Let's talk about Nvidia's RTX 3080 CTD Issues - Are SPCAPS and MLCCs to blame?

[Dicehunter]

I have a theory as to why these issues are popping up and it may be a feasible one.

Apparently Nvidia, Or some of their engineers, Have seen what RDNA2 will deliver from a price/performance standpoint, Lisa Su did say RDNA will do for their GPU's what Zen did for their CPU's, From a business perspective if you are used to being number 1 and you see this situation unfolding that will make you a little pale and your shareholders a little jittery.

With that in mind I think Nvidia wanted to get these out as fast as possible even though components were spread thin among the various AIB's due to the human malware issue and factories only working at 10% capacity, The components their board partners could get hold of were sparse so the factories that churn these out quite likely had to forego their usual thorough quality control, Not just the surface mounted components but everything that goes with it, Traces, PCB quality etc...

Now the FE cards aren't suffering from these same issues and Nvidia always source the best components for themselves as well as no nonsense PCB designs to ensure rock solid cards as well as keeping back the best silicon for themselves.

It makes sense and it's not really too "out there".

 

[KingNosser]

There are many factors involved with the production and shipping in all that kind of thing atm, so yea that is understandable from how many can actually get into peoples hands, but they should be functioning GPU's regardless, it's most likely a combination of things.

Where Nvidia went really wrong and it's a fact, is they tried to play price war football with TSMC and TSMC didn't break down and say oh sure Nvidia were cut you a deal better price, so they went with samsung, TSMC know that their 7nm is a high yield high demand why would they cut any deal other than the exact price for everyone else.

RDNA2 isn't just going to be strong it could be so strong that the 6900 will beat a 3090 then what is nvidia going to do, and if AMD are agressive like they were with Intel taking into account the gddr6 rather than X and less ram but performing better and under cutting the 3090 by a long way, well it'll be really bad for Nvidia.

tbh regardless of Nvidia's launch and issues all AMD need to do is to get as smooth as launch as possible, the best drivers they can with as few issues as possible and meet a good amount of demand, then Nvidia is going to be in a real tight spot.

I'm sure they will get on top of the issues over time, but Nvidia is going to need to do a lot better going forward, the 3070 launch is pretty much a given that it will end in the same way but i can 100% say now it won't be a 2080ti matching card once them fps numbers come out, if i'm dead wrong and it's even slightly better then fair enough, but i think Nvidia are not in a good place after them two launches.

 

[Dicehunter]

There are many factors involved with the production and shipping in all that kind of thing atm, so yea that is understandable from how many can actually get into peoples hands, but they should be functioning GPU's regardless, it's most likely a combination of things.

Where Nvidia went really wrong and it's a fact, is they tried to play price war football with TSMC and TSMC didn't break down and say oh sure Nvidia were cut you a deal better price, so they went with samsung, TSMC know that their 7nm is a high yield high demand why would they cut any deal other than the exact price for everyone else.

RDNA2 isn't just going to be strong it could be so strong that the 6900 will beat a 3090 then what is nvidia going to do, and if AMD are agressive like they were with Intel taking into account the gddr6 rather than X and less ram but performing better and under cutting the 3090 by a long way, well it'll be really bad for Nvidia.

tbh regardless of Nvidia's launch and issues all AMD need to do is to get as smooth as launch as possible, the best drivers they can with as few issues as possible and meet a good amount of demand, then Nvidia is going to be in a real tight spot.

I'm sure they will get on top of the issues over time, but Nvidia is going to need to do a lot better going forward, the 3070 launch is pretty much a given that it will end in the same way but i can 100% say now it won't be a 2080ti matching card once them fps numbers come out, if i'm dead wrong and it's even slightly better then fair enough, but i think Nvidia are not in a good place after them two launches.

All sounds agreeable minus the 3090 beating part, Don't get me wrong I REALLY want that to happen but I cannot see it, I can see performance being between a 3070 and 3080 which in itself coming from their highest performing card, 5700XT which matches a 2070S, Is damn good, All down to the price and drivers after that.

 

[looz]

Seems to me that this caused limited launch supplies as AIBs shuffled their cap configurations, and some review units ended up unstable.

Which is a separate issue from crashes caused by boost behaviour - but with limited data available, people jumped gun and even some youtubers went click hunting. And then there's crashes from lemon cards, inadequate PSUs and the like.

Seems pretty typical for a freshly launched product, early adopters being beta testers is annoyingly par of the course. However, if sub-par designs are still being sold and the driver update didn't fix the issue, it looks rather worrying.


EDIT: Weren't there rumours that AIBs were given very limite drivers, and couldn't run stability tests apart from Nvidia's own CLI tools?

 

[Warchild]

https://videocardz.com/newz/msi-qui...-gaming-x-trio-design-amid-stability-concerns

MSI redesign their PCBs.

https://videocardz.com/newz/asus-al...eforce-rtx-3080-tuf-and-rog-strix-pcb-designs

And so do Asus.

Exactly what I knew would happen. But expect to see the cost of the cards rise. AIBs will not let their XXX% markup change at all.

EDIT: Weren't there rumours that AIBs were given very limite drivers, and couldn't run stability tests apart from Nvidia's own CLI tools?

Correct. Nvidia allowed only certain softwares to be useable for testing. So prior to launch not even the AIBs knew how cards performed. This is why I think this CAP usage was known before launch but nothing could be done about it at the time, except limiting production while they fixed the issue.

 

[KingNosser]

Yep the prices on the readjustments will cost more the price of that change is actually a fair bit more once you add it all up, so the tuf version that was £649 will be more like £699 same for any others that use a full set but adding just one making it 2+4 would be a lot cheaper and also exactly what the FE cards use.

6900XT vs 3090 will be a far closer run thing, than you might think your litterally talking about a small % above a 3080 and AMD were aiming for 50% + performance over a 2080ti that would put it beyond that mark if they reached that and also at a lot lower wattage.

AMD didn't call it the Nvidia Killer for no reason, they are dead serious.

 

[Tolemac]

I'm merely speculating here but we all remember the GTX 970 fiasco and the R9 290X. I reckon this could be another case of what happened back then. Although they called the 290X a sun heater because of the reference blower cooler. The Sapphire AIBs "Tri X and Vapor X" were solid cards and did age very well.

 

[trawetSluaP]

Seems the latest drivers NV put out have stopped the crashes. Wonder if it really was that simple or we'll be seeing people crashing still.

 

[Warchild]

Seems the latest drivers NV put out have stopped the crashes. Wonder if it really was that simple or we'll be seeing people crashing still.

Well I dont know what happens if someone wants to overclock their card and push limits. Its not the same kind of crashing as it would be with the GPU chip itself I guess. Because even if you were able to pump more mV which would usually stabilise the GPU core, these cheaper caps would still be unstable.

 

[WYP]

If I'm honest, it looks like Nvidia had problems with their GPU-boost software. That's why a driver fix addressed things and why the issue only seemed to impact some software.

I can't speak for all GPUs, but not all of them used "cheap caps".

Here's Gigabyte's statement on the issue.

https://www.overclock3d.net/news/gp...ors_on_geforce_rtx_3080_3090_graphics_cards/1

 

[Warchild]

If I'm honest, it looks like Nvidia had problems with their GPU-boost software. That's why a driver fix addressed things and why the issue only seemed to impact some software.

I can't speak for all GPUs, but not all of them used "cheap caps".

Here's Gigabyte's statement on the issue.

https://www.overclock3d.net/news/gp...ors_on_geforce_rtx_3080_3090_graphics_cards/1

However... aren't they experiencing the issues on their other models which have all 6 POSCAPS?

I get what they say. But it does seem a little "too" coincidental that every GPU experiencing the crash has all 6 of the cheaper variants in use. And if we look across the market, all brands have one or two models affected. Its not like its an entire ASUS stack or all EVGA models.

It really does seem like its only those with the "cost cut" components. I say Cost cut, even though we know they are still pricey. But a great test would be to see how the newer MSI/ASUS cards behave on the previous driver.

Obviously, they know far FAR more than us armchair warriors and our OTT theorising, but so far the evidence adds up. And I trust igorslab far more than almost any reviewer on the net.

Lets use Igor lab pic as an example.

3080 Founders on left
Zotac Trinity Centre
ASUS TUF on the right.

There are no reports on this issue for the founders, and also none regarding ASUS TUF. There are however a massive amount of complaints of Zotac Trinity (sad that this is the one I reserved )

Now look at which one is using the lower priced Caps. We cant expect them to use the best of the best everwhere, but at the same time, they should not comporomise a good product by cutting costs everywhere and expect a smooth operation. There has to be balance. Seems like 5/6 was able to create better stability.

edit* Dice might even have a point that maybe component shortages and rushed launch led to this. Using whats available to get these out on the market.

I still believe that this instability was known from the start. Or at least AIBs were concerned. But they cannot be blamed fully as Nvidia limited their stress testing from the start, with locked down benchmarking software.

 

[WYP]

However... aren't they experiencing the issues on their other models which have all 6 POSCAPS?

I get what they say. But it does seem a little "too" coincidental that every GPU experiencing the crash has all 6 of the cheaper variants in use. And if we look across the market, all brands have one or two models affected. Its not like its an entire ASUS stack or all EVGA models.

It really does seem like its only those with the "cost cut" components. I say Cost cut, even though we know they are still pricey. But a great test would be to see how the newer MSI/ASUS cards behave on the previous driver.

Obviously, they know far FAR more than us armchair warriors and our OTT theorising, but so far the evidence adds up. And I trust igorslab far more than almost any reviewer on the net.

Lets use Igor lab pic as an example.

3080 Founders on left
Zotac Trinity Centre
ASUS TUF on the right.

There are no reports on this issue for the founders, and also none regarding ASUS TUF. There are however a massive amount of complaints of Zotac Trinity (sad that this is the one I reserved )

Now look at which one is using the lower priced Caps. We cant expect them to use the best of the best everwhere, but at the same time, they should not comporomise a good product by cutting costs everywhere and expect a smooth operation. There has to be balance. Seems like 5/6 was able to create better stability.

I have seen reports on ASUS cards having this issue. The problem can impact all models regardless of cap design. It is true that the problem is more prevalent on some designs.

With the Zotac, they use lower-end 330u SPCAPS for their Trinity model. Gigabyte uses 470U capacitors for their Gaming OC and Eagle OC designs.

Even then, Zotac exceeds Nvidia's reference specifications. The fact that a driver fixed this says a lot TBH.

 

[looz]

Exactly what I knew would happen. But expect to see the cost of the cards rise. AIBs will not let their XXX% markup change at all.

Those redesigns were done between review units and actual release.




Edit: Gigabyte released a statement:

Statement Regarding the SP-CAP and MLCC Capacitor on GeForce RTX 3080 Graphics Cards

In response to the recent reports speculating that the use of POSCAP capacitors on the GeForce RTX 3080 graphics cards could lead to stability issues and crashes, we would like to clarify the issue with the following statement:

It is false that POSCAP capacitors independently could cause a hardware crash. Whether a graphics card is stable or not requires a comprehensive evaluation of the overall circuit and power delivery design, not just the difference in capacitor types. POSCAPs and MLCCs have different characteristics and uses, thus it is not true to assert that one capacitor type is better than the other.

The GIGABYTE GeForce RTX 30 graphics cards are designed in accordance with NVIDIA specifications, and have passed all required testing, thus the product quality is guaranteed. GIGABYTE GeForce RTX 3080 GAMING OC and EAGLE OC series graphics cards use high-quality, low-ESR 470uF SP-CAP capacitors, which meet the specifications set by NVIDIA and provide a total capacity of 2820u in terms of GPU core power, higher than the industry’s average. The cost of SP-CAP capacitors is not lower than that of MLCCs. GIGABYTE values product integrity highly and definitely does not reduce costs by using cheap materials.

NVIDIA has released a driver (version 456.55) on September 28, 2020 that improves stability. Users are advised to update to the latest driver for optimized performance. For users who have any concern, please contact our local service centers or representatives.

GIGABYTE has been constantly improving and optimizing product quality, especially in terms of thermal designs, to provide the best gaming experience to the consumers for decades. For the latest AORUS GeForce RTX 30 graphics card series, we have also paid extra attention to the cooling performance and introduced industry-leading solutions such as MAX-Covered Cooling to ensure that the operation of each component is stable.