AmpereGate! RTX 3080 instability

Thats not Fixing the issue. That’s avoiding it.
Until the root cause is found it provides a temporary solution to at worst unusable, and at best unstable cards.
This 'fix' allows function of the cards while determining the true issue is and evaluation whether a recall is needed.
Not sure what fix definition means to you but that pretty much sums it up for me even if its a temporary one.
 
Found this poking around. Since it had Gainwards stance on the 30 series use of caps I decided to post.
6mh1dym95up51.png
 
Thats not Fixing the issue. That’s avoiding it.
As long as it still boosts to the advertised speeds then no one will have a leg to stand on. You shouldn't expect boosts over 2000 on cheap cards. Does it suck? Sure. The only fix I see them doing is releasing a FW update that locks the cheap cards to under 2000mhz.
 
As long as it still boosts to the advertised speeds then no one will have a leg to stand on. You shouldn't expect boosts over 2000 on cheap cards. Does it suck? Sure. The only fix I see them doing is releasing a FW update that locks the cheap cards to under 2000mhz.
Yup, but the question is will they be heavy handed and push it aggressively lower than that to be on the safe side?
 
Found this poking around. Since it had Gainwards stance on the 30 series use of caps I decided to post.

Source? Gainward’s wording seems to imply that the use of MLCCs was per Nvidia’s spec.
 
I can’t help to think maybe they flew too close to the sun if slight capacitor differences causes instability.

It’s sort of a cluster if they backdown on stock clocks but in reality probably what they should do.
 
I can’t help to think maybe they flew too close to the sun if slight capacitor differences causes instability.

It’s sort of a cluster if they backdown on stock clocks but in reality probably what they should do.
I see them putting a hard limit in the 1800 area for boosts just to be safe, even if its 1899.
 
As long as it still boosts to the advertised speeds then no one will have a leg to stand on. You shouldn't expect boosts over 2000 on cheap cards. Does it suck? Sure. The only fix I see them doing is releasing a FW update that locks the cheap cards to under 2000mhz.

Sure you do. The cards are supposed to work out of the box in their default configuration. Underclocking is not the end users responsibility. If the card isn't stable enough to boost to 2000, it shouldn't boost to 2000.
 
Until the root cause is found it provides a temporary solution to at worst unusable, and at best unstable cards.
This 'fix' allows function of the cards while determining the true issue is and evaluation whether a recall is needed.
Not sure what fix definition means to you but that pretty much sums it up for me even if its a temporary one.

It appears the root cause has already been found. How it's going to be addressed is what remains to be seen.

My idea of a fix is about as simple as it gets. When the cards work the way they are supposed to work (without the end user having to take steps to slow them down from their default configuration) then it's fixed.

The best option for anyone with one of these cards is to return them while they still can.
 
Sure you do. The cards are supposed to work out of the box in their default configuration. Underclocking is not the end users responsibility. If the card isn't stable enough to boost to 2000, it shouldn't boost to 2000.

Agreed; also agree with the other poster who said they should just release a firmware update that locks clocks below a certain threshold. OCing is not guaranteed and as long as you get the advertised levels I see no problem for them.
 
Yup, but the question is will they be heavy handed and push it aggressively lower than that to be on the safe side?
I doubt it will be noticable to most people. The issue seems to be when you push it too around 2000mhz. Which most the cards don't even reach at stock.
Sure you do. The cards are supposed to work out of the box in their default configuration. Underclocking is not the end users responsibility. If the card isn't stable enough to boost to 2000, it shouldn't boost to 2000.
I agree. That's why there more then likely will be a FW upgrade to "fix" the issue. As long as it is still above the advertised specs of the card customers won't have a leg to stand on. AiBs didn't do their due diligences with this release.
 
Last edited:
Agreed; also agree with the other poster who said they should just release a firmware update that locks clocks below a certain threshold. OCing is not guaranteed and as long as you get the advertised levels I see no problem for them.

Yes, but stability is guaranteed otherwise the product is defective. It is the manufacturer's responsibility to ship firmware with boost bins that the hardware can actually handle.

My understanding is that the issue doesn't have anything to do with end user overclocking. It's the automatic boost mechanism that's taking clocks above the stable limit.
 
Honestly, anyone with one of these cards showing stability issues really should return it and get their money back. Don't let nVidia and Co hang on or keep your money while they figure out the best way to slow your card down. Yeah, it'll suck to have to be without your new 3080 for a month or two while the dust settles and inventory is available, but the end result is you'll get a better card, and if enough people demand their money back, the odds of a repeat lessen moving forward.
 
Yes, but stability is guaranteed otherwise the product is defective. It is the manufacturer's responsibility to ship firmware with boost bins that the hardware can actually handle.

My understanding is that the issue doesn't have anything to do with end user overclocking. It's the automatic boost mechanism that's taking clocks above the stable limit.

Yes but these are the typical problems new releases have. I was speaking more from a RMA perspective that it's really nbd to underclock your card temporarily until a fix is out.
 
GeForce GTX 970 Fiasco all over again.
Was going to grab a 3070, glad I waited to see a head to head review with Big Navi.
 
Fact is, this is on Nvidia. I'm not a fanboy, I rock Nvidia and plan to buy the 3080 once available and confirmed this issue is resolved.

However, they should have known (and maybe did) that this was a problem, and maybe didn't give the right specs to their partners. Additionally, an early working driver would have allowed AIB partners to catch this issue before shipping to consumers. Two strikes.

I'd like to see more information to exactly what happened, as it still seems a bit unknown, but it's a botched launch if I ever saw one. I also had crashing issues on my 2080 Ti and had an RMA (and had to RMA a Titan X Pascal before that for VRAM corruption). Not a good look.
 
Fact is, this is on Nvidia. I'm not a fanboy, I rock Nvidia and plan to buy the 3080 once available and confirmed this issue is resolved.

However, they should have known (and maybe did) that this was a problem, and maybe didn't give the right specs to their partners. Additionally, an early working driver would have allowed AIB partners to catch this issue before shipping to consumers. Two strikes.

I'd like to see more information to exactly what happened, as it still seems a bit unknown, but it's a botched launch if I ever saw one. I also had crashing issues on my 2080 Ti and had an RMA (and had to RMA a Titan X Pascal before that for VRAM corruption). Not a good look.

The conspiracy theorist in me thinks the reason this was a paper launch is because they already knew about the problem but didn't want to have to RMA large amounts of cards so they only manufactured a small amount to by themselves time to figure out what the problem is.
 
The conspiracy theorist in me thinks the reason this was a paper launch is because they already knew about the problem but didn't want to have to RMA large amounts of cards so they only manufactured a small amount to by themselves time to figure out what the problem is.
These companies are nefarious enough to make a plan like that, just wish we could find one engineer that's retiring that knows the story and ready to burn the place down on his way out.
 
GeForce GTX 970 Fiasco all over again.
Was going to grab a 3070, glad I waited to see a head to head review with Big Navi.
Are you communicating from the future? How's the civil war going and did Katy Perry lose the baby weight?

Anyhoo, back here in present reality I'm looking to get my mitts on a EVGA 3080 soon.
 
Last edited:
Granted, the RRoD on Xbox360 was much worse, but we have precedent for big companies knowingly shipping faulty product.

As expected from every new tech launch. It's just how bad the issue is and how the parties responsible respond to it that makes it a bigger issue it is or not.
 
As long as it still boosts to the advertised speeds then no one will have a leg to stand on. You shouldn't expect boosts over 2000 on cheap cards. Does it suck? Sure. The only fix I see them doing is releasing a FW update that locks the cheap cards to under 2000mhz.

Gotta be honest, if most of the "issue" is furore of customers not being able to overclock their cards past advertised frequencies, I'm kind of on the side of the manufacturers here. I have a Gigabyte 3080 Eagle OC arriving later this week, and honestly, if it's stable at stock factory-OC clocks I'm fine with this. Nowhere do nV or AIB ever advertise that you'll get a single mhz over what's on the box, and I'm fine with that.

That said, it's a bit unprecedent for a this number of chips to be so unstable at near-spec frequencies. I also highly suspect nV and AIBs knew about the "issue" prior to launch: they were hell bent for leather to get Ampere out to have a month window before RDNA 2 is released.
 
Last edited:
Gotta be honest, if most of the "issue" is furore of customers not being able to overclock their cards past advertised frequencies, I'm kind of on the side of the manufacturers here. I have a Gigabyte 3080 Eagle OC arriving later this week, and honestly, if it's stable at stock factory-OC clocks I'm fine with this. Nowhere do nV or AIB ever advertise that you'll get a single mhz over what's on the box, and I'm fine with that.

That said, it's a bit unprecedent for a this number of chips to be so unstable at near-spec frequencies. I also highly suspect nV and AIBs knew about the "issue" prior to launch: they were hell bent for leather to get Ampere out to have a month window before RDNA 2 is released.
The issue isn't customers not being able to overclock their cards... it's the cards being sold that are factory overclocked and failing. It IS the stock factory-OC that's failing.
 
The issue isn't customers not being able to overclock their cards... it's the cards being sold that are factory overclocked and failing. It IS the stock factory-OC that's failing.

It's not even factory OCs. These are stock clocked cards that are failing when they boost too high. In the video above the card is running TimeSpy fine at ~1950 and then fails when it instantaneously boosts to ~2055.
 
Not sure what rushing this has accomplished since most folks have not been able to buy one. The point of creating hype is to grab a chunk of sales before the competition launches, as of last count 90% of H folks polled have not been able to get one and if Big Navi is indeed very good and available in large numbers a month from now, that is a valid option. Also note that both next generation consoles will be Navi based and with Xbox series X, AMD should have an expanded driver team for Windows so I expect that that situation to improve as well. NV and partners have a month to flood the market and yet they are creating artificial scarcity...what gives?
 
Not sure what rushing this has accomplished since most folks have not been able to buy one. The point of creating hype is to grab a chunk of sales before the competition launches, as of last count 90% of H folks polled have not been able to get one and if Big Navi is indeed very good and available in large numbers a month from now, that is a valid option. Also note that both next generation consoles will be Navi based and with Xbox series X, AMD should have an expanded driver team for Windows so I expect that that situation to improve as well. NV and partners have a month to flood the market and yet they are creating artificial scarcity...what gives?

I had assumed that Nvidia giving AIBs drivers at the last second was intentional to avoid leaks. So not rushed but stupid.

As for launch inventory nobody knows how much stock was available. It would have sold out no matter what.
 
So the "fix" is just run at stock?
Not sure that's even a problem with 95% of the people buying one.
 
Back
Top