Rtx 3080 lower quality capacitor Issue

[pBay]

Starting to regret changing my pre-order from the TUF to the Vision OC! Who knows if they will alter the config on that being a later model, or whether they just dressed up an Eagle to look fancy.

 

[TheOnlyRyan]

I have a Ventus 3X on preorder, I think I was bout 750 in the queue, I'm hoping any issues in manufacturing are ironed out by the century I get my card aha.

At the very least I'll have 1 of these little doodads.

 

[Dave2150]

Rushed launch, last minute drivers, not enough time for proper testing and bug fixes. Seems like Nvidia deliberately screwed their AiB partners. #Maybe this is why certain models of the the 3080's aren't out yet, as they caught the issues there?

I wouldn't blame Nvida, since every card they've sold has had the 6 good quality caps on it. If Asus can do it, why can't the rest?

 

[3t3P]

I think this is like in audio gear people try to predict how an amp will sound Vs another based on capacitors used and other componentry when it's ludicrous to do so.

 

[Electronic Punk]

This might all still be a driver issue *crosses fingers*

 

[opethdisciple]

Hmm... There was a video I watched yesterday about the situation in the Canada and the guy said the shops he visited in the whole of Vancouver only had like 3 for the entire city and apparently the cards got 'recalled' for a bios update.

He said at the time it sounded like excuses. But in light of this thread maybe it now makes some sense.

 

[Inferno99]

https://www.youtube.com/watch?v=pKi-T88jHRs&t=4m49s
I don't know the credibility of this statement, but it could mean that pre-release firmware may have fixed an issue for some partners. Then again, could be totally unrelated.

 

[opethdisciple]

https://www.youtube.com/watch?v=pKi-T88jHRs&t=4m49s
I don't know the credibility of this statement, but it could mean that pre-release firmware may have fixed an issue for some partners. Then again, could be totally unrelated.

Yea that's the video I was talking about.

 

[Aretak]

I wouldn't blame Nvida, since every card they've sold has had the 6 good quality caps on it. If Asus can do it, why can't the rest?

The FE has four of the cheaper caps and two of the more expensive arrays. That goes for the 3090 FE too. So clearly you don't need six.

 

[whippyice]

hummm,
got the XC3 on Pre order still (guesstimate that I'm around 16 in the queue,)
kind of hope a respected version arrives before they send mine out.

 

[Inferno99]

https://www.overclockers.co.uk/foru...egards-to-the-1-3-lower-performance.18899501/
Thinking this could also be related. Could it be that the capacitor configuration is an intentional design to limit the capabilities of the GPU?

 

[SeeNoWeevil]

People are going to look pretty stupid when this turns out to be something unrelated after being so confident they've found the cause for these 3080s crashing and going on a witchhunt over capacitors.

 

[Aretak]

https://www.overclockers.co.uk/foru...egards-to-the-1-3-lower-performance.18899501/
Thinking this could also be related. Could it be that the capacitor configuration is an intentional design to limit the capabilities of the GPU?

Alternatively, it could be that Zotac limited the card after discovering the issue in their internal testing. AIBs apparently only had a very limited amount of time to do internal testing due to Nvidia being late with drivers, so it may be that they found it late and the best they could do in the situation was a band-aid BIOS fix that lowers the boost slightly to prevent the instability from occuring (providing you never overclock it).

People are going to look pretty stupid when this turns out to be something unrelated after being so confident they've found the cause for these 3080s crashing and going on a witchhunt over capacitors.

Still, at least those people are actively looking for reasoning and solutions rather than trying to set themselves up to look like the smartest guy in the room via contrarianism.

 

[Rroff]

As I said in the other thread not sure this is the problem - I'm far from an expert on them but it doesn't make much sense to me - usually these capacitors are there to extend the tolerances of the GPU and it will generally work even without them - even a basic capacitor would give considerable filtering capabilities here. The symptoms of them not being upto the task would normally be more in the realm of crashing 1 in 10,000 times kind of thing. The MLCC types tend to be a bit better for filtering high frequency noise/instantaneous current demand and the tants a bit better for help with longer demands for power (though that would mostly be handled elsewhere - but the tants are more local to the demand). Then there is a whole load of ESR related stuff that I barely understand.

On a crowded PCB like Ampere I can see these capacitors being more critical but I would still expect issues related to them being insufficient or even missing to be more the order of very random and fairly rare crashes not something that is easily reproducible or someone has really screwed up on the design and you'd likely see the FE cards have some degree of symptoms of it as well.

EDIT: My guess here would be that there is a fairly big variation in individual core ASIC quality and the binning of cores for OC models is insufficient to catch the full range resulting in some with marginal stability. (Hence why some people are finding you can turn down the voltage quite a lot with little or no impact on stability and performance).

 

[CSClemens]

JayZTwoCemts has updated to say that it isn't just MLCC and POSCAPS. There is a third possibility. Some believed to be POSCAPS may be SPCAPS which should be better than MLCC. So POSCAPS may be an issue but these charts categorizing them as POSCAPS and MLCC only may not be accurate, because they are based on a visual identification that can't distinguish between POSCAPS and SPCAPS. Looking at the capacitors is not an accurate method to determine if your card is at risk.

 

[chrcoluk]

It sounds like some cards with the cheaper caps are actually fine at high clocks, and some of the ones with the correct config have been crashing. So no idea what is going on!

Igor suggested there is potentially two issues here.

One is OC cards having potato GPU chips on them (not binned properly or at all). This one seems either downplayed or not mentioned at all by the reviewers.

The other is the capacitor issue, lower binned chips will need more voltage/power to hit same clocks so naturally would stress these capacitors more.

 

[chrcoluk]

As I said in the other thread not sure this is the problem - I'm far from an expert on them but it doesn't make much sense to me - usually these capacitors are there to extend the tolerances of the GPU and it will generally work even without them - even a basic capacitor would give considerable filtering capabilities here. The symptoms of them not being upto the task would normally be more in the realm of crashing 1 in 10,000 times kind of thing. The MLCC types tend to be a bit better for filtering high frequency noise/instantaneous current demand and the tants a bit better for help with longer demands for power (though that would mostly be handled elsewhere - but the tants are more local to the demand). Then there is a whole load of ESR related stuff that I barely understand.

On a crowded PCB like Ampere I can see these capacitors being more critical but I would still expect issues related to them being insufficient or even missing to be more the order of very random and fairly rare crashes not something that is easily reproducible or someone has really screwed up on the design and you'd likely see the FE cards have some degree of symptoms of it as well.

EDIT: My guess here would be that there is a fairly big variation in individual core ASIC quality and the binning of cores for OC models is insufficient to catch the full range resulting in some with marginal stability. (Hence why some people are finding you can turn down the voltage quite a lot with little or no impact on stability and performance).

Yep what I said in another thread as well, I remember my EVGA 970 FTW, crashed on its shipped configuration, this problem has happened for multiple generations, but it seems in this generation its affecting way more people, I think this is because the shipping config is pushing the chips very close to their limits, so the binning quality perhaps needs to be quite high to get the stability.

 

[Rroff]

Igor suggested there is potentially two issues here.

Yeah it has to be more than just this - if it was just down to these capacitors, even if due to AIBs being cheap, it would be someone like me doing the engineering who knows enough to be dangerous not the work of someone (in reality a team) with years of engineering experience and qualifications.

 

[Chapchap27]

Palit 3080 GamingPro has one MLCC. Have not had any crashing either.

 

[ChrisUK1983]

https://forums.evga.com/m/tm.aspx?m=3095238&p=1

EVGA admit it was an issue and the cause of the delay in getting their FTW3 cards out. Don't know why they couldn't have said that initially rather than blaming "unprecedent demand".