Did All The AIB's Know About the SP Cap Problem before the 3080/3090 Launches?

[Briggsy]

tinfoil hat time!

 

I've watched Buildzoid's and Jayztwocent's videos about the MLCC and SP Caps issue on the 3080 and 3090 cards, originally discovered by Igor over at IgorsLab. Review models like the FTW from EVGA, or the ROG Strix and TUF Gaming from Asus are now using MLCC caps in full or at least partially.

 

Oddly enough, the backside pictures of several cards on Newegg shows them all with 6 SP Caps. (see images below) 

 

Did AIB's know something before launch, and have to ship launch cards back to the factory to be upgraded? I'm just speculating here, but it seems like there's more to this story than we currently know, as many have been blaming AIB's for cutting corners on Nvidia's reference spec, myself included until now.

 

I mean, it could at least explain why supply is so tight for every single model of 3080 and 3090 right now.

 

What this does suggest to me is that AIB's did not cut corners, but that Nvidia's reference spec likely was 6 SP Caps.

 

Below are pictures from Newegg, which all show 6 SP Caps and zero MLCC's.

 

TUF Gaming 3080 

Spoiler

 

ROG Strix 3090

Spoiler

 

MSI Ventus 3090

Spoiler

 

EVGA FTW 3090

Spoiler

Spoiler

 

 

Edited September 28, 2020 by Briggsy
clarity and citing sources

[bellabichon]

No tinfoil hat necessary, that's exactly what happened. From a forum post by Jacob Freeman, product manager at EVGA:

Quote

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.

It would seem that this issue, combined with Nvidia giving their board partners less than average R&D time to work on their cards, greatly contributed to the lack of supply at launch.

[PeachGr]

Full respect to the people that do the research and found the problem, but in electronics, that is an over simplification. Don't think that you know the whole story after a short text. 

[Briggsy]

1 minute ago, bellabichon said:

No tinfoil hat necessary, that's exactly what happened. From a forum post by Jacob Freeman, product manager at EVGA:

 

good to know, but it appears it wasn't just EVGA. 

[Airdragonz]

Until Nvidia themselves come out with an official statement, no one (including the AIBs) are 100% sure what the problem is exactly. The poscap / spcaps could be the or an issue, it could be a driver issue, it could be a completely unrelated issue, or even a combination of problems. 

Remember, Igorslab was the one who released the initial findings about the caps, and as far as I'm aware, the only site which discovered this. Most other sites and tech reviewers just reiterated the report. 

[bellabichon]

Just now, Briggsy said:

good to know, but it appears it wasn't just EVGA. 

True, but I would assume that all the other manufacturers encountered the same problem during the testing of their factory overclocked cards, and used the same or similar solutions. I'd guess that the reason that we're hearing from EVGA specifically on this matter is they're typically one of the more transparent manufacturers when it comes to discussing product issues.

[Briggsy]

22 minutes ago, Airdragonz said:

Until Nvidia themselves come out with an official statement, no one (including the AIBs) are 100% sure what the problem is exactly. The poscap / spcaps could be the or an issue, it could be a driver issue, it could be a completely unrelated issue, or even a combination of problems. 

Remember, Igorslab was the one who released the initial findings about the caps, and as far as I'm aware, the only site which discovered this. Most other sites and tech reviewers just reiterated the report. 

yep I'm aware of Igor being the original finder and I've updated the OP to include this information. appreciate the reminder.

 

I guess one of the big revelations for me was that Buildzoid in his video discussed Nvidia's reference spec likely being to blame, but what that spec was exactly was unknown, and some people are blaming AIB's for cutting corners. 

 

Based on the pictures from newegg it seems like none of them cut corners at all.

[Spotty]

21 minutes ago, Briggsy said:

Oddly enough, the backside pictures of these cards on Newegg shows them all with 6 SP Caps. 

The images they use for marketing and promotional material would have been early samples. They need to have those images ready to go on their website when the product launches and they need to distribute that material out to vendors so they can create their product pages - BEFORE they get the cards in stock.

It's likely that original samples of those cards shown in the early pictures used the SP-Caps and then perhaps once the AIB partners got drivers from Nvidia the cards were failing their internal validation and they changed the design to use the MLCCs.

 

I wonder if the AIB partners had already started shipping cards out before realising they weren't stable. Apparently reviewers got drivers for the cards before some of the AIB partners did.

 

21 minutes ago, Briggsy said:

Did AIB's know something before launch, and have had to ship a lot of cards back to the factory to be upgraded?

I wonder what they will be doing with the returned cards. The cards are all run through pick and place machines which place the capacitors during the assembly line. I doubt they're going to take returned cards and desolder the SP-caps and solder on MLCCs manually.

[-CLASSIFIED-]

asus shows the reg tuf with the more expensive 6x full arrays. and the OC tuf with the 6 reg caps. pretty sure they updated both pcbs to have all arrays in the spots. 

[porina]

2 hours ago, Spotty said:

I wonder what they will be doing with the returned cards. The cards are all run through pick and place machines which place the capacitors during the assembly line. I doubt they're going to take returned cards and desolder the SP-caps and solder on MLCCs manually.

In business the question is going to eventually boil down to costs.

 

It isn't that difficult to replace the components by hand, but you're still going to have to have someone do it, and then test the cards afterwards to ensure they are working correctly. How much would they gain from doing that? At best you'd have B-grade units afterwards. Is it less loss to throw them directly in recycling? There will be new stock coming in to replace them in due course. 

 

Or maybe get really creative. Create a new bios optimised for low power consumption, even if it means lower clocks, and rebrand it as an eco version. Note I've not kept up to the exact claims of when the problems happen, but for this purpose I'm assuming it is at higher clocks/power usage. Backing off on either or both would help.

[Mark Kaine]

I don't think resoldering different capacitors would be all that costly tbh. (I do have some experience with this)

[LogicalDrm]

-> Moved to Graphics Cards