3080 FE crashing at 2Ghz

[Method]

Memory overclock seems to do great. +600 no issues. +45 on core clock seems stable. +50 however seems to bring it above 2ghz (cooling permitted) and timespy benchmark crashes pretty consistently. 

 

I've seen all the stuff on the power delivery possibly being a culprit for AIB's, but I have founders edition. Maybe a driver issue? I'm definitely on latest with clean install.

 

Any thoughts?

 

Edit: I have a 1000watt G2 powersupply, so I'm certain its not that. 

[dfsgsfa]

what you expect over base clock?

[Shlouski]

Would be nice if it's only a driver issue that can be quickly fixed, would save a lot of people some trouble.

[xg32]

23 minutes ago, Method said:

Memory overclock seems to do great. +600 no issues. +45 on core clock seems stable. +50 however seems to bring it above 2ghz (cooling permitted) and timespy benchmark crashes pretty consistently. 

 

I've seen all the stuff on the power delivery possibly being a culprit for AIB's, but I have founders edition. Maybe a driver issue? I'm definitely on latest with clean install.

 

Any thoughts?

 

Edit: I have a 1000watt G2 powersupply, so I'm certain its not that. 

at this point we can only hope it's not the power delivery problems that plagues the other cheap cards, but on a higher clock on the FE.

 

Or it could jsut be simple silicon lottery.

[Briggsy]

buildzoid did a video recently on the power delivery, although he spent half the video rambling about poscaps.

 

frequency very well could be hitting a wall with Samsung's 8nm process node, or the architecture itself. There reaches a point when you can't keep a chip stable no matter how much power you throw at it. Reminds me of first gen Ryzen and the 4Ghz barrier.

[VeganJoy]

has any overclocking on ampere yielded actually decent gainz? or is this gonna be a gen when you just get whatever looks good and isnt especially bad? itd be weird to get a strix or ftw3 if the tuf or xc3 overclock and cool similarly.

[xg32]

38 minutes ago, VeganJoy said:

has any overclocking on ampere yielded actually decent gainz? or is this gonna be a gen when you just get whatever looks good and isnt especially bad? itd be weird to get a strix or ftw3 if the tuf or xc3 overclock and cool similarly.

afaik it's all crap and within 3% of each other, the stock settings around 1950 is already at exponential part of the voltage curve (think if an old zen 2 with 4.2 stock, and the wall is at 4.3), undervolting the card to 1800-1850 at 0.8v 250w seems to be the way to go, but if thats the case, rip aib models, though some people will still wanna squeeze out the last 5-7%.

 

We already knew the launch bins were crap, let see if they improve over time, i doubt it'll be significant either way. best hope is a card 5% faster than the 3080 for about 800usd later down the line with more vram.

[Method]

EVGA precision seems to be giving me trouble. Getting _terrible_ 3dmark scores even with it just open and defaulted. Afterburner however got me to 2040 Mhz and I was able to be stable with it and got this score:

 

https://www.3dmark.com/pr/336660

 

+50 core, +500 memory

 

It seems the moment I hit 2055 mhz, I crash.

[VeganJoy]

14 minutes ago, xg32 said:

afaik it's all crap and within 3% of each other, the stock settings around 1950 is already at exponential part of the voltage curve (think if an old zen 2 with 4.2 stock, and the wall is at 4.3), undervolting the card to 1800-1850 at 0.8v 250w seems to be the way to go, but if thats the case, rip aib models, though some people will still wanna squeeze out the last 5-7%.

 

We already knew the launch bins were crap, let see if they improve over time, i doubt it'll be significant either way. best hope is a card 5% faster than the 3080 for about 800usd later down the line with more vram.

hmm thats interesting, at least many of the "budget" models like the tuf seem pretty solid with cooling and all that. with the 3090 being revealed to not be a titan i wonder if theyve got a real titan on the backburner thats even more kickass, with an accompanying 3080ti to shit on rdna 2. exciting times lmao

[Maticks]

Seems little point overclocking this generation it appears that GPU Boost this time around actually auto boosts the card to its maximum limit.

1950 and the odd 2000 spike is the limit for these things on air, maybe water we will see 2000 stable but i doubt 2050 will be a thing.

 

Good job for Nvidia not leaving anything on the table this time around though but for the overclocking community there isn't a lot left for us to do now.

Just play the games and enjoy..

[VeganJoy]

14 minutes ago, Maticks said:

Seems little point overclocking this generation it appears that GPU Boost this time around actually auto boosts the card to its maximum limit.

1950 and the odd 2000 spike is the limit for these things on air, maybe water we will see 2000 stable but i doubt 2050 will be a thing.

 

Good job for Nvidia not leaving anything on the table this time around though but for the overclocking community there isn't a lot left for us to do now.

Just play the games and enjoy..

i wonder which makes consumers happier; gpu running at 99% of available performance out of the box or maxwell with 30-50% overclocking gainz? im guessing they went right up to the absolute limit this gen to hook people who waited out turing on this gen. 

[Maticks]

1 minute ago, VeganJoy said:

i wonder which makes consumers happier; gpu running at 99% of available performance out of the box or maxwell with 30-50% overclocking gainz? im guessing they went right up to the absolute limit this gen to hook people who waited out turing on this gen. 

I guess in previous generations its been a 25% speed increase out of the box and overclocking can get another 20%.

This generation isn't really that big a leap its the same as always, they've just managed to get the auto boost technology to boost a lot higher hitting nearly that 100% overclock threshold we've usually had to manually tweak to get.

 

[Blademaster91]

46 minutes ago, VeganJoy said:

hmm thats interesting, at least many of the "budget" models like the tuf seem pretty solid with cooling and all that. with the 3090 being revealed to not be a titan i wonder if theyve got a real titan on the backburner thats even more kickass, with an accompanying 3080ti to shit on rdna 2. exciting times lmao

Theres rumors of a Titan card with 48GB of vram, but who knows with AMD being quiet about their RX 6000 cards but if AMD has anything i'm sure Nvidia would have a 3080Ti with more VRAM to beat AMD.

5 minutes ago, VeganJoy said:

i wonder which makes consumers happier; gpu running at 99% of available performance out of the box or maxwell with 30-50% overclocking gainz? im guessing they went right up to the absolute limit this gen to hook people who waited out turing on this gen. 

I'd be less worried about how reliable these 3080's are if there was more room to OC, I usually just let the GPU boost on it's own, there's really not much point in overclocks unless the card is on a water loop or going for benchmark scores.

[Maticks]

10 minutes ago, Blademaster91 said:

Theres rumors of a Titan card with 48GB of vram, but who knows with AMD being quiet about their RX 6000 cards but if AMD has anything i'm sure Nvidia would have a 3080Ti with more VRAM to beat AMD.

I'd be less worried about how reliable these 3080's are if there was more room to OC, I usually just let the GPU boost on it's own, there's really not much point in overclocks unless the card is on a water loop or going for benchmark scores.

I get my EKWB in 2-3 weeks time will see on water what it does.. the data online is showing that we won't get much if anything at all. it seems to be the most the card will let you do is going over 2050Mhz at 30c. I doubt a waterloop is going to get that cold. will be interesting to see if the few Watts from having the Fan power draw removed make any difference.

At this point a few Mhz is going to make not much difference 1FPS if you are lucky but with likely instablity.

Getting it stable at 2000Mhz is the best we can hope for.

 

Where the RX 6000 is going to sit performance wise and in addition what overclocking headroom it has will both be an interesting point here.

Even if it's a toaster oven of power usage, if the performance is there overclockers will use it.

 

The RTX3080's certainly are watercooled for aesthetics and nothing more in this generation from what is online so far with maybe 1%-3% performance benefit.

[Mark Kaine]

2 hours ago, Method said:

EVGA precision seems to be giving me trouble. Getting _terrible_ 3dmark scores even with it just open and defaulted. Afterburner however got me to 2040 Mhz and I was able to be stable with it and got this score:

 

https://www.3dmark.com/pr/336660

 

+50 core, +500 memory

 

It seems the moment I hit 2055 mhz, I crash.

yeah ok but as long you aren't crashing with stock settings you kinda got what you paid for ? And about overclocking that is a different issue than what many AIBs seem to have.

 

Also drivers can't be mature at this point either

 

[Mark Kaine]

2 hours ago, VeganJoy said:

i wonder which makes consumers happier; gpu running at 99% of available performance out of the box or maxwell with 30-50% overclocking gainz? im guessing they went right up to the absolute limit this gen to hook people who waited out turing on this gen. 

most people ("consumers") absolutely hate overclocking... it's why they often buy expensive AIB cards, or Alienware () computers and be done with... (they also do not overclock RAM etc, way too dangerous!)

[Briggsy]

7 hours ago, Mark Kaine said:

yeah ok but as long you aren't crashing with stock settings you kinda got what you paid for ? And about overclocking that is a different issue than what many AIBs seem to have.

 

Also drivers can't be mature at this point either

 

Yeah for performance gains at "stock" clocks it will be interesting to see if there's any more headroom left in these cards through driver optimizations.  As for stability, anyone hoping that drivers can fix this are barking up the wrong tree. Firmware update is the only option for existing cards if someone doesn't want to tweak settings in afterburner/tools, other than someone hardware modding their card and adding/replacing capacitors.

[Mark Kaine]

1 hour ago, Briggsy said:

Yeah for performance gains at "stock" clocks it will be interesting to see if there's any more headroom left in these cards through driver optimizations.  As for stability, anyone hoping that drivers can fix this are barking up the wrong tree. Firmware update is the only option for existing cards if someone doesn't want to tweak settings in afterburner/tools, other than someone hardware modding their card and adding/replacing capacitors.

it's just so weird to me they'd use the "wrong" capacitors, maybe they're faulty? Or maybe not, in any case I think there should be recalls for affected cards as it's a product that's not functioning as advertised and also a marketing disaster in the making if they don't fix the issue properly. Downclocking (or gimping) with a software update doesn't strike me as a good idea at all...

[Panoramix97]

I have read about all this but guys from what I understand the cards works totally fine except if you overclock them ?

 

What is the problem then ? Since when Nvidia garantee you will have overclocking room for any card ?

 

From what I saw the cards not overclocked doesn't go over 1950 mhz ?

[Briggsy]

8 minutes ago, Panoramix97 said:

I have read about all this but guys from what I understand the cards works totally fine except if you overclock them ?

 

What is the problem then ? Since when Nvidia garantee you will have overclocking room for any card ?

It's not really an issue with manual overclocking. Under certain conditions the GPU boost will push the clocks beyond stability, without any manual overclocking. The cards with cheap capacitors behind the core don't filter out enough noise to keep the cards stable under these 2+GHz boost conditions. It's general consensus at this point that AIB's will probably release a new Bios to nerf GPU boost, as opposed to any recall happening. It seems Zotac already did this before launch with the Trinity card, which explains why it doesn't boost as high.

[BTGbullseye]

36 minutes ago, Mark Kaine said:

I think there should be recalls for affected cards as it's a product that's not functioning as advertised

Except that they already countered that particular issue by making the guaranteed boost clock only 1710MHz, and the additional boost on top is not guaranteed at all.

[Mark Kaine]

1 hour ago, BTGbullseye said:

Except that they already countered that particular issue by making the guaranteed boost clock only 1710MHz, and the additional boost on top is not guaranteed at all.

oh yeah that's right for the FE cards , still for others it might be an option. In any case this is a mess and while maybe expected not a great show for these cards. And it'll be interesting to watch how it all unfolds and eventually be resolved.

[Method]

After more testing and playing more games, it does appear my FE crashes with no manual overclock. MSI afterburner is set to 0 across the board, and I can consistently crash in COD multiplayer after about 5 minutes. GPU temp really doesn't go up much, maybe 60c, so I think it does boost very close to 2ghz and eventually crashes. Setting -50 in Afterburner seems to keep the gpu at 1950mhz and so far has been stable.

[Mark Kaine]

A true overclocking beast. Nvidia did it again! 

 

Spoiler

(I'm sorry)

 

[Briggsy]

5 hours ago, Mark Kaine said:

it's just so weird to me they'd use the "wrong" capacitors, maybe they're faulty? Or maybe not, in any case I think there should be recalls for affected cards as it's a product that's not functioning as advertised and also a marketing disaster in the making if they don't fix the issue properly. Downclocking (or gimping) with a software update doesn't strike me as a good idea at all...

EVGA said it took a week of testing to even catch the problem and diagnose. Adds credence to the theory that Nvidia rushed these cards to market, especially given that some FE cards are having the problem as well with 2 arrays of the smaller caps. According to Buildzoid the smaller caps are more stable at higher frequencies, which could explain why its happening above 2Ghz for the bigger caps. So in theory the bigger caps can hold more charge, but are slower at feeding the power to the core. It would be nice to have an official response from Nvidia that wasn't a bunch of bs. Like its not even that big of deal, but I'm not going to feel confident buying any model of 3080 until there's a non-bs resolution. So far I've seen reps on youtube and elsewhere trying to downplay it all and say its all just overclockers. We know that isn't true, and I wish they'd realize that their customer base is partly made up of nerds that will pick apart a bs response.