Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by kurtextrem, Mar 20, 2019.
This is that quadro drivers are better placebo crap all over again.
Looking forward to Unreal Engine 4.22 and RTX on a GTX lighting builds...very cool, especially with new CUDA/GPU calcs thanks to a few dedicated coders taking the lighting builds away from the CPU to the GPU.
As it stands, CPU lightmas builds take forever, but with these new 'hacks' massive time saves can be had, just using a 1060GTX.
Impressive. Most impressive.
so maybe that's why they are making creator ready drivers, i see rtx features on gtx more usefull in productivity since games benefits more from rt cores.
Indeed. Just as a bit of back story, Unreal Engine has always used the CPU (C++) to calculate the photon/radiosity of a level/map. It has been a little bit of a bug bear, knowing that the GPU can do these calculations much faster than a CPU, using vram, and Epic Games steadfastly refused to use anything other than the CPU.
The coder in question released his code onto the forums and the feedback is self-evident:
And if the 70 page thread and it's comments over the last few months and not impressive enough, Epic have now hired the individual directly...
You don't need RTX features to run a CUDA program on GTX cards.
There's some post on nvidia reddit, here's a picture:
I see massive difference in DX11 minimal framerates in Uningine Heaven benchmark, did anyone test it and found similar results or it's case-specific scenario here?
Not seeing that here (3570k/GTX970), min was like 29 fps.
There is absolutely no difference in configuration between these two drivers for Geforce parts, there are no optimisations or driver changes snuck in, it is the same driver
You see what you want to see, which is wishful thinking.
Your system is never in the exact same state no matte how you try to compare like for like, this is normal margin for error.
The Heaven min FPS is not counted by any body desiring to base their system performance on Unigine, as there is a momentary hitch in the beginning which tanks fps.
You are correct but all his other results are within normal margin of error except for those FPS drops in Heaven. I was just wondering did anyone else did any tests or it's that guys setup.
I posted above I am not seeing low min fps in Heaven with GRD 419.67, have not tried CRD.
Who knows why he got those results, it might even be a bug with even switching between D3D11 and OpenGL.
Don't take that heaven min fps as a sure thing..
Benchmark is flawed by min fps, one time it can show 9-10fps the next 30-35fps.
Same by metro benchmarks
Yes, the first run of the Heaven bench always has a hitch or two. Metro is the same, the first run through has a hitch or two. That's why I always let them run through once prior to running it in benchmark mode. If you do that, you should have very accurate and repeatable results.