Rtx 3080 lower quality capacitor Issue

[willhub]

If it's the more expensive ones that are the issue with is the TUF ok with all more expensive?

 

[ALXAndy]

Thats the one I have, no crashes unless I OC over 55Mhz, but it will run at 2050 - 2070Mhz with a 55Mhz OC

Then it's fine.

 

[adrichardson81]

i think evga worded it wrong and threw every other company under the bus that’s the bit nvidia won’t like.
I don’t think it will be an bios update they chose I think nvidia if the power issues and boost clock is the issue will change it at a driver level and all cards won’t boost as high all for one and one for all will be nvidia moto they need there partners they won’t call them out

Plus side, if you're on a decent card you'll gain some oc potential back I suppose...

 

[ihatelag]

MSI said in their livestream that they are aware of the 3080 issue and said it can be fixed with a driver update, so I guess they are going down the "free" approach of fixing their cards.

 

[ALXAndy]

MSI said in their livestream that they are aware of the 3080 issue and said it can be fixed with a driver update, so I guess they are going down the "free" approach of fixing their cards.

To be fair even though he was totally wrong as usual Jay did raise a good point. IE, your card is supposed to boost to the frequency on the box and if it does so then you have little reason to complain.

I've bought some real stinkers in the past. Had a 670 that wouldn't even do an extra 50mhz but I never sent it back.

 

[69Dude]

Let's look at some absolute facts.

The SP Caps and MLCC are for power filtering and stabilisation.
EVGA confirmed the 6 SP Caps cause real world failures in testing.
We can assume that this is true since they delayed a product off the back of the research.
EVGA now have 4 + 2 and a 5 + 1 3080 models.

The Founders has 4 + 2 and 3090's mostly have 4 + 2.
This would show confidence that 4 +2 is a sufficient or optimal configuration

Many AIB's initially had pictures on ther own product pages of 6 SP Caps (like the MSI Ventus) but then the real world products have 5 + 1 and those pictures have been replaced.
This would show that non EVGA AIB's have knowledge of the issue and that 5 + 1 is sufficient in their follow up test criteria.

The issue is now public and everyone is looking to confirm and verify cases, many via overclocking.

Here's an example of a guy where the stock boost config is fine but he can show the issue, only when overclocking the boost even further:
https://www.youtube.com/watch?v=aeWHJzLdUl8

People are now going out and looking to generate the issue.
Reports from end users can't be lab verified to show that a 5 + 1 is actually a problem.

It might even be possible that the SP Caps and MLCC in combination soak signal noise better than 6 of 1 kind.

The below article might be enlightening about how people act when trying to reverse engineer cause and effect without enough knowledge.

https://en.wikipedia.org/wiki/B._F._Skinner#'Superstition_in_the_pigeon'_experiment

 

[SniperSmithy]

These crashes that are happening almost all speak about clocks of around 2000mhz. Given that the stock boost clocks on at least the Gigabyte Eagle are meant to be 1755mhz they are either being overclocked past their stable range which is user fault or the boost algorithm is taking the clocks too high, which sounds like a driver error to me. I think it's the latter.

 

[SniperSmithy]

Let's look at some absolute facts.

The SP Caps and MLCC are for power filtering and stabilisation.
EVGA confirmed the 6 SP Caps cause real world failures in testing.
We can assume that this is true since they delayed a product off the back of the research.
EVGA now have 4 + 2 and a 5 + 1 3080 models.

The Founders has 4 + 2 and 3090's mostly have 4 + 2.
This would show confidence that 4 +2 is a sufficient or optimal configuration

Many AIB's initially had pictures on ther own product pages of 6 SP Caps (like the MSI Ventus) but then the real world products have 5 + 1 and those pictures have been replaced.
This would show that non EVGA AIB's have knowledge of the issue and that 5 + 1 is sufficient in their follow up test criteria.

The issue is now public and everyone is looking to confirm and verify cases, many via overclocking.

Here's an example of a guy where stock boost config is fine but he can show the issue, only when overclocking the boost even further:
https://www.youtube.com/watch?v=aeWHJzLdUl8

People are now going out an looking to generate the issues.
Reports from end users can't be lab verified to show that a 5 + 1 is actually a problem.

It might even be possible that the SP Caps and MLCC in combination soak signal noise better than 6 of 1 kind.

The below article might be enlightening about how people act when trying to reverse engineer cause and effect without enough knowledge.

https://en.wikipedia.org/wiki/B._F._Skinner#'Superstition_in_the_pigeon'_experiment

Not all SP caps are the same/equal. It's like saying all cars are slow because cheap cars have smaller engines.

 

[TrM]

These crashes that are happening almost all speak about clocks of around 2000mhz. Given that the stock boost clocks on at least the Gigabyte Eagle are meant to be 1755mhz they are either being overclocked past their stable range which is user fault or the boost algorithm is taking the clocks too high, which sounds like a driver error to me. I think it's the latter.

the boost algorithm also regulates voltage and power draw. might be its pulling to much juice also and not the clockspeed is the issue it is a whole new gpu architecture and bound to have issues.

 

[SeeNoWeevil]

EVGA confirmed the 6 SP Caps cause real world failures in testing.
We can assume that this is true since they delayed a product off the back of the research.

You can't extrapolate from this that 6 SP caps is a problem on all boards at all the varying boost clocks they use. What clocks were EVGA talking about when they stated this? They don't say. I'm sure it does limit OC ability somewhat but that's irrelevant on a stock card. There's no entitlement to these higher clocks on an RRP card.

 

[DFroN]

According to Hardware unboxed some TUF and FE cards have the same issue. If true it means that every card with every combination of capacitors from 0 MLCC to 6 MLCC has the same issue, so its probably not the caps: https://twitter.com/HardwareUnboxed/status/1309659834468298753?s=20

 

[Aretak]

These crashes that are happening almost all speak about clocks of around 2000mhz. Given that the stock boost clocks on at least the Gigabyte Eagle are meant to be 1755mhz they are either being overclocked past their stable range which is user fault or the boost algorithm is taking the clocks too high, which sounds like a driver error to me. I think it's the latter.

It's not an error. That's how Nvidia's boost algorithm has always worked. It'll boost as high as it can until it runs into some sort of limiting factor (usually the power or voltage ceiling). I've never owned an Nvidia card that only delivered the "guaranteed" boost clock written in the specs. They all boost way higher right out of the box by design. Maybe Nvidia have made it even more aggressive this time around and that's causing issues, but that remains to be proven of course.

 

[TearSoup]

Palit 3080 GamingPro has one MLCC. Have not had any crashing either.

What about when you put the power limit to its maximum position and then play a gpu intensive game with it?

 

[Gamer_3983]

These crashes that are happening almost all speak about clocks of around 2000mhz. Given that the stock boost clocks on at least the Gigabyte Eagle are meant to be 1755mhz they are either being overclocked past their stable range which is user fault or the boost algorithm is taking the clocks too high, which sounds like a driver error to me. I think it's the latter.

Isn't the boost algorithm meant to boost the clock speed as high as it can until it hits power limit.

 

[Vidar]

Gigabyte RTX 3080 Gaming OC 10GB

This card has genuinely surprised me! With a +125 on the core and boosting to a rock solid 2070Mhz after hours of gaming and benchmarks I've finally started tackling the memory. Each +100 actually seemed to bring my Port Royal score down but I remembered something Gamers Nexus said about sometimes a really high number will work when lower ones will not. So I just dropped a +800 randomly and it not only increased my score to a whopping 12075 but again is rock solid in games.

Annoyingly 3D Mark isn't letting me validate, saying I've got no internet connection or some such but here you go...

I'm far from being an expert by any stretch of the imagination but having a card that's supposed to have the "cheap caps" problem and getting results like this leads me to believe this has to do with the Silicon Lottery, I did see something pre-launch that said Nvidia and 3 different bins of chips, I'm wondering if this is where the real issue is. If you've got a rock solid chip then the caps shouldn't be an issue when boosting to higher than specified clocks, yet a lower quality chip and you are likely to have issues.

Think I will be keeping a hold of my card, though it would be fascinating to see what other people get. The fact is with a 4 year warranty from Gigabyte, I'm not worried at this point.

 

[SkinSpotlights]

I've finally started tackling the memory. Each +100 actually seemed to bring my Port Royal score down but I remembered something Gamers Nexus said about sometimes a really high number will work when lower ones will not.

GDDR6X is error correcting memory, so if it gets unstable then more errors occur and then it will start trying to fix them which then causes dips. Whats really interesting is though how a higher number works but lower one doesn't, guess it offsets the errors by having more speed?

 

[KompuKare]

I'm surprised that so many people don't know how Nvidia's boost works and had worked for years.
Yes, drivers could probably scale the max boost back but surely consumers won't be happy if their cards are noticeably slower than the reviews were.
It would be as if reviewers were seeded golden samples which most consumers would never see. It would certainly be a bit too close to bait-and-switch for me.

 

[SeeNoWeevil]

I'm surprised that so many people don't know how Nvidia's boost works and had worked for years.
Yes, drivers could probably scale the max boost back but surely consumers won't be happy if their cards are noticeably slower than the reviews were.
It would be as if reviewers were seeded golden samples which most consumers would never see. It would certainly be a bit too close to bait-and-switch for me.

I'm just speculating here, but potentially the issue could be resolved by eliminating the brief spikes to high clocks under light workloads which wouldn't affect overall performance. Although factory OC cards may still need more substantial downclocking to remain stable.

 

[TearSoup]

I got my Gigabyte 3080 gaming yesterday and havnt had any crashes yet while playing with it. Just heard about this issue and seems this card has the worst setup of capacitors so hoping it doesnt cause issues. Will keep testing over the next few days to see if anything happens but so far so good.

Its pretty quiet considering the last windforce cooler i had was super loud, tops out around 67 degrees at 75% fan in my setup. Will push the OC a bit and see if it induces any crashes like mentioned.

It's not something that I'd entirely recommend but you could really try to make sure and attempt a simple overclock on your card to make sure that the issue isn't being masked due to you being a mere inch away from the problem clocks, you know? Nvidia boost ramps up clock speeds automatically to as speeds above 2,000 mhz. You overclocking (while also having fans running at max speed) can help completely rule out capacitor power delivery issues to your GPU at high frequencies. This would give you some peace of mind because if there really was an issue with this, you can rest assured that you have thoroughly investigated it. This leads to this issue not cropping up in the distant future when your return window is finished when perhaps your room was a little cooler and you pc begins crashing randomly.

 

[TearSoup]

When you get your card, don't take it out of the sealed bag.
Take a look at the back first....That's what I will be doing.

And doing what? If it's 5 PODCAPS and 1 MSCC would you end up returning it or keeping it?