3080/3090 - the current issues, problems and solutions as of 30/09/2020

[LukeSavenije]

last update: 30/09/2020

 

Issue 1: CTD

The first issue seems to be a crash to desktop (CTD), where during higher loads the GPU would crash the game and return to the desktop. This mostly happened at clocks of 2ghz and higher.

 

Issue 2: OCP

The second issue is a hard shutdown on the PSU side, likely on the Over Current Protection (OCP). This looks very similar to cutting the power.

 

Issue 3: availability

The cards have been incredibly hard to get, some restocks going as far as next year already as estimated delivery date. Especially founders has been incredibly hard to get after launch.

 

Problem 1: capacitors

sources: Igorslab, Buildzoid

as indicated fairly well already by Igor's lab, owned by Tomshardware writer Igor Wallossek, the cards seem to have some issues, where the capacitors, mainly behind the core are insufficient, mostly blamed on the use of Polymer Tantalum capacitors, in Igor's article as POSCAPs, but this is actually a very specific series from Panasonic, not a general name for the type of cap. This is proven on production models of Zotac Trinity and MSI Gaming X Trio (has since been silently revisioned).

 

Problem 2: incorrect guidance from Nvidia

in the reviewers guide, Nvidia gave two options regarding the capacitors on the back of the GPU core, between MLCC and Polymer Tantalum capacitors. as Polymer Tantalum is cheaper to do, some have opted to try these, but have mostly been restricted to reviewer/pre-production samples and only in the case of the Gaming X Trio and Trinity have been proven on production samples.

 

Problem 3: statements from manufacturers regarding their GPUs

sources: Videocardz  , tomshardware, kitguru

 

Nvidia

Spoiler

Regarding partner board designs, our partners regularly customize their designs and we work closely with them in the process. The appropriate number of POSCAP vs. MLCC groupings can vary depending on the design and is not necessarily indicative of quality.

Colorful

Spoiler

The manufacturer was the first to report on an issue to the press. Review samples that had already been sent out, were later recalled.

EVGA

Spoiler

Recently there has been some discussion about the EVGA GeForce RTX 3080 series.

During our mass production QC testing we discovered a full 6 POSCAPs solution cannot pass the real world applications testing. It took almost a week of R&D effort to find the cause and reduce the POSCAPs to 4 and add 20 MLCC caps prior to shipping production boards, this is why the EVGA GeForce RTX 3080 FTW3 series was delayed at launch. There were no 6 POSCAP production EVGA GeForce RTX 3080 FTW3 boards shipped.

But, due to the time crunch, some of the reviewers were sent a pre-production version with 6 POSCAP’s, we are working with those reviewers directly to replace their boards with production versions.
EVGA GeForce RTX 3080 XC3 series with 5 POSCAPs + 10 MLCC solution is matched with the XC3 spec without issues.

Also note that we have updated the product pictures at EVGA.com to reflect the production components that shipped to gamers and enthusiasts since day 1 of product launch. Once you receive the card you can compare for yourself, EVGA stands behind its products!

Asus

Spoiler

ASUS has changed the design in pre-production. No official statement was made.

MSI

Spoiler

MSI acknowledged the issue during the MSI Insider Livestream, suggested it could be a driver issue.

Update: MSI has changed the design of the RTX 3080 GAMING X TRIO.

 

MSI sent over images and a statement disputing any claims that it has changed the design on its 30-series GPUs. Included are four images of the backs of the cards, showing the capacitor configuration. The MSI RTX 3090 Gaming Trio and MSI RTX 3090 Ventus 3X have a layout with 4x SP-Caps and 2x MLCC clusters, while the MSI RTX 3080 Gaming Trio and MSI RTX 3080 Ventus 3X have a layout with 5x SP-Caps and 1x MLCC cluster. These are the cards that actually shipped to consumers and reviewers; earlier variants may have existed but apparently did not ship to anyone outside of MSI.

Zotac (owned by PC partner)

Spoiler

Galax/KFA2 (owned by Palit)

Spoiler

About the SP-CAP capacitors and MLCC capacitors of GALAXY RTX 3080/3090 products

Dear player friends:

Hello, everyone. Recently, many users have come to inquire about the specific usage of the capacitors on the back of the GALAXY RTX 3080/3090 series of graphics chips. After verification, about the RTX 3080/3090 released by GALAXY. The capacitors used on the back of the model chip are as follows:

1. GALAXY RTX 3080 Heijiang/Metal Master product, the number of SP-CAP capacitors on the back of the chip: 5, the number of MLCC capacitors: a set of 10. This version is currently on sale and is the original commercial version.

2. GALAXY RTX 3090 General/Metal Master product, the number of SP-CAP capacitors on the back of the chip: 4, the number of MLCC capacitors: two groups of 20. This version is currently on sale and is the original commercial version.

3. GALAX RTX 3090 GAMER trial production samples, currently only 6 pieces are in the hands of the media and KOL. The first batch of this sample uses 6 SP-CAP capacitors. After confirmation, the GAMER products officially produced and sold will be used for capacitor materials. Make optimization improvements. Note: This product is not currently on sale.

I am very grateful to the players and friends for their support and love to GALAXY. GALAXY is also consistent in its pursuit of product quality. It is our glorious mission to provide you with better and stronger hardware. In addition, the current full range of GALAXY graphics card products support three-year warranty and personal warranty service. If you have other doubts or questions, please feel free to leave us a message to discuss, thank you!

Gainward (owned by Palit)

Spoiler

Announcement on SP-CAP Capacitors and MLCC Capacitors of Gainsun 30 Series Graphics Card Products

Dear Gainward consumer players:

Thanks to the friends who bought and supported Gainward. Recently, we received the voice of market players’ inquiries. Many players are very concerned about our company’s just released 30 series products. Regarding the specific usage of the capacitors on the back of the chip, we hereby explain the situation:

All the RTX 3080 10GB graphics cards released by Gainward currently use 5 SP-CAP capacitors on the back of the chip and 10 MLCC capacitors. The versions currently on the market are all the original commercial versions.

All the RTX 3090 graphics cards released by Gainward currently use 4 SP-CAP capacitors on the back of the chip and 20 MLCC capacitors. The versions currently on the market are all the original commercial versions.

As a long-term AIC partner of Nvidia, Gainward has always been adhering to the product standard to design and produce completely according to Nvidia’s requirements. Therefore, currently players are concerned about the problem of capacitors and new product failures on the Internet. Currently Gainward has not generated such feedback. .

In addition, all Gainward graphics card products support three-year warranty and personal warranty service. Thank you consumers and players for your support and love to Gainward.

Inno3d (owned by PC partner)

Spoiler

Gigabyte

Spoiler

“In response to the recent reports speculating that the use of POSCAP capacitors on the GeForce RTX 3080/3090 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 3080/3090 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 29, 2020 that improves stability. Users are advised to update to the latest driver for optimized performance. For users who encounter power-related issues with GIGABYTE GeForce RTX 30 series graphics cards, GIGABYTE will provide product replacement, free of charge.

 

“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.”

 

Problem 4: peak power

The third has been covered very little as of the time of writing, but there are indications that the 3080 and 3090 cards have a very high peak power usage, triggering some PSUs to make a hard shutdown. It's as of right now unknown if this is the fault of the capacitor choice or because of other reasons. Possible is that Nvidia has broken ATX specifications on transients. The following units might cause tripping on these cards:

Spoiler

- Seasonic Prime Titanium 650w - 3080 founders (tweakers.net)

- Seasonic Prime Platinum 850w (SSR-850PD) - 3090 founders (linustechtips 3090 review)

- pre January 2018 Seasonic Prime Gold 550/650 -not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- Seasonic Focus FM/FX/PX v1 - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- Bitfenix Formula 450/550 - too little 12v wattage on VGA rail

- Bitfenix Whisper 450/550 - too little 12v wattage on VGA rail

 - pre January 2018 Antec High Current Gamer Gold - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- pre January 2018 Antec Earthwatts pro Gold - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- pre January 2018 Antec Neo Eco Gold - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- pre January 2018 Asus ROG Strix - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- pre January 2018 Phanteks Revolt Pro - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- pre January 2018 Riotoro Enigma G2 - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- Seasonic G - not confirmed, but shown issues with vega, 1080 ti and 2080 ti
- Seasonic SSP-RT - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Seasonic S12G RT - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Seasonic G RM - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- XFX TS Gold - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- XFX XTR <=750W - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Antec True Power Classic - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Antec Edge <=750W - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Corsair HX650 - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Fractal Design Edison M - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- LDLC QS+ - not confirmed, but shown issues with vega, 1080 ti and 2080 ti
- SAMA Armor -not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Sama Forza - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Silentiumpc Supremo FM2 - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Silentiumpc Supremo L2 - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Silentiumpc Supremo M2 - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Vinga VPS-G - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- VIVO 24k - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- Modecom Volcano - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- NOX Hummer X Gold - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- iTek GF - not confirmed, but shown issues with vega, 1080 ti and 2080 ti

- pre January 2018 Huntkey MVP Gold - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

- pre January 2018 Huntkey MVP Platinum - not confirmed, but shown issues with vega, 970 strix, 1080 ti and 2080 ti

please do report in detail below in case you have similar issues, preferably with the following information

- load (which game/stresstest/program you were running before shutdown)

- psu model

- system specs

 

As of the time of writing, there's nothing conclusive on how high the peak power actually is, since measurements have been ranging from 340 to 750w of peak power in various testing.

 

Problem 5: pushed to it's limit
this is mostly personal speculation based off the data out so far on continued power usage, temperatures on the GPU core+GDDR6x and related information. Nvidia has pushed their cards both in thermals and power to a high limit, where there is little threshold for overclocking and possibly even causing stability issues on some models.

 

Problem 6: poor binning and/or too agressive boost?

This specifically is nothing more than just speculation. It could be that because of the rush that AIBs didn't have enough time to properly bin the cards, and let them turbo too high or too agressively, causing high peaks in power or even shutdown. If this is the case and would overwhelm the already limited capacitance on some cards, this would explain some of the issues, but this is as of the time of writing not proven in any way more than just speculation.

 

Problem 7: rushed launch

sadly, most of the issues will likely come down to this one issue. Nvidia has reportedly been rushing the launch for both themselves and AIB partners, where most, if not all issues can be traced back to. This is something already concluded, it's now just to find which problem links to which issue and how to fix it.

 

Solution 1: downclocking

As of right now, various AIBs have been pushing firmware that limit how far the card boosts, to hopefully give a temporary solution to people having these issues. This can hopefully be ruled out on a later date with more targeted software and hardware updates.

 

Solution 2: capacitor changes

This is, as of the time of writing done by both MSI and Zotac on their Gaming X trio and Trinity cards, replacing the capacitor design close to the core to counter the issues found by Igor's lab. Video from Der8auer related to this:

 

Solution 3: take some time

this is mostly a solution to problem 6. The launch should've gone more careful on Nvidia's side to allow AIB to have enough time for testing, for them to have some production ready for the first batch of consumers and so on. This is something Nvidia should've considered by starting earlier, or planning a later release date. But by the looks of it, they wanted to be quicker than AMD's RDNA2 cards... we'll soon see what they have to bring.

 

Note 1: (possibly) affected 30xx's

OCP issue - all possibly

capacitor problem - Zotac Trinity, early revision MSI Gaming X Trio confirmed, others possibly

peak power - all possibly

availability - all

pushed to it's limit - all possibly

3070 - unknown, release date at 15 October 2020

[Juular]

Yay, Luke isn't lazy !

[VeganJoy]

ah, you're missing solution 4: dont pay 800-1700 bucks to effectively beta test new hardware. they dont call it "bleeding edge" for nothing

[Moonzy]

27 minutes ago, LukeSavenije said:

Solution 2: capacitor changes

This is, as of the time of writing done by both MSI and Zotac on their Gaming X trio and Trinity cards, replacing the capacitor design close to the core to counter the issues found by Igor's lab.

derbauer just released a video

he replaced 2 "POS" caps with 10 MLCC each and only saw 20mhz improvement

so it doesn't really explain the 100-200mhz different between the cards

 

did anyone else swap caps and saw big improvements?

 

imo it's more to do with being overly aggressive with the clocks, and the variance in the capability of the die of 30 series is very large.

 

edit:

"POS" caps = Piece of Sh*t caps?

[GoldenLag]

@GabenJr would it be possible for LTT to do some exploration into issue nr 4. PSUs tripping that really shouldnt. 

 

seeing as:

1. you have a 3090 that trips the Prime
2. i assume you have a decent bit of PSUs on hand to test cards
3. its not that much of a covered topic due to being overshadowed by other more widespread issues. 

 

[LukeSavenije]

27 minutes ago, Moonzy said:

"POS" caps = Piece of Sh*t caps?

POSCAP: Conductive Polymer Tantalum Solid Capacitors, it's a couple lines from Panasonic

27 minutes ago, Moonzy said:

derbauer just released a video

I added it into the main post, just in case

[Haro]

I still haven't watched Build zoid's video because i'm lazy, guess i need to watch it now. 

thanks lol.

[FaxedForward]

Regarding crash causes, I have seen a surprising amount of people report crashes that ended up being caused by the use of a single “daisy chain” 2x8-pin power cable instead of two separate 8-pin  power cables. All of the pre-release info seemed pretty clear to me that you can’t use a “daisy chain” cable but apparently this is getting missed in the real world.

[Juular]

23 minutes ago, FaxedForward said:

but apparently this is getting missed in the real world.

People are morons, who knew ?

[ewitte]

42 minutes ago, Juular said:

People are morons, who knew ?

It should be apparent after the human malware experiment..

[Emily Young]

2 hours ago, GoldenLag said:

@GabenJr would it be possible for LTT to do some exploration into issue nr 4. PSUs tripping that really shouldnt.

This is on our radar.

[Ankerson]

1 hour ago, FaxedForward said:

Regarding crash causes, I have seen a surprising amount of people report crashes that ended up being caused by the use of a single “daisy chain” 2x8-pin power cable instead of two separate 8-pin  power cables. All of the pre-release info seemed pretty clear to me that you can’t use a “daisy chain” cable but apparently this is getting missed in the real world.

 

Well with way too many morons (idiots) telling people they can use one daisy chained cable on both the forums and YT it's not surprising to me. 

 

The PSU companies and NVIDIA etc are telling people not to do it and it's in the GPU literature. 

 

But it seems there are always those people on the forums and YT who think they know more than the ones who actually design and build the things. 

 

 

[Ankerson]

3 hours ago, LukeSavenije said:

As of the time of writing, there's nothing conclusive on how high the peak power actually is, since measurements have been ranging from 340 to 750w of peak power in various testing.

 

 

Likely one of the reasons why they specified a 750W min PSU in the 1st place...

 

 

 

 

[LukeSavenije]

43 minutes ago, GabenJr said:

This is on our radar.

coughs you can always send me a couple PSUs and a 3080

[vong]

So no word from Gigabyte? I'm queued for one of their Aorus master cards.

[Finalcountdown]

Very informative post 
Thanks for this LukeSavenije 

[LukeSavenije]

56 minutes ago, vong said:

So no word from Gigabyte? I'm queued for one of their Aorus master cards.

not that I've personally seen

[whaka]

1 hour ago, Ankerson said:

 

 

Likely one of the reasons why they specified a 750W min PSU in the 1st place...

 

 

 

 

ahem... if you peak at 750W for the gpu only, on let's say, half loaded 750W psu...

but nvidia did not say 750 is a mandatory minimal. they also say less can be fine depending of the whole setup. that's on their website.

[Ankerson]

33 minutes ago, whaka said:

ahem... if you peak at 750W for the gpu only, on let's say, half loaded 750W psu...

but nvidia did not say 750 is a mandatory minimal. they also say less can be fine depending of the whole setup. that's on their website.

 

Well if one has to read the fine print then they don't have enough PSU in the 1st place.

 

That said MOST of the AIB's are saying a min of 750W PSU.

 

 

[FinTheHuman]

Actually from what I read, MSI denied changing their design for the 3080 Gaming X Trio. The article here was update: https://www.tomshardware.com/news/msi-stealthily-revamps-geforce-rtx-3080-design-amid-stability-concerns

 

Here's the snippet: "Update 9/29/2020, 7:30am PT: MSI sent over images and a statement disputing any claims that it has changed the design on its 30-series GPUs. Included are four images of the backs of the cards, showing the capacitor configuration. The MSI RTX 3090 Gaming Trio and MSI RTX 3090 Ventus 3X have a layout with 4x SP-Caps and 2x MLCC clusters, while the MSI RTX 3080 Gaming Trio and MSI RTX 3080 Ventus 3X have a layout with 5x SP-Caps and 1x MLCC cluster. These are the cards that actually shipped to consumers and reviewers; earlier variants may have existed but apparently did not ship to anyone outside of MSI."

[FaxedForward]

8 hours ago, vong said:

So no word from Gigabyte? I'm queued for one of their Aorus master cards.

Gigabyte has a statement now:

https://www.kitguru.net/components/graphic-cards/dominic-moass/gigabyte-issues-statement-in-response-to-rtx-3080-3090-capacitor-controversy/
 

“In response to the recent reports speculating that the use of POSCAP capacitors on the GeForce RTX 3080/3090 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 3080/3090 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 29, 2020 that improves stability. Users are advised to update to the latest driver for optimized performance. For users who encounter power-related issues with GIGABYTE GeForce RTX 30 series graphics cards, GIGABYTE will provide product replacement, free of charge.

 

“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.”

[LukeSavenije]

edits according to statements provided by @FaxedForward and @FinTheHuman have been done

[Intoxicus]

On 9/30/2020 at 12:30 AM, LukeSavenije said:

edits according to statements provided by @FaxedForward and @FinTheHuman have been done

Perhaps something to add is to not use daisy chain cables for GPU power for 3000 series.

Seasonic and EVGA have said it's ok in a 3X8pin setup. But if you research into it their advice doesn't make sense.

TL;DR The 8 pin power connectors are rated for 150W and the daisy chain cables are often 18 gauge when they should be 16 gauge according to recommended specs. If we get a 500W+ peak it exceeds what the PSU can supply to a daisy chain cable on a 3x8pin GPU. The PSU can only send 150W through its single connection to the daisy chain asking for potentially up to 300W. This seems like an obvious recipe for magic smoke and/or dead components. If we really get into the details it seems like it's insane that these daisy chain power cables even exist in the first place.

I'm coming to suspect that people with randomly dying 3080s and 3090s are often using daisy chain power cables and that might be the cause of some of the dead GPUs being reported online. That is speculative at this time though.