AmpereGate! RTX 3080 instability

Ummm yeah but so far all the reports are that performance has not changed. Not sure what we're talking about.
Ok, never mind. It was based on your comment "You pay for a card that boosts up to 1710Mhz". I just took it that it would be OK for the test cards to boost 2000mhz but since a buyer looking at those reviews should maybe be happy if it only boost to 1710mhz since that is what he payed for.
 
From what I've been reading, the drivers lowered max boost speeds on some cards (not all apparently) and increased power consumption slightly. However, performance has stayed relatively the same or increased in some instances. People are hypothesizing that the cards that are not boosting as high are sustaining their higher boost clocks for longer which keeps overall performance the same or even better. And lastly, there is conversation that its not necessarily the high boost clocks that causes the cards to crash, but rather the very quick ramp up to max boost clocks that would cause it. As a result, certain capacitor arrangements couldn't keep up with this ramp up.

But who knows, we need actual professional, repeatable testing performed with variables under control.
Which historically made first reviews of video cards more of a first run and expect changes, sometimes immediately that can change the numbers, just the changes are not routinely noted on these reviews, updated note and link to more recent results would be nice if reasonably possible for a site to do.
 
Ok, never mind. It was based on your comment "You pay for a card that boosts up to 1710Mhz". I just took it that it would be OK for the test cards to boost 2000mhz but since a buyer looking at those reviews should maybe be happy if it only boost to 1710mhz since that is what he payed for.

That’s not what I meant. I was trying to understand what you meant :)

You paid for the specs on the box and the performance in reviews. So far it seems neither of those things have changed.
 
  • Like
Reactions: noko
like this
Looks like first tests with new drivers did not adversely affect performance at all, very good news. Also seemed to have solved the issue of CTD. Day one reviews should be good for performance capability.


Okay, that seemed like the final word to me.
 
Looks like first tests with new drivers did not adversely affect performance at all, very good news. Also seemed to have solved the issue of CTD. Day one reviews should be good for performance capability.



Good summary of the entire situation even though a little heavy on the self praise. Lol at the major shade he threw on JayzTwoCents.
 
Yep, still disagree with you, but it's ok. So if a card was "OCd" to 1500mhz base clock, and boosted to the factory 1710mhz... And another card kept base clocks the same but boosted to 2000mhz... Your consider the first one overclocked and the second one factory clocked? Man these guys must be marketing wrong, they should just dump boost clocks back down to 1710 and just give a +50mhz base clock so they can sell all of their base models as overclocked models!! Ok, I get it you want to call it boost clocks except that already has a term and a number, which is officially 1710mhz. So we have a boost frequency (clock) that is > advertised boost frequency (clock)... Let's just call this greater than. So we have clocks greater than the advertised boost frequency (I think we are still in agreement to this point?). So we have this thing that we are fine calling a higher (or greater) frequency... But if we replace higher with over... And clock for frequency (which are synonymous), it's now deemed the wrong thing? In my mind, whether you agree or not, if something is running at a frequency higher than spec, then it's overclocked. Whether it is the base or boost clocks, its just a base OC vs a boost OC. If that doesn't make sense to you, that's ok, feel free to not use it and continue on in life, it really doesnt change anything for either one of us.

Oh, sorry never saw this reply.

Hey, I am not arguing with you, I am not really even trying to change your mind. You can believe what you want, you are certainly doing some amazing mental gymnastics to make it seem like you have some sort of valid argument.

Nope, all I am doing is pointing out the facts and that the way you are thinking is wrong.

And the simple fact is that the Boost clock boosting past spec is not a factory overclock. No matter how try to twist it to seem otherwise.

But, You are right about one thing, your opinion on this does not matter, at the end of the day facts are facts.

I am surprised that somebody like you in the technical community is having difficulty with this?
 
But who knows, we need actual professional, repeatable testing performed with variables under control.

I feel like we don't really need anything at this point... the problem has been resolved and performance hasn't suffered.

Mystery hunters can certainly keep investigating for curiosity sake, but it will probably end up like Kyle's space invaders hunt. Interesting in the end, but ultimately moot since the problem was resolved in the meantime.
 
Good summary of the entire situation even though a little heavy on the self praise. Lol at the major shade he threw on JayzTwoCents.

It’s not like there wasn’t a problem, der8auer’s test shows that replacing an sp-cap with mlcc did in fact increase OC and stability, just not by 100 MHz BUT he did stop only after adding one array so it’s possible he could’ve gotten even higher OC by adding additional arrays. So while that may not cover the entire spectrum of the problem, it was certainly a factor and shouldn’t be dismissed. ASUS and EVGA wouldn’t have gone back to change their designs if it didn’t make a difference.

Jayz2Braincells might have created a clickbait title about all this but his video just regurgitated Igors findings and a bit of his own ramblings. I still don’t understand the smug self back patting of Steve in that HWUB video, yeah it’s obviously a multifaceted engineering problem way beyond the scope of most of us but you can’t dismiss empirical evidence either and say “oh look Jayz is a fucking idiot and we’re so smart because we chose to keep quiet”. The driver fix did reduce boost, even if HWUB didn’t see a performance drop, others (eg real end users) did see one from 0.5-3+% in some games. While it wouldn’t warrant changing day 1 numbers, it will for sure reduce boost ceilings for some cards.
 
Last edited:
It’s not like there wasn’t a problem, der8auer’s test shows that replacing an sp-cap with mlcc did in fact increase OC and stability, just not by 100 MHz BUT he did stop only after adding one array so it’s possible he could’ve gotten even higher OC by adding additional arrays. So while that may not cover the entire spectrum of the problem, it was certainly a factor and shouldn’t be dismissed. ASUS and EVGA wouldn’t have gone back to change their designs if it didn’t make a difference.

Jayz2Braincells might have created a clickbait title about all this but his video just regurgitated Igors findings and a bit of his own ramblings. I still don’t understand the smug self back patting of Steve in that HWUB video, yeah it’s obviously a multifaceted engineering problem way beyond the scope of most of us but you can’t dismiss empirical evidence either and say “oh look Jayz is a fucking idiot and we’re so smart because we chose to keep quiet”. The driver fix did reduce boost, even if HWUB didn’t see a performance drop, others (eg real end users) did see one from 0.5-3+% in some games. While it wouldn’t warrant changing day 1 numbers, it will for sure reduce boost ceilings for some cards.

While this debacle isn’t fantastic it’s honestly a non-issue for me. I just ordered a 3080.

The boosting to 1900 and a bug causing it to go to 2055 makes sense and not a big deal to fix.

Reminds me of my 1080 a little. I frequency/voltage locked since that would let me OC 50Mhz higher (no random boosts at low temp / without enough voltage). To me it appeared voltage lagged frequency increase.

More caps help but it’s in the tiny % range. You shouldn’t be crashing due to caps... they have a pretty large variance both production and temps. A card should be stable even with 1/2 of the caps if you are mass producing them and want a high success rate.

Sounds like it’s fixable via drivers and not firmware which is great.
 
It’s not like there wasn’t a problem, der8auer’s test shows that replacing an sp-cap with mlcc did in fact increase OC and stability, just not by 100 MHz BUT he did stop only after adding one array so it’s possible he could’ve gotten even higher OC by adding additional arrays. So while that may not cover the entire spectrum of the problem, it was certainly a factor and shouldn’t be dismissed. ASUS and EVGA wouldn’t have gone back to change their designs if it didn’t make a difference.

There's also evidence of cards with the "good" caps suffering the same fate. I think it's fair to say that the hardware was adequate for the intended boost range and misbehaving software was the issue.

Jayz2Braincells might have created a clickbait title about all this but his video just regurgitated Igors findings and a bit of his own ramblings. I still don’t understand the smug self back patting of Steve in that HWUB video, yeah it’s obviously a multifaceted engineering problem way beyond the scope of most of us but you can’t dismiss empirical evidence either and say “oh look Jayz is a fucking idiot and we’re so smart because we chose to keep quiet”. The driver fix did reduce boost, even if HWUB didn’t see a performance drop, others (eg real end users) did see one from 0.5-3+% in some games. While it wouldn’t warrant changing day 1 numbers, it will for sure reduce boost ceilings for some cards.

Yeah HWUB were far too proud of themselves. They made the right call in the end but it was far from obvious to everyone what was going on at first. Jay2TwoCents is getting grief for his delivery style. He took Igor's speculation and ran with it like it was gospel truth. I think the drama would've happened either way though. Igor's article was getting plenty attention too.
 
Yep, still disagree with you, but it's ok. So if a card was "OCd" to 1500mhz base clock, and boosted to the factory 1710mhz... And another card kept base clocks the same but boosted to 2000mhz... Your consider the first one overclocked and the second one factory clocked?
Yes, because the first one is guaranteed to be clocked at minimum 1500mhz from the factory, and be stable there. The second one only boosts higher because the environment it is in and the components it has allow it to do so, not because it was configured to do so. In a poor environment, the second card would struggle to reach 1700mhz, but both cards will reach their advertised base clock.

Iirc, it's been this way since the 1080 or earlier, and people made a big stink about it back then, too. Don't know why people are still confused about it now...
 
Much ado about nothing. Early days and driver bugs. There doesn’t seem to be much more performance to be gained from Ampere OC’ing. Is it nice to be at the top end? Yes. Absolutely. But the gains seem to be minimal in real world FPS for the effort and cost :/

Nvidia seems to have squeezed as much as they can and pushed it to the ragged edge. Even the mammoth ASUS strix version can’t produce significant gains despite going over 2Ghz regularly. So I suspect AIB’s will get wacky to differentiate amongst themselves.

The FE version, for me anyway, is the most appealing. Not the coolest or highest clocking I grant you, but it looks the bees knees for not much lost in terms of performance.

Tight variances this generation
 
What if Nvidia purposely made shitty drivers so that they could release a new driver that would prevent you from overclocking the 3080 to get more speed out of it, to justify the $1500 dollar pricetag of the 3090?
 
What if Nvidia purposely made shitty drivers so that they could release a new driver that would prevent you from overclocking the 3080 to get more speed out of it, to justify the $1500 dollar pricetag of the 3090?
GTFO here with you conspiracy theories. It been proven that the drive fix didn't change much performance wise. If anything some people are getting slightly better performance.
 
GTFO here with you conspiracy theories. It been proven that the drive fix didn't change much performance wise. If anything some people are getting slightly better performance.

But can you overclock the 3080 without it crashing? Is what I'm getting at.
 
But can you overclock the 3080 without it crashing? Is what I'm getting at.
These cards are pretty much clocked as high as most can go without actually binning the chips. Regardless of which driver they were on. The drivers barely changed the boost limit, if at all, on many cards from what has been reported thus far.
 
It is not the same thing and not a bandaid fix. Shot was rushed out and boasted higher then it should have been. You are still going to get higher then advertised boost clocks. Sure if the driver update cause the GPUs too boost under advertised clocks then it would of been a issue. It is not like you can't OC it yourself still.
Wrong this is a band-aid fix. The cheap capacitors are creating a resonating circuit which is causing part of the issue. The driver modification was designed to band-aid the poor design.
 
Wrong this is a band-aid fix. The cheap capacitors are creating a resonating circuit which is causing part of the issue. The driver modification was designed to band-aid the poor design.

Do you have any proof of this? der8auer’s data showing it stabilizes at 1900Mhz and random shoots to 2050+ and it being a bug makes a lot more sense.
 
Do you have any proof of this? der8auer’s data showing it stabilizes at 1900Mhz and random shoots to 2050+ and it being a bug makes a lot more sense.

Right, all hardware will fail at some threshold clock. It would make sense to blame the caps if clocks were dialed back significantly to solve the issue but that isn't the case. The observed clock range appears to be the same, minus the sudden spikes.

der8auer's video actually provided some of the best evidence that the failures were due to large sudden jumps in clock speeds that the caps couldn't handle. These spikes were a bug not a feature.
 
Do you have any proof of this? der8auer’s data showing it stabilizes at 1900Mhz and random shoots to 2050+ and it being a bug makes a lot more sense.
This all depends on which cards are being tested and which capacitor setup. Everyone is basing there opinion from a single person posting on a website just google around a lot of electrical engineers chimed in on the resonance issue. But what ever not sure why people are losing there shit for difference of opinion.
 
This all depends on which cards are being tested and which capacitor setup. Everyone is basing there opinion from a single person posting on a website just google around a lot of electrical engineers chimed in on the resonance issue. But what ever not sure why people are losing there shit for difference of opinion.

I am an EE. I just asked if you had seen data of this someplace since it’s something you could oscilliscope. Not typically something a person would guess at.

Der8raur is one of the best overclockers on the planet, has data, and knows the card’s intimately so there’s a good chance it’s solid and he’s right about the software bug.

I have one of these on the way and a oscilliation issue is a way different issue lol.
 
I am an EE. I just asked if you had seen data of this someplace since it’s something you could oscilliscope. Not typically something a person would guess at.

Der8raur is one of the best overclockers on the planet, has data, and knows the card’s intimately so there’s a good chance it’s solid and he’s right about the software bug.

I have one of these on the way and a oscilliation issue is a way different issue lol.

datasheet.ciiva.com/29312/420965-da-01-en-poscap-tpb-6-3v-150uf-29312653.pdf

The resonant frequency depends on the series and is approximately 100KHz to 1MHz and above this frequency the impedance increases as it is dominated by the series inductance. Can you see this on page 19

www.cde.com/resources/catalogs/ceramperf.pdf

(MLCC) soldered in parallel with POSCAP are normally needed to lower the impedance at the higher frequencies and have a resonant frequency of approximately 10MHz to 100MHz .

What does this all mean? POSCAP and MLCC capacitors are normally connected in parallel on the PCB These planes have extremely low series inductance and provide low power supply impedance from 500MHz to above 5GHz. The capacitance value however is much lower than MLCC and POSCAP types.

So with out MLCC you may have signal integrity issues depending on the supply current and the capacitance needed.

My point is that this is not as simple as its just a bug in the driver.
 
Wow, Jay said he was on the phone with AIB's September first and none of them knew about the new card. Reviewers got the card before the AIB's did. lol what the hell?!
 
Wow, Jay said he was on the phone with AIB's September first and none of them knew about the new card. Reviewers got the card before the AIB's did. lol what the hell?!

GN basically said the same thing regarding drivers, reviewers got stuff before AIBs. I guess the alleged reasoning was to prevent leaks, but clearly that didn't work anyway.
 
GN basically said the same thing regarding drivers, reviewers got stuff before AIBs. I guess the alleged reasoning was to prevent leaks, but clearly that didn't work anyway.

Always works out that way, trying to avoid one hole and end up falling into one you didn't know was there.
 
Back
Top