AmpereGate! RTX 3080 instability

M

mls1995

Guest
Looks to me like only Nvidia FE and Asus cards are OK and that everything else is garbage but I just started looking into it
 

WorldExclusive

[H]F Junkie
Joined
Apr 26, 2009
Messages
11,014
Looks to me like only Nvidia FE and Asus cards are OK and that everything else is garbage but I just started looking into it

But the FE cards could be artificially being held back to prevent crashing. They still have 4 of the cheap parts (maybe one of the cost cuts to hit the $699 price point) while ASUS have none.
The cards doesn't have much OC headroom....and we now see why.

These cards have stout power demands, so who knows what will happen long term as the capacitors degrade.

This is the time to sit on the sidelines and wait to see how things pan out....like the BF4 launch.
 
M

mls1995

Guest
But the FE cards could be artificially being held back to prevent crashing. They still have 4 of the cheap parts while ASUS have none. These cards doesn't have much OC headroom....and we now see why.
These cards have stout power demands, so who knows what will happen long term as the capacitors degrade.

This is the time to sit on the sidelines and wait to see how things pan out....like the BF4 launch.

Well 4 of the cheap parts might be good enough, it's certainly better than what other companies are doing. I agree though that Asus seems to be the safest bet
 

Ready4Dis

2[H]4U
Joined
Nov 4, 2015
Messages
2,369
Seems like a scape goat for poor die binning. Nvidia gave everyone the build sheet, these items where said to be perfectly acceptable. To now blame the parts and not be pointing the finger at NVidia for telling the AIB's they were acceptable is literally just trying to paint the AIB's as the bad guy for using the cheaper parts that nvidia told them they could use. No, they followed the build with the parts that were suited for the job (or so they were told by nvidia in the reference design/BOM). No matter if this is the case or not; I am skeptical, it's just that one guy found this to be a single difference between some models so he suggested it'd be worth looking into, then you have click bait articles saying that THIS IS THE REASON ABSOLUTELY WITHOUT A DOUBT... lol, it was simply a guess, maybe a good one, maybe it has nothing to do with it. No matter what though, this was Nvidia's reference design and AIB's where told it would be sufficient. Just because you can lay 99.999% gold traces on a board because it's better, doesn't mean you should. You use the parts that do the job, If it's the low end model, you use the cheaper parts that can get the job done. Regardless of whether this is indeed the issue or not, the fault lies on NVidia for the design, and the AIB's for sh!t QA (although as I understand it, they didn't get a lot of time to build the boards since nvidia was rushing this launch something fierce, which leads to these kinds of issues not being found).
 

t1337duder

n00b
Joined
Sep 7, 2014
Messages
47
Looks to me like only Nvidia FE and Asus cards are OK and that everything else is garbage but I just started looking into it
Do we know for sure the Asus cards don't share the same cap setup which is causing the issue? Or just speculation?

From the EVGA post, are we to assume the FTW cards don't suffer the same issue?
 

st1650

n00b
Joined
Aug 10, 2020
Messages
40
Good thing that the FTW3 3080 in the stepup program is the “good” version. It pays to be patient I guess.
 
M

mls1995

Guest
Do we know for sure the Asus cards don't share the same cap setup which is causing the issue? Or just speculation?

From the EVGA post, are we to assume the FTW cards don't suffer the same issue?

Confirmed on Reddit and by Jayz2cents
 

5150Joker

Supreme [H]ardness
Joined
Aug 1, 2005
Messages
4,445
But the FE cards could be artificially being held back to prevent crashing. They still have 4 of the cheap parts (maybe one of the cost cuts to hit the $699 price point) while ASUS have none.
The cards doesn't have much OC headroom....and we now see why.

These cards have stout power demands, so who knows what will happen long term as the capacitors degrade.

This is the time to sit on the sidelines and wait to see how things pan out....like the BF4 launch.

FE cards aren't prone to it, they've been tested beyond 2 ghz and found they work fine.
 

exlink

Supreme [H]ardness
Joined
Dec 16, 2006
Messages
5,151
It appears that a few cards went with the 6x POSCAP arrangement initially when looking at product photos on Newegg and Amazon then changed their design for release cards. Even the ASUS TUF version has 6x POSCAP according to those store photos. ASUS probably made a last minute change to 6x MLCCs on the TUF just like EVGA changed from 6x POSCAP to 4x POSCAP/2x MLCC last minute on the FTW3.
 

Mylex

Gawd
Joined
Aug 30, 2018
Messages
707
The wife has been playing on her FTW3 the past 4 hours without issues. Build quality is pretty good to me and the design choices meaningless in her windowless case. It makes since if that was the reason for the delay. I guess I should put her card in my system to play at more enjoyable settings.
 

sethk

[H]ard|Gawd
Joined
May 3, 2005
Messages
1,849
Isn't evga saying no production cards shipped with the issue they "confirm"?
 

MangoSeed

[H]ard|Gawd
Joined
Oct 15, 2014
Messages
1,434
Nvidia gave everyone the build sheet, these items where said to be perfectly acceptable. To now blame the parts and not be pointing the finger at NVidia for telling the AIB's they were acceptable is literally just trying to paint the AIB's as the bad guy for using the cheaper parts that nvidia told them they could use. No, they followed the build with the parts that were suited for the job (or so they were told by nvidia in the reference design/BOM).

That is not accurate. Nvidia’s reference spec called for 1 MLCC group and 5 POSCAPs.
 

cybereality

Supreme [H]ardness
Joined
Mar 22, 2008
Messages
6,317
I'd still say Nvidia was to blame. The AIB partners should have received working drivers in time to QA, they would have caught these issues before release.
 

MangoSeed

[H]ard|Gawd
Joined
Oct 15, 2014
Messages
1,434
I'd still say Nvidia was to blame. The AIB partners should have received working drivers in time to QA, they would have caught these issues before release.

Yes Nvidia is to blame for shipping drivers late. However that would only help AIB’s to catch the defect not prevent it. Their decision to go below spec was unrelated to drivers.
 

Ready4Dis

2[H]4U
Joined
Nov 4, 2015
Messages
2,369
That is not accurate. Nvidia’s reference spec called for 1 MLCC group and 5 POSCAPs.
Do you have a different copy of it? I'd like to see what it actually called for. If that's the case, then EVGA was building cards not up to spec as well? "During our mass production QC testing we discovered a full 6 POSCAPs solution cannot pass the real world applications testing."

According to the posted images of the reference, this was in spec. https://www.igorslab.de/en/what-rea...tabilities-of-the-force-rtx-3080-andrtx-3090/
"The BoM and the drawing from June leave it open whether large-area POSCAPs (Conductive Polymer Tantalum Solid Capacitors) are used (marked in red), or rather the somewhat more expensive MLCCs (Multilayer Ceramic Chip Capacitor). The latter are smaller and have to be grouped for a higher capacity. " "According to the list and specifications of Nvidia, both are possible."

So, yes.. it is accurate, and the reference does not call for 1MLCC (unless there is/was a different spec that you know of). It leaves it open and both should work perfectly fine for low end boards. For higher end boards and more OC, yes it wouldn't make sense to cheap out.


Yes Nvidia is to blame for shipping drivers late. However that would only help AIB’s to catch the defect not prevent it. Their decision to go below spec was unrelated to drivers.
Again, it wasn't below spec... they were within spec, see above. EVGA knew of the problem prior to launch (hence their recall) yet Nvidia allowed all the other manufacturers to ship. NVidia absolutely deserves the majority of the blame here. The more information that comes out, the more it confirms (Especially the part about EVGA knowing about the issues prior to launch, I'm 100% positive they were in contact with NVidia about this issues to resolve).
 

tunatime

Well...OK
Joined
Sep 15, 2011
Messages
4,066
Now we know why there is no stock.....bet all the next batch of cards are going to be fixed and
that aib/nvidia found the problem but still sold them hoping no one would notice till they fixed them.
 

AVATARAT

Weaksauce
Joined
Jun 16, 2020
Messages
107
Can someone owner on 3080/3090 that have crashes to try to lower 50-100MHz the VRAM as fix for the problem ?
 

Ready4Dis

2[H]4U
Joined
Nov 4, 2015
Messages
2,369
I’m going off the same pic in the Igor article showing 1 MLCC array. If the reference spec in fact allows for 6 cheap caps then this is on nvidia.
Yes, that image of the single most pack was a test they did to see if it made a difference. That was a hand soldered in place replacement of the cheaper caps with the better caps and actually ran better, but it wasn't required by nvidias reference.
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,181
Me and JaysNoCents got in to on Twitter. Rofl. He dropped f bombs hahaha. “He basically said it’s AIBs fault but you are still getting the performance promised and may not exactly get exact extra boost but still what you were promised if you don’t like it return it so someone else can appreciate the card” I basically said he tries so hard to not piss of Jensen’s leather jacket. Like nvidia has no blame here for not keeping AIBs in check when testing their boards and it’s all AIBs fault. And making it sound like those having issues some how don’t appreciate it? So others can appreciate open box? Wtf lol.

He bsaically responding doing this is putting him on nvidias shit list and I can fuck off. Yea saying that nvidia is safe form class action and users are getting what they promised and they can return it and be happy is really getting on nvidia’s shit list. 🙄.

I am sure I will be blocked soon from his Twitter.
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,181
Looks to me like only Nvidia FE and Asus cards are OK and that everything else is garbage but I just started looking into it


No hardware unboxed has issues with asus tuf OC variant. They said they are not sure about the non OC because they don’t have it to test. I saw their Twitter.
 

tunatime

Well...OK
Joined
Sep 15, 2011
Messages
4,066
No hardware unboxed has issues with asus tuf OC variant. They said they are not sure about the non OC because they don’t have it to test. I saw their Twitter.
So the caps might be a smoke screen to bad cores? Be a little shady if they patch the bost to be 10% lower or something
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,181
I am getting popular. My tweet to jay is getting lot of hate by fans. I guess even bad marketing is good marketing even if people are yelling at me. One dude said I am stupid because not all people use boost I was like wtf? It’s a feature you don’t enable that shit boost is automatic excluding if you overclock further. He made it seem like people can turn off boost to lower performance below stock that shit boosts at stock 😂😂😂😂
 

exlink

Supreme [H]ardness
Joined
Dec 16, 2006
Messages
5,151
No hardware unboxed has issues with asus tuf OC variant. They said they are not sure about the non OC because they don’t have it to test. I saw their Twitter.
The issue is apparently occurring on all versions of the card both AIB and FE, but doesn’t happen to all cards. People say that most of the reports seem to be coming from Zotac cards but I also believe Zotac cards could be the most shipped to date.

So essentially we’re still too early in the problem investigation to draw any conclusions just yet and people should sit tight. The cap arrangement could only be one part of the problem or could be a completely separate issue.
 

III_Slyflyer_III

Limp Gawd
Joined
Sep 17, 2019
Messages
153
Kind of glad I could not snag a 3090... confirmed I'm gonna play the sit and wait game now!

Its like all those 2080Ti's that had issues... I waited awhile and got one with Samsung memory and never had a single issue over 1.5 years. Does pay to wait a few months for sure sometimes, even tho I want a 3090 bad.
 
M

mls1995

Guest
No hardware unboxed has issues with asus tuf OC variant. They said they are not sure about the non OC because they don’t have it to test. I saw their Twitter.

OK. I trust Hardware Unboxed, I consider them to be an outstanding source of information. One of my favorites actually
 

Ready4Dis

2[H]4U
Joined
Nov 4, 2015
Messages
2,369
On going list of the different poscap/mlcc designs from all the manufacturers here if you feel more confident in which power delivery/filtering selection you want to make:
https://www.reddit.com/r/hardware/comments/izmi1k/ampere_poscapmlcc_counts/
Yes, but it's not completely constrained to only cards with the POSCAP's... it's also been seen on models with MLCC designs as well, it just seems more common on the POSCAPs, but not limited to them. Of course, it could be multiple issues (aka, POSCAP issue, and driver issue), but at this point we don't know enough to even say you're safe if you buy an FE or MLCC board (just that you'll likely have less chance of issues).
 

NKD

Supreme [H]ardness
Joined
Aug 26, 2007
Messages
8,181
Yes, but it's not completely constrained to only cards with the POSCAP's... it's also been seen on models with MLCC designs as well, it just seems more common on the POSCAPs, but not limited to them. Of course, it could be multiple issues (aka, POSCAP issue, and driver issue), but at this point we don't know enough to even say you're safe if you buy an FE or MLCC board (just that you'll likely have less chance of issues).

Hopefully it’s not Samsung’s shit process where the some chips just can’t hold boost that high and that’s the issue where nvidia has to now drop performance in favor or stability. Good leaker on Twitter said it might be that some chips are not as good of quality as others
 

Ready4Dis

2[H]4U
Joined
Nov 4, 2015
Messages
2,369
Hopefully it’s not Samsung’s shit process where the some chips just can’t hold boost that high and that’s the issue where nvidia has to now drop performance in favor or stability. Good leaker on Twitter said it might be that some chips are not as good of quality as others
Yeah, I wouldn't doubt Nvidia are giving out low binned chips, then coupled with low end circuitry, ends up being prone to crashing. Still waiting on an official statement?
 

exlink

Supreme [H]ardness
Joined
Dec 16, 2006
Messages
5,151
Another update from Igor’s Lab:

“Owners of cards, which still run really stable despite six solids, owe everything to the very good chip quality. Owners of cards with MLCC, where errors still occur, may be annoyed about a GPU that does not even consistently cope with the stored voltage/frequency curve. This is exactly the point where the board partners couldn’t test anything at all with the first cards due to the lack of suitable drivers. There are certainly many cards in circulation here that would not have been suitable as OC cards.

The fact that NVIDIA has split the GPU’s power supply voltages between NVVDD and MSVDD also shows that they are well aware of the problem. I noticed that the MSVDD has much less changes and is generated independently from the NVVDD, so you should be able to get along with a well equipped MLCC group. So more MLCC does no harm if the rest of the layout allows this interpretation. Without it, however, becomes slow and sluggish.”

https://www.igorslab.de/en/nvidia-g...-so-important-and-what-are-the-object-behind/
 

Mylex

Gawd
Joined
Aug 30, 2018
Messages
707
Another update from Igor’s Lab:

“Owners of cards, which still run really stable despite six solids, owe everything to the very good chip quality. Owners of cards with MLCC, where errors still occur, may be annoyed about a GPU that does not even consistently cope with the stored voltage/frequency curve. This is exactly the point where the board partners couldn’t test anything at all with the first cards due to the lack of suitable drivers. There are certainly many cards in circulation here that would not have been suitable as OC cards.

The fact that NVIDIA has split the GPU’s power supply voltages between NVVDD and MSVDD also shows that they are well aware of the problem. I noticed that the MSVDD has much less changes and is generated independently from the NVVDD, so you should be able to get along with a well equipped MLCC group. So more MLCC does no harm if the rest of the layout allows this interpretation. Without it, however, becomes slow and sluggish.”

https://www.igorslab.de/en/nvidia-g...-so-important-and-what-are-the-object-behind/
I think this is a simple and reasonable hypothesis, probably as close to the truth as we will get.
 
Top