Rtx 3080 lower quality capacitor Issue

[Dumphole]

https://videocardz.com/newz/manufacturers-respond-to-geforce-rtx-3080-3090-crash-to-desktop-issues

 

[TearSoup]

Not all SPCAPs are equal. EVGA may well have found issues with the caps they were using, doesn't mean the other manufacturers had the same problem. Besides which, are we supposed to take a large corporation at their word? I mean think of the incredible spin potential. Cards delayed and customers angry -> Tell customers that actually you're behind because you're improving the quality unlike those other guys -> Customers love that you did this and are now happy to wait and also consider your product to be better than competition.

Not one EVGA shipped with the all SPCAP solution allegedly so if that's true there'll be no issues on those cards then right? Let's wait and see on that.

Hardware reviewers get pre production samples, with incomplete drivers/bios etc. They expect crashing, but they also know that it's not likely to be an issue in consumer cards and are usually told "we know these have some issues but they will be fixed before launch" and they go along with this. Whether this is morally right or not doesn't matter but that's why they come out after the fact saying x product crashed in testing for us and we didn't report it at the time of review.

Hardware unboxed have stated that the have had the crashing issue on an Asus TUF which features 0 SPCAPs. So no one can point at the caps and say it's definitely them.

On a positive note, this could mean that more people cancel their preorders on the cards with only 1 MSCC. When I receive it and I see crashes and instability with actual empirical evidence on the source of the issue I can either return it or RMA it if need be. Win-win.

 

[gbgt2]

On a positive note, this could mean that more people cancel their preorders on the cards with only 1 MSCC. When I receive it and I see crashes and instability with actual empirical evidence on the source of the issue I can either return it or RMA it if need be. Win-win.

Only if you consider your time to be free. Alternatively you could wait for the vendor to fix their faulty product.

 

[cutterslade]

Here is another post which hit the nail on the head as it looks
https://www.reddit.com/r/hardware/comments/j09yj5/comment/g6r7ce2

Still no explanation why some people have seen Asus cards also crashing.

 

[pc-guy]

Get the cards put under a scope isnt easy. You need the scope and open bench setup and know how.

no one has the circuit board design so it is a guess as to where you stab the contact. But buildzoid has identified a couple of contacts look like a good start.

anyway I think the reviewers don’t have expertise or don’t want to go into this level of technicality for whatever reasons ( maybe cheesing off their financial sponsors). Then the average Joes who have cards and just don’t have a clue. Or someone maybe have the card and does have a clue but frankly cba with it all as not wanting to void the warranty of the card.

Let’s face it, if you paid for the card yourself you don’t want to void that warranty by ripped your card apart and start probing it. Even if you are wealthy beyond imagination, it’s not like you can get another card readily atm if you damage it.

 

[Grim5]

Here is another post which hit the nail on the head as it looks
https://www.reddit.com/r/hardware/comments/j09yj5/comment/g6r7ce2

Still no explanation why some people have seen Asus cards also crashing.

My theory: many cards boost past the point of stability and the fewer MLCC units the card has the quicker it passes the point of stability.

Probably we just need a new driver to fix the issue and many cards will boost to lower clocks on the new driver

 

[SeeNoWeevil]

My theory: many cards boost past the point of stability and the fewer MLCC units the card has the quicker it passes the point of stability.

Probably we just need a new driver to fix the issue and many cards will boost to lower clocks on the new driver

Peak boost on these cards is only seen very briefly during spikes on light workloads and waaay past reference boost spec of 1710Mhz. They generally spend most of their time quite a bit lower. Like buildzoid said, "What's the problem? Most of these cards aren't boosting past 2Ghz" but then seeing even the base cards can spike very briefly to high clocks. Which would probably be enough to cause a crash. Limiting these spikes would no doubt have next to no performance repercussions. I think personally the fix is going to be fairly trivial and everyone will forget about all this. Although it seems like there could be a bunch of driver related problems thrown into the mix to confuse matters too.

Also, if Nvidia's own spec allowed either capacitor, I don't think it's very accurate to describe an all sp-cap design as 'cheaping out'. All in, I imagine this topic is pretty infuriating to listen to for people with electrical engineering knowledge.

 

[GioUK]

I think there are too many ppl out there (youtubers) saying bananas about what haven't got a clue. First MLCC were the the expensive ones, then started calling manufactures cheap for not using then. Now, they are the cheap ones.
There is a problem indeed, but no one have an idea what is, not even the manufactures apparently. Anyway, as someone already said, if the boost is 1750, why the hell is hitting over 2000. If this is the reason, reduce the max boost. Would you will be crying if the boost would be exactly as advertised??

My MSI Ventus with 1 MLCC is 100% fine and I have seen ppl with Zotac with 0 MLCC with zero issues. As I have seen Asus TUF with 6 MLCC having crashes. NO ONE KNOWS what to blame heheheh. If the 2000mhz is the problem, limit to not go over. Theses cards are a the best. I am telling you, a lot of ppl buying WILL not use 100% of the card. This card will use its full potential only on 4K.

I know my cpu is the bottom of the line for AMD(3600, changing soon for a better one). But it is bottlenecking on 5820x1080 (Triple Screen). How many 1080p users is how there who will use the 3080. I bet a lot.

 

[Kaapstad]

My theory: many cards boost past the point of stability and the fewer MLCC units the card has the quicker it passes the point of stability.

Probably we just need a new driver to fix the issue and many cards will boost to lower clocks on the new driver

Don't need a new driver at all.

The manufacturers need to fix/replace the cards not nerf them.

As this is a NVidia screw up they really should be picking up the bill.

 

[KompuKare]

Probably we just need a new driver to fix the issue and many cards will boost to lower clocks on the new driver

Meanwhile all the reviews were done with the higher boost...
Does anyone have any details about how Nvidia's boost algorithm works?
Does it take into account the ASIC quality which it can read somehow?
Because I thought TPU's GPU-Z utility and its ASIC quality reading wasn't supposed to be very useful for Nvidia cards.
With the size, complexity and huge transistor count of GPUs, surely any rating has to be made out of multiple parts anyhow?
More like a table with leakage etc. for each SM /CU and so on.

 

[cutterslade]

This problem may have an impact on overclocking - we wont be able to pass through 2 GHz. Most likely, there will be only a few cards capable of doing so, anyway. But still...

Reviewers have also stated how bad those cards are to overclock in general (max 3-4% more performance increase). Maybe on overclocked cards it is more likely that the GPU boosts beyond 2 GHz and just crashes, so the overclock looks unsuccessful.
A solution might lead to better overclock capabilities, who knows.

 

[cutterslade]

About the ASUS one crashing:
I think the reviewers got mixed cards, some with MLCC capacitors and some without. Might be an amateur explanation why they are reported to be crashing.

 

[SkinSpotlights]

About the ASUS one crashing:
I think the reviewers got mixed cards, some with MLCC capacitors and some without. Might be an amateur explanation why they are reported to be crashing.

Hardwareunboxed that claims ASUS Instability has the full array of MLCC Capacitors just like all the retail units did, there was no mix.

Don't need a new driver at all.

The manufacturers need to fix/replace the cards not nerf them.

As this is a NVidia screw up they really should be picking up the bill.

The problem lies with a mix of silicon quality, capacitors (because EVGA said they had to change the design so yes it can be a cause), factory overclock and GPU Boost.
Easiest fix is GPU Boost nerf which would be a driver update otherwise all the manufacturers would have to do a huge recall of current stock and flash a lower factory overclocked bios on them so GPU Boost doesn't boost them to that crashing point.

 

[Simon Belmont]

Don't need a new driver at all.

The manufacturers need to fix/replace the cards not nerf them.

As this is a NVidia screw up they really should be picking up the bill.

Yes. Nvidia’s greed is to blame. I hope they suffer.

 

[Adzassin]

After a bit more testing with the Gigabyte, i was getting more CTDs during games when boosting past 2ghz with an overclock. However there was actaully very little difference in boost clocks whether at stock or with a small 50-75mhz overclock, as a previous poster i think stated with his card. Also worth noting that i couldnt extend the power limit of this card in Afterburner past 100%.

So just decided to leave at stock and let it sit with a stable 1965 boost for now, and decide later whether its worth to return or not as more info comes out.

 

[KiNgPiN83]

So basically what we are seeing is that these Nvidia cards are maxed out as standard? If the only way to fix them is to downclock them that's bad news. I really really really hope AMD are taking notes right now... Hopefully the people moaning about them taking ages to release their product on their set schedule also now see why not rushing is important.

 

[Joxeon]

My guess is that they didn't have time to properly bin out the dies so you got high quality dies on low speced cards without the power delivery to cope with it.

 

[Nexus18]

Hopefully people will learn their lesson but alas, sadly the mentality of "must have shiny new right now!!!!!" and brand loyalty will still blind people's judgement.

 

[TrM]

Don't need a new driver at all.

The manufacturers need to fix/replace the cards not nerf them.

As this is a NVidia screw up they really should be picking up the bill.

so you have made your mind up on what the issue is on basis of a youtuber ? sorry but WE dont know what the issue is

 

[TrM]

How is it speculation if EVGA have confirmed?

https://videocardz.com/newz/manufacturers-respond-to-geforce-rtx-3080-3090-crash-to-desktop-issues

they confirmed they had issue's with SP-CAPS they dont give details of the whole power delivery system. They dont give any specifications of the caps used. silicon quality or anything of real substance. when gigabyte msi asus frounders edition zotac all use differnt Parts to make there system how can EVGA talk for all of the cards if everyone would use same stuff there wouldnt be there own designs and nvidia would just make all the boards for them