NVIDIA GeForce Version 430.39 WHQL - Download and Discussion

Performance up from the previous driver "430.00" and smoother in ESO "stutters gone". And higher fps.
Windows 10 ver 1903
Going to stay with these for a bit.

 

Ouch, didn't notice that 800 series got retired to legacy status. I wonder when it'll be 900 series' turn?

 

Played it for a few hours with these drivers and didn't notice anything weird, running stable Win10 not Insiders.

 

I have fast boot enabled, but hardware fastboot is off and it looks ok

https://i.**********/rRNwSrZY/nvcontainer.png

 

Hey, benchmarks are not totally useless, if you run a number of synthetic benchmarks on a driver and see a general downturn, then it's not a long shot to think that it can be quite likely to occur in actual games too. I do run game benchmarks too as additional comparisons though. To be fair, NVidia drivers have over the last couple of years always given pretty much identical benchmark runs on synthetics as well as game benchmarks on my GTX 1070, but I like to run them anyway just to make sure I'm not running a dud driver! A few years ago it used to be the thing that NVidia would state "7% performance increase in [such & such general scenario]", so some drivers used to provide performance improvements kind of 'across the board', but the last couple of years there doesn't seem to have been any magical drivers that have given general performance improvements. For specific games they sometimes release performance improvements, but a long time since 'general performance improvements' seen.

 

Does fast startup still works with HW fastboot off?

 

after 600 and 700 gets retired

 

i have fast boot too and nvcointainer never gives me any issue,

but i have 6 of them

 

When I turn on the computer, the NVIDIA container task manager all the time shows 15% of the CPU, when I restart the computer, the NVIDIA container process becomes normal 0% CPU, Windows 10x64 1903 18362.53, NVIDIA Driver 430.39 Standart. Sorry for my english.
425.31 driver and older versions are ok, the problem is only in driver 430.39.

 

Are you on 1809?

 

Anything the GPU does is extremely parallel. However, the problem is coherent fast communication between multiple GPUs. If they could solve that, they would make many-die GPUs instantly instead of huge monolightic ones that cost way too much money to make.

 

I spoke too soon, first time seeing it, mine was at 17%, i resumed from sleep started office word and then when I pressed copy/paste form web browser it stalled the program, I checked taskmaneger and saw nod32 & nvcontainer both at 17%.. Nod stopped but nvcontainer (service) is still 17% as Im writing this, never saw such thing before.. Will install something else now, probably Quadro variant otherwise back to Vulkan 425 driver.


Also Prey v2017 crashes, or has black screen by ingame start menu and doesn't want to start properly with DSR either, in that case it can crash with dx11 api crash.

 

I see, thanks for reporting this. Good thing I didn't mess with the BIOS yet. I bet hotfix is already in the works.

 

Not really. Something which is time dependent can't be parallel and there are a lot of time dependent shaders in modern rasterization which can't run in parallel despite being natively parallel on their own. This severely limits the application of mGPU rasterization methods like AFR.

This "problem" isn't a problem, it's a given for any mGPU system. And the biggest issue here is that this communication get exponentially more complex and costly once you go above just two GPUs.

You will never be able to have the same extra bandwidth as your intra bandwidth. You can improve it, sure, but it will still be several times or magnitudes less than what you'll have inside one multiprocessor. Thus this will always be an issue for something like AFR in modern rasterization techniques.
RT on the other hand can run almost independently on one node and scale to an unlimited number of nodes easily.

 

IIRC that NVContainer high CPU usage bug happened before and was fixed later, seems to have resurfaced.

So yes, hotfix it is then. Again.

 

With these new drivers I cannot change "Output color format", "Color depth" or "Output dynamic range" anymore.

Is there a way around this?

 

no, 1903 18362.53, is this a 1809 issue only?

 

I think it only affects 1903

 

I'm still on Win 10 1803, but I had a problem with this driver and had to revert back to 425.31. Everything worked well until my G-Sync monitor went to sleep. It would easily wake up again, but then the system would become very unresponsive in that the mouse would move / screens would scroll with big pauses interspersed with short periods of normal activity. Nothing short of a reboot would put things back to operating normally until the next time the monitor went to sleep. There wasn't any CPU or Drive Access activity going on when things bogged down, so I'm guessing it's something related to WDDM 2.6 trying to run on 1803. In any event, going back to 425.31 completely resolved the issue.

 

Strange I just now got another prompt even after doing the rename trick... You guys notice the same thing?