Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by RealNC, Mar 15, 2021.
that is on a 5950x btw, not a 1600 like they tested for ampere
Showing results on an extremely high-end CPU would, of course, remove any possible driver overhead.
So can my 3060 use 456.14?
I dunno dude.I have a 9900k/3080 and I encounter kinda frequent random frametime spikes (some games have it worst and some not much)Thankfully I have a g-sync compatible monitor so it softens the micro stutters alot
Honestly I think it's more of a thing I noticed since I got my 3080.On my old 1080ti I used to stay on driver 442.59 and the frametime was def better then my current 3080 so I don't know if this driver overhead issue came after or before this driver but being on a 3080 I can't go back that far.
Wierdly enough I encounter more random frametime spikes (micro stutters) on 1440p then on 1080p which makes no sense since upping the resolution should put more load on the gpu then the cpu.(yea the the frametime is definitely better on 1080p then 1440p)
Currently using 466.11 on my 3080.
You're looking at 1080p Ultra Low. This game at such setting will likely be CPU limited on any CPU and if a GPU has a higher CPU overhead then it will affect the results. Which isn't the case here, which in itself is an interesting result considering that Dirt 5 is using DX12.
In some cases you can hit a ram bandwitdh limitation too.
Would this count as part of the overhead issue?Seems to mainly happen on 1440p.(Much lesser degree or nothing on 1080p)
Another thing I tried is disabling CFG globally which does visibly improve things.
All your examples there are running on D3D11.
Is that just an observation that you noticed or is there something you would like to add as to why that would matter?(honest question)
Nv doesn't have a driver overhead issue in D3D11. This thread is about D3D12 and Vulkan.
Oh.I guess what I'm having is unrelated then.
Considering disabling CFG made a difference I should have guessed it was an OS issue and not actually a driver issue :/
No response from Nvidia on this one after so many months ? Are they even aware of the overhead problem ?
why is nvidia going to address an issue that never existed.
The issue exist and is sometimes seen in games even on high end CPUs but its actual impact is very limited (how many people have systems like 1600+3080 these days?) and thus they are unlikely to bother doing anything about it.
I'm not aware of any followup tests by those youtubers. They seem to just have gone silent about this. Which kind of suggests to me they were full of crap? Made a video to get clicks and then pretend nothing happened.
I believe that if this was a true issue, they'd be making videos about this like no tomorrow.
Iam running RTX 3090 OC with my Celeron, Nvidia should fix it.
Let them update the control panel from geforce 2 era first then they can fix the software base scheduler.
they weren't full of crap.but they weren't empty of crap either,as usual.
the extent to which this causes problems has been intentially left out.
they showed it's a problem on ryzen 1600 (which is about the slowest single core gaming cpu you'll see these days) to 5600x,probably the fastest single core gaming cpu which showed no impact.nothing in between.but that's what hw unboxed do.
watched their today's video and lol'd when they said they still prefer a higher vram buffer over better lighting in games,quoting mfs2020 as an example
Geforce 2 had the nvcp built into Display.cpl
Not just single-core, especially multi-core latency is terrible in most games + there is only half-rate AVX2. Sticking with Zen 1/+, while beeing able to afford recent games or GPUs, is very much self-own.