 
  • Like
Reactions: colson79
Anybody who buys one of these cards, right now, before the problems are 100% identified and completely sorted out, is especially foolish. They SHOULD fear that there are problems with the fundamental design of the card itself, because this is a new and probably somewhat rushed design, and there might be significant revisions required to the architecture to resolve the problems. If the problems aren't component specific and are down to design flaws, it could be an even larger problem than it seems to be.

It shouldn't come as any surprise that Nvidia hasn't come forward and said, Well, yeah, you're right. We messed up. Let us go back to the drawing board and fix it and we'll let you know when it's all good". That, isn't likely to happen.

Especially if this is primarily a board partner issue. I haven't seen yet whether this is a problem on FE cards or not.
 

yourilevoye

Reputable
Jan 6, 2017
221
10
4,715
Anybody who buys one of these cards, right now, before the problems are 100% identified and completely sorted out, is especially foolish. They SHOULD fear that there are problems with the fundamental design of the card itself, because this is a new and probably somewhat rushed design, and there might be significant revisions required to the architecture to resolve the problems. If the problems aren't component specific and are down to design flaws, it could be an even larger problem than it seems to be.

It shouldn't come as any surprise that Nvidia hasn't come forward and said, Well, yeah, you're right. We messed up. Let us go back to the drawing board and fix it and we'll let you know when it's all good". That, isn't likely to happen.

Especially if this is primarily a board partner issue. I haven't seen yet whether this is a problem on FE cards or not.

The issue seems to be happening on all cards. Also FE, as tested by Paul's Hardware in this video that was uploaded yesterday. It seems like some will just have issue's with it sooner than others.
View: https://www.youtube.com/watch?v=BkTgYMlCl4E&t
 
  • Like
Reactions: hannibal

hannibal

Distinguished
True. The caps Are not the issue! Firmware and driver boost just allow the cards to run too high clockspeeds. Both Are easily corrected via new firmware and updated drivers.
But this should not have happened! So don`t rush the releases and give aib enough time to test. Result leaks will hapen more, but that is better than bad rep from unstable behaviour in release!

here is one more excelent video about this matter!
View: https://youtu.be/ud6NrbJllzk


De8auer did test the situation and find out the same. Normal sp-caps Are just fine! Ceramics gives a tiny bit more oc headroom, but that is all.
 
Last edited:
  • Like
Reactions: sizzling

yourilevoye

Reputable
Jan 6, 2017
221
10
4,715
True. The caps Are not the issue! Firmware and driver boost just allow the cards to run too high clockspeeds. Bot Are easily corrected via new firmware and updated drivers.
But this should not have happened! So don`t rush the releases and give aib enough time to test. Result leaks will hapen more, but that is better than bad rep from unstable behaviour in release!

here is one more excelent video about this matter!
View: https://youtu.be/ud6NrbJllzk


De8auer did test the situation and find out the same. Normal sp-caps Are just fine! Ceramics gives a tiny bit more oc headroom, but that is all.

Indeed.
Here a statement from Nvidia around 14 hours ago:

"Regarding partner board designs, our partners regularly customize their designs and we work closely with them in the process. The appropriate number of POSCAP vs. MLCC groupings can vary depending on the design and is not necessarily indicative of quality."

I figured many people just hopped on the capacitor train, point at them because some people say it's the issue, while the people themselves don't really understand what the issue is.

It's good for Nvidia to give AIB enough time to test indeed, so I hope they learn from this. The 2000 series release wasn't 100% perfect either.
Most people say if you keep things stock you will be just fine. At the end, if a CPU doesn't really reach an OC that was not advertised anyway people won't really complain either honestly. This does exclude the cards that did have issue's with stock settings though, although many haven't.

I noticed that there wasn't much to be seen about Gigabyte. Many people make statements like "Gigabyte uses the same poscaps (which aren't poscpas) as Zotac", which isn't the case. They are different. So is any card. There is more to it than the capacitors.

Personally I don't mind an update that will limit the OC. As long as it's making it stable and doesn't bring any downsides. Some think differently though, but I am not a type of person that likes to fiddle around with overclocking. At the end, it's still as advertised if it gets limited, although I understand how some people will not agree on that. But it's either that or a crashing card from what I can tell.

PS. Also seen many things about manufacturers showing pictures of their GPU with different capacitor layouts on their site than what they actually come with. From my understanding from people that have experience with this, the engineers and marketing team does not always have 100% knowledge of eachother. Not sure how to say this properly, but what they mean by this is that the pictures do not always represent the card on an engineering level, if that makes sense (kinda hard for me to explain in an other language).
 

spongiemaster

Admirable
Dec 12, 2019
2,273
1,277
7,560
The issue seems to be happening on all cards. Also FE, as tested by Paul's Hardware in this video that was uploaded yesterday. It seems like some will just have issue's with it sooner than others.
View: https://www.youtube.com/watch?v=BkTgYMlCl4E&t
No, not really. None of his cards were unstable a stock speeds during his testing according to him. They all crashed when he overclocked them too far which isn't a revelation. Every video card ever made will crash if you overclock it too far. There is nothing useful that can be concluded from this video.
 

yourilevoye

Reputable
Jan 6, 2017
221
10
4,715
No, not really. None of his cards were unstable a stock speeds during his testing according to him. They all crashed when he overclocked them too far which isn't a revelation. Every video card ever made will crash if you overclock it too far. There is nothing useful that can be concluded from this video.

Sorry if I gave the impression it was about the cards on stock speed. hannibal mentioned that it was unsure if the FE seem to have "the issue", whcih the video of Paul showed that basically "the issue" is the same on all cards, but on some sooner than others.
It did get reported that some cards (personally only seen this with the peopel who bought the MSI Ventus early) had this issue out of the box, but this wasn't seen in the video, which also concludes that this either a very rare case, or those cards are updated, or the cards in the video don't have this issue at all.

The video showed that it doesn't matter what capacitors are being issued, leaning towards the impression it's something more/else than that.
 
  • Like
Reactions: hannibal

spongiemaster

Admirable
Dec 12, 2019
2,273
1,277
7,560
Indeed.
Here a statement from Nvidia around 14 hours ago:

"Regarding partner board designs, our partners regularly customize their designs and we work closely with them in the process. The appropriate number of POSCAP vs. MLCC groupings can vary depending on the design and is not necessarily indicative of quality."

I figured many people just hopped on the capacitor train, point at them because some people say it's the issue, while the people themselves don't really understand what the issue is.

It's good for Nvidia to give AIB enough time to test indeed, so I hope they learn from this. The 2000 series release wasn't 100% perfect either.
Most people say if you keep things stock you will be just fine. At the end, if a CPU doesn't really reach an OC that was not advertised anyway people won't really complain either honestly. This does exclude the cards that did have issue's with stock settings though, although many haven't.
This is a new product. There are going to be problems and they all aren't going to be caused by the same thing. More than one AIB is confirming caps caused some problems. One of the AIB's (Colorful? don't remember offhand) recalled their review units and replaced them even before the cards were released, supposedly because they had found the cap problem before it went public. You're not going to recall a review sample for something that can be fixed with software. So there is definitely something going on with the caps, but they aren't the cause of all the problems. Boosting didn't seem to be working correctly, with boost speeds spiking to unstable levels which caused problems with all cards. Apparently, adjustments were made in the just released drivers leading to lower boost clocks and better stability.

These cards have sort of been available for less than 2 weeks. There are likely to be more problems found. People will preach that the sky is falling and don't ever buy anything new, but with Nvidia at least, the hope is that most problems will be fixed relatively quickly. I doubt these early problems will still be widespread a year from now like is sometimes the case with other companies.