GeForce RTX 3080 CTD Issues likely due to POSCAP and MLCC configuration

quick tests to try to replicate the crash reports, high mhz before temp rise for the boost be the highest possible:

full hd, boosts above 2100
https://i.ibb.co/KGX4Hz3/3080-456-55.png

rendered in game 4k, boost jumps between 2080 and 2040, no more 2100+ boost haha, again, no stability, crash, problem whatsoever, but now, like print shows, perfcap hits hard as the card power consumption raised

 

In multiple Spanish stores, the pictures of GPU have the new changes. I have downladed the file and the metadata says that it's from 14/09/2020, so maybe they knew it from some time ago and in Europe they will directly sell only the updated revision.

 

Here's a link to the updated MSI GAMING X TRIO 3080 product page on MSI's website , notice the 2 top quality yellow MLCC on the backplate

https://cn.msi.com/Graphics-card/GeForce-RTX-3080-GAMING-X-TRIO-10G/Gallery

 

Nvidia think they might have fixed it with a driver update. They also suggest that the caps thing might be a red herring.

https://www.nvidia.com/en-us/geforc...x-3080-board-stability-new-driver-capacitors/

 

I dont think they are suggesting it might be a red herring. Seems to me they just want to give the appearance they are addressing it but not really wanting to clarify it any better. Its also why the topic is "locked".

"The appropriate number of POSCAP vs. MLCC groupings can vary depending on the design and is not necessarily indicative of quality..."

That may be a truthful statement but its tempered with the words "not necessarily".

 

Corporate lawyers don't allow absolutes. This is America man, get with it.

 

remember when some youtube (jayz ?) explained that AIBs were only provided a pass or fail basic test and no driver ?
remember that when cooling designs had already been released for pretty much all models they still didn't have drivers
remember that this time nvidia wanted to be the no1 videocard seller because keeping their best bins wasn't enough already ?

so now we have AIBs that clock lower are based on an inferior PCB and have troubles, but still sold for more money, thanks nvidia
100% their fault
they only forgot that screwing up your partners make the entire product look bad and now no one following videocard news wants one, great job

btw guys you can't take their websites pictures as facts, it's marketing has to go through 2-3 departements before being updated, it's meaningless EDIT : it does prove tough that they have changed the capacitors at one point unlike what they say
and....my system updated the drivers again and now i get 2x more crashes on my 1080ti time to roll back again, drivers still a problem

 

Sorry for the quality, not my plans being youtuber, no light and with smartphone recording hehe... "issue installed , tested and repaired" in 10min , better quality uploading !

 

Thanks. Interesting that vcore didn't go as high after the driver update.

 

So. Safe to say it's an end user issue as it only happens when overclocking the GPU.

So no manufacturer has to honor a warranty because the end user is taking the hardware out of spec. Good things!.! HA!!

 

Ok now i understand the issue a bit better.

those are the specs as reported by guru3d, 1710 is the boost clock.
The 3080 is targeted at 1440/1710.
Over 1710 could work, but it doesn't.
What are we even talking about?

 

Updated to 465.55 but still having completely crash and freeze up issue.

The only working workaround for now is to down clock by 50MHz.

P.S. Zotac Gaming RTX 3080 Trinity

 

I have a question here, never owned the rtx 2xxx series before.

In regards about the unstable boost clock that's supposedly causing the games to crash, is there a way to keep 3D base clock in conjunction to your boost clock???

I know with the maxwell bios editor II, I can keep it together, for instance, running @ 1506 mhz on the 970's with other voltage tuning, I can keep it that speed range.

Has anyone bios edited these cards yet? I know its still extremely to early but was curious have anyone done so yet.

I wonder if the bios setup is the same as the 2xxx series.

If you can keep your 3D base clock speed together along with the boost clock speed, it may help stabilize the system, again, never owned the 2xxx turing gpu's so can't say.

 

This men has balls!


It's a nice demostration, confirmed is risky work, but it's a very good work

 

How it safe to say that?
What evidence do you have that makes this a safe statement?

 

Those are minimum boost clocks, the cards automatically boost a lot higher.
Problems have been found in some cards that boost higher than 1950MHz, even more above 2000MHz.

 

Well it seems that the `minimum` is the minimum in spec, all the rest is for free. Or isn't.
I think a lot of noise for something that should be solved with a bios update limiting the boost clocks to 1900.

 

This might help your rather closed understanding.
https://wccftech.com/nvidia-officia...ries-crashes-issues-sp-cap-vs-mlcc-groupings/

 

Arghh, re Update #4, the driver update that limits the GPU clock - now that is the wrong kind of band aid! I don't want my cards to be limited by the driver, I want to be able to overclock them and let them run free according to the stability of the silicon I might have lucked out within my GPU.....some kind of artificial/arbitary frequency caps put in place by the driver is just tying you down, applying a one size fits all to the problem, and that sucks...they need to sort the problem in the hardware.