No crashes yet with palit 3080 gaming oc. Using Seasonic 1000w titan

as posted in the 456 driver page I've got the same crashes on a 1080ti and if you don't know who you're talking with I have 440hours of subnautica 1625hours of skyrim etc...I easily play more than 40hrs a week and have done so for more than 10 years I know some games better than devs ever did (with modding tools) so if I say it's the drivers..it's the drivers I rolled back to 452 no more CTDs sad to see people doing x-files theories when the drivers are 100% faulty in the 1st place and yes a higher clock/overclocking doesn't help, it's the 1st thing I did when I started CTD every xx minutes on borderlands 3 on which I already had 130hours with zero crash before those 456 drivers..going back to a slower clock around 1920Mhz (which is the 100% default clock on my msi) only allowed me to play longer before CTD than 1974Mhz but they were still ther I also have more fps with the 452 drivers so something is wrong with those jayztwocents made a video about overclocking 3080 memory and it kills the fps because they do so many errors so it may be that the faster 3080 memory get it worse than me yes I know the 2080ti launch was a nightmare, but when they glitched I didn't have the same exact problem with a previous gen card, and now I have the exact same problem 3080 have at the same time that was on intel but I have a threadripper pcie 4.0 same thing with the 1080ti and a 1200w psu and my 1080ti draws 320+watts too (with a 384watts peak once) so none of those are the reason of the crashes in case you tought they were

prices for the cheapest rtx 3080 in Europe = 1012 dollar / 850 euro nvidia founders edition 840 dollar / 720 euro even these 3080's aren't gonna sell That good i think , at least not in Europe i'm good with my 2080ti oc , that's oc'd to 2100 for now that's way too much , they're taking the piss again , nvidia , with these prices rtx 3090 ... 2000 euro / 2300 dollar just laughable !!

Clanmate had it yesterday, not in every game though. Looks like it's only in certain games where crossing the 2000 Mhz barrier causes the issue for him.

Well, this does actually make me feel a bit better for not being able to get my hands on one... As for the FE, yea, that part in the middle gets super, super hot. Say at Jay's 2 cent that he put a 120mm fan pushing air in on that area and temps went down a LOT. So I guess the AIBs with their triple fan designs will prove to be the more sensible choice for the long run. Either way, a lot of research and comparisons required before making a purchase. Given the shortage situation, it will give me time to do just that.

Done an AMD.

Is this the new GTX 480 generation?

TheDeeGee:

Is this the new GTX 480 generation?

Big, hot, compute heavy, crashes, overpriced...fermi 2.0. Remember amd at that time had revolutionary hd 5000 series and it was actually winning. Its time...

If uderclocking and undervolting is the temporary solution, this suggest, power delivery issues which is very much depended on the PSU and that is a fact since founders edition doesnt have that problem only the OC AIB releases. Something to think about.

Astyanax:

Guess you got one of the good ones 😉 The blackscreen isn't a crash as opposed to AMD's. It's literally VRR/Freesync displays going out of range and can be mitigated by turning the feature off screen side.

Some of them were, my initial display with the 5700XT Pulse didn't even have free-sync or adaptive sync support. Over time and especially post the 20.2.1 drivers though much of this has been resolved but not everything. Perhaps NVIDIA could deploy a similar fix to what AMD did in a newer driver and tweak the algorithm a bit maybe make it a bit less aggressive or at worst do a driver hot fix release and just cap max boost speeds a bit lower temporarily while doing some tweaks to how the card scales with power and temperature and other limits and factors for what I guess NVIDIA Boost Version 4 handles this. Hmm maybe GFE on it's own and a higher power draw could mitigate the problem but if it's temperature related for VRM's or the VRAM it's just going to make it worse short of underclocking. Testing will have to confirm effectiveness and any possible mounting issues or cooling for components besides the GPU core too for some of these models I suppose. EDIT: Could be reading this quoted bit about AMD incorrectly. Well also that AMD could have defined "Black screen." a bit better besides half a dozen fixes for it without details on what they actually do for fixing them. (PCI 4.0 compatibility, display cabling both HDMI and DP, power delivery and PSU quality and power cabling especially if it's daisy chained plus hardware sensitivity in general and additional stuff I'm not very good at.) EDIT: And for AMD's little tweak the Navi10 cards are a bit more conservative in boosting now so with the target clock speeds so there's a bit more of a gap between Wattman and actual achieved clocks and users might need some extra power or lowering the GPU temps before it can boost up to full near 2100 Mhz again or sustaining these for a longer period of time. Depending on the card and the GPU binning, some require a bit more to go past 2.0 Ghz before getting close to 2050 - 2100 Mhz. (Earlier drivers like the 19.7's or the 19.8's were really quick to target max clocks in comparison.) Suppose that's the same here though, some Ampere cards might land around 1900 Mhz instead of 2.0 Ghz but that's still a good bit above the 1750 Mhz listed boost clock speed. 🙂 Well the driver code is pretty flexible, probably some interesting things NVIDIA could do to add or refine the boost algorithm in later updates here or other micro code or firmware enhancements perhaps. Depends on what the issue is I guess, getting a few sample cards and replicating the error shouldn't be impossible for a answer to that but it could take a little while. (Likely a priority issue for NVIDIA though to get this resolved quickly.)

Next up, The 3090

It's difficult to speculate where the actual problem lies at the moment with just a subset of none scientific results being reported in. Currently it seems to indicate multiple board designs being affected when they run at full whack. Perhaps it is a firmware / BIOS update as advised by nvidia that is required. As mentioned above, both AMD and nVidia have issues with new releases, it's not a surprise as using multiple AIB's is quite complex to cover all the permutations. This is why graphics cards have more issues than consoles on launch, because consoles are tightly controlled on the permutations and allows for better testing / debugging. nVidia has not helped here though. Their board design, the new funky one for the cooling on the FE, is completely different to a traditional 2 or 3 fan design the AIB partners now have been doing for years. In the past nvidia even restricted what the board partners could do board wise. Not here. Also nvidia seems to have limited drivers to the AIBs until the last moment etc. I think they new they had a killer product and wanted a shock and awe for release unlike the drip of results. It worked, they got the awe factor. Now users are getting the shock factor ! 😀 I'm not going to be too critical though. It was nice to really be impressed by the generation gap this time and the cost for that ... like the good old days. Having said that, the people who rushed to buy the new shiny thing have to know that what test pilots might have to put up with as they glow in the fact they are one of the first to have it ... I'd wait for prices to settle and bugs to iron out, reading the reviews and judging the results of the performance and build quality to make an informed decision that is obviously the best 😉 I can then do this to everyone [youtube=eH56UOjDQ4A]

hamltnblue:

Next up, The 3090

Well the reviews for that and the initial available shipments for this GPU will be later today but as for the issue itself with the extra hardware will the card even boost to the same levels as the 3080 though with said extra cores and what not does it need to before exhibiting similar issues? EDIT: Suppose it depends on if it's a heat issue, voltage or power draw problem above certain speeds just spiking for what the GPU demands to be stable at these clocks. Needs testing but could be worked around until a full solution has been worked out by just reducing the boost a bit if that's where it runs into stability problems. 🙂 (Temporary workaround but I guess that would need testing too for if it's enough.) Or bios updates perhaps as a alternative. Though I don't see why tweaking the algorithm and behavior through the driver software couldn't work and thus avoid needing more than a driver update.

In days gone by, things were built with plenty of spare safe running capacity. Nvidia appears to have sacrificed some of that capacity for performance gains. Hence the very low OC potential on recent cards. Just as when you OC to the limit you start getting random crashes, maybe these cards are already there.

ThermaL1102:

prices for the cheapest rtx 3080 in Europe = 1012 dollar / 850 euro nvidia founders edition 840 dollar / 720 euro even these 3080's aren't gonna sell That good i think , at least not in Europe i'm good with my 2080ti oc , that's oc'd to 2100 for now that's way too much , they're taking the piss again , nvidia , with these prices rtx 3090 ... 2000 euro / 2300 dollar just laughable !!

One can still find 3080s for 750€ in Sweden at least (not in stock of course). It's pretty much confirmed that nVidia is giving discounts to wholesalers or stores in the beginning but that the discount will disappear in a few weeks.

Netherwind:

Phew, hadn't had that problem yet but on the other hand the card only boosts over 2GHz occasionally and doesn't really stay there for a long time. Tried some OC while running FS Extreme and the card had no problem boosting up to 2070MHz for a short while but on the other hand such tests only last a moment. Played Control yesterday which is quite demanding but I'd say that my clock average was 1950-1980. I'd check with HWInfo but since the card clocks down completely when entering menus or inventory it wouldn't give an accurate reading unless I just fool around for 2-3 minutes.

Could you,please,put a picture of HWInfo (GPU&VRM) after you play Control or other game(after 1 hour) who push the limits of the card,to see max/peak voltage and ampers, on G3D Rtx 3080 owners Thread.Thank you. Could be anything,but many users have some older Psu who dont have 2 separate cables for this 2x8 12V power and cannot sustain this new power hungry video card.Underclock is a good solution,but for how long?

ThermaL1102:

prices for the cheapest rtx 3080 in Europe = 1012 dollar / 850 euro nvidia founders edition 840 dollar / 720 euro even these 3080's aren't gonna sell That good i think , at least not in Europe i'm good with my 2080ti oc , that's oc'd to 2100 for now that's way too much , they're taking the piss again , nvidia , with these prices rtx 3090 ... 2000 euro / 2300 dollar just laughable !!

In the Netherlands you'll get an ASUS TUF non-OC for €729,- So not the €850 you say is the cheapest in Europe, cause last time I checked, the Netherlands are still in Europe 😉 And the cheapest RTX3090 is €1549,- And the average waiting list in my country is ~300 people per AIB in some cases for the RTX3080, a couple of thousand are in pre-order here. So I dont know where you get your info from, but its wrong. They sell like hot buns

hamltnblue:

Next up, The 3090

The 3090 is basically a "halo" product with a limited supply. So not such a big impact if there are issues, which I am inclined to think there will not be.

I was eager to get one. Not anymore. Doubt AMD will have anything interesting , so i guess i will wait for the refresh of this card.

yikes staying on my 2070s and waiting for Big Navi is looking better and better