NVIDIA AIC Partners Clarify RTX 3080/3090 Crash to Desktop Issues, Capacitor Choices

[EarthDog]

The Nvidia Silence is deafening regarding this and the messy Rollout...

They already acknowledged the 'messy rollout' several days ago...like a day or two after the 3080 launch.

 

[Pumper]

As someone who never OC'ed my GPU, because I have always experienced crashing do desktop/BSODs with at least one game when overclocked even mildly, I don't really see this as an issue. The manufacturer only guarantees stability on stock specs, so if you OC your GPU and end up with problems, it's up to you to either roll back to stock or keep milking every last MHz out of it, but stop complaining about it.

 

[ZeppMan217]

They already acknowledged the 'messy rollout' several days ago...like a day or two after the 3080 launch.

That was in regards to availability, not hardware defects.

 

[Vayra86]

Please stop calling them POSCAPs. They're not POSCAPs. POSCAPs are Panasonic's Tantalum based caps, the caps in question are actually Aluminum based caps. So it'd be more appropriate to call them SP-CAPs, but that's still like calling any TV an LG or any wet paper tissue Kleenex. It's a brand name. Technically the generic term would be Conductive Polymer Aluminum Solid Electrolytic Capacitors.

Name fits the bill

P.O.S. caps

I just remembered Intel, when it was telling users that they should avoid overclocking their expensive brand new highly overclockable and unlocked i7 CPU.

Intel does it, Nvidia does it, AMD does it... over the past generations they have steadily been chipping away at 'our' overclocking headroom with clever boost algorithms, and they're now pushing things over the edge at times. As every overclocker knows, for 24/7 you want to nudge back one step from the maximum you can get... Seems like corporate greed doesn't really sit well with that idea. And all for single digit percentage wins over the other for the biggest epeen...

 

[fynxer]

Off cause they blame "...overzealous Boost algorithms ..."

SO OBVIOUS! NOW THEY ARE SOFTENING UP PEOPLE so they will accept a new firmware with nerfed boost clocks.

Lowering performance is the only way they can fix their poor capacitor choices without taking back the GFX card.

IF YOU HAVE A 3080, RETURN IT NOW WHILE YOU HAVE A CHANCE, DO NOT ACCEPT LOWER PERFORMANCE.

Some manufacturers made poor capacitor choices to make more money, it was greed pure and simple.

Who the flock saves a couple of measly cents on a $699 card, they must be real cheap.

What I heard is that ASUS and GIGABYTE went all in even on the cheapest versions, have anyone heard of any problems with their cards?

 

[steen]

What if the solution is GA103...?

Edit: 456.55 -> lower max boost + 10W = higher Vramp? Needs TPU/Igors test rig. PCAT is useless <100ms.

 

[Legacy-ZA]

"EVGA stands behind its products!"
— Jacob Freeman

Oh, that's strange, why is it that I experienced the opposite, in fact, a nice middle finger. Save it, buddy, you don't, not with all your customers, not for me. Or maybe it's only some countries that take preference?

 

[Mr McC]

Something is amiss here.

It can't be the fault of the AIBs, because Nvidia has to greenlight their designs and there are reports of the Founders Edition experiencing the same problems.

There are reports of cards that have not been overclocked crashing as well, it seems foolish simply to try to explain this away as a few individuals being overzealous with Afterburner.

Whilst Nvidia may not guarantee anything above 1710 MHz, they market the cards with boost, so if the "solution" requires downclocking the cards or limiting the boost, they have engaged in false advertising.

It is too soon to say what is at play here, but there are too many people jumping to Nvidia's defence: whatever happens, they are not going to come out of this smelling of roses.

 

[EarthDog]

That was in regards to availability, not hardware defects.

Uhh huh... and the guy I replied to specifically mentioned "this" (= CTD issues and what this thread is about) AND a messy rollout. Hence why I responded saying they already mentioned the messy rollout part days/week+ ago. With me?

 

[Mr McC]

Boost will not go over 2GHz from 1710MHz.

It is not clear that the issue is only affecting cards that go over 2 GHz.

 

[KarymidoN]

Something is amiss here.

It can't be the fault of the AIBs, because Nvidia has to greenlight their designs and there are reports of the Founders Edition experiencing the same problems.

There are reports of cards that have not been overclocked crashing as well, it seems foolish simply to try to explain this away as a few individuals being overzealous with Afterburner.

Whilst Nvidia may not guarantee anything above 1710 MHz, they market the cards with boost, so if the "solution" requires downclocking the cards or limiting the boost, they have engaged in false advertising.

It is too soon to say what is at play here, but there are too many people jumping to Nvidia's defence: whatever happens, they are not going to come out of this smelling of roses.

Thats what i don't get it. Partners can't just grab the chip and put it on a cardboard and sell it. Nvidia has Specifications, guidelines, strict details about componentes and minimum specs that can be used with their chips, every model has to be aproveed by Nvidia Engineers, not just the partners Q/C.

Other thing that people are not realising, even if you're not overclocking your card, by default if your system is cool enough and have enough power budget your GPU will boost during certain workloads, a lot of people are reporting crashes during the Automatic GPU Boost, thats not ok, some AIB models will boost close (and over) 2ghz (as shown in @W1zzard reviews), what a mess

 

[DeathtoGnomes]

Someone has to say it.

Just blame Furmark for blowing up the caps !

-----------------

Something is amiss here.

It can't be the fault of the AIBs, because Nvidia has to greenlight their designs and there are reports of the Founders Edition experiencing the same problems.

There are reports of cards that have not been overclocked crashing as well, it seems foolish simply to try to explain this away as a few individuals being overzealous with Afterburner.

Whilst Nvidia may not guarantee anything above 1710 MHz, they market the cards with boost, so if the "solution" requires downclocking the cards or limiting the boost, they have engaged in false advertising.

It is too soon to say what is at play here, but there are too many people jumping to Nvidia's defence: whatever happens, they are not going to come out of this smelling of roses.

Hopefully Nvidia doesnt have to go thru the false advertising bit again.

 

[Chomiq]

Based on that input from AIB's one thing is for sure - it's SP-CAP and not POSCAP.

 

[birdie]

Here's a clarification for y'all:

The cards are 100% stable at stock and start crashing when people are OC'ing them too much which has always been the case. I mean the beast already consumes 320W, why would you want to add 2% OC on top of that while increasing your power consumption by additional ~ 50W?

 

[Deleted member 50521]

Here's a clarification for y'all:

The cards are 100% stable at stock and start crashing when people are OC'ing them too much which has always been the case. I mean the beast already consumes 320W, why would you want to add 2% OC on top of that while increasing your power consumption by additional ~ 50W?

These cards have to generate power to feed the whole system and some folks would still complain about it. Internet mob mentality.

Last time I checked it took the mob half a year to accept that Navi had driver stability issues. I mean it is not like most of the vocal folks gonna go out to buy Ampere anyway so who cares.

 

[Xaled]

please order by manufacturer name, alphabetically, thank you!

There is no rule, even unwritten, that forces to do so, he can even sort them, at worst case , however he wants.

A friendly reminder to all the fanboys

Do you include yourself by the reminder?

This is overclocking only issue

Your source?

Cherry picked cards video

You want people to pay more for, custom, OC versions and ask them not to overclok?

Now this is new, but expected from Fanatic Nvidia users

They already acknowledged the 'messy rollout' several days ago...like a day or two after the 3080 launch.

I agree. Also partners cant open their mouth without Nvidia's permission

 

[BorisDG]

@techpowerup Buildzoid told ya it's not POSCAP... come on guys.

 

[asdkj1740]

there are more and more evga xc3 users complainning about crashes...

 

[Deleted member 50521]

there are more and more evga xc3 users complainning about crashes...

source?

 

[Raevenlord]

Based on that input from AIB's one thing is for sure - it's SP-CAP and not POSCAP.

EVGA uses POSCAP in their communication.

Gainward and GALAX use SP-CAP.

They are either using the terms interchangeably or they know what they're talking about.

Either way, I've updated the story so that any mention of POSCAP that's not a quoted statement now reads POSCAP/SP-CAP.

 

[birdie]

source?

"I've seen seen some people on the Internet who cannot even prove they own the cards in question".

I'm pretty sure this whole situation is blown out of proportions by select few who either recklessly OC their cards either manually or by autoOC'ing them in MSI Afterburner (which uses a very simple OC'ing algorithm which cannot really guarantee stable OC) or have lousy PSUs.

Meanwhile we have two reviews from verified owners and they don't report any stability issues: https://www.newegg.com/evga-geforce-rtx-3080-10g-p5-3885-kr/p/N82E16814487520

 

[hathoward]

This guy found something interesting as well:

Skip to 11:32

Everyone is probably mum because no one is yet 100% certain what the issue actually is.

 

[Cheeseball]

The new 456.55 driver may be helping mitigate this issue.

 

[birdie]

Here's another interesting video:

If it's really the case NVIDIA might fix or already have fixed it in drivers. It almost looks like NVIDIA's secrecy in terms of not allowing OEMs to have the actual drivers before releasing the cards have hurt NVIDIA.

 

[john_]

Intel does it, Nvidia does it, AMD does it... over the past generations they have steadily been chipping away at 'our' overclocking headroom with clever boost algorithms, and they're now pushing things over the edge at times. As every overclocker knows, for 24/7 you want to nudge back one step from the maximum you can get... Seems like corporate greed doesn't really sit well with that idea. And all for single digit percentage wins over the other for the biggest epeen...

Yeah, why give it to you (the customer) as an extra feature, when they can sell it to you as an extra feature and at the same time help their models move up on the performance charts in hardware sites?