The Initial Performance Of NVIDIA's R515 Open-Source Linux GPU Kernel Driver

[phoronix]

Phoronix: The Initial Performance Of NVIDIA's R515 Open-Source Linux GPU Kernel Driver

As outlined in yesterday's extensive article about NVIDIA's new open-source Linux kernel GPU driver, currently for consumer GeForce RTX GPUs the driver is considered of "alpha quality" while NVIDIA's initial focus has been on data center GPU support. In any event with having lots of Turing/Ampere GPUs around, I've been trying out this new open-source Linux kernel driver on the consumer GPUs. In particular, I've been curious about the performance of using this open-source kernel driver relative to the default, existing closed-source kernel driver. Here are some early benchmarks.

https://www.phoronix.com/vr.php?view=31119

[V1tol]

I wonder if performance of open part could be better if built with something like

-march=native -O3

[jeisom]

That was much better than I was expecting with power management still to be implemented.

[Vorpal]

Really cool that nvidia is moving in the right direction. Unfortunately I have 10-series cards that now sit in the awkward spot between nouveau and this new driver. For my desktop I guess I can upgrade in a few years time when it becomes a problem. For my laptop though...

[RejectModernity]

Will this open driver work with Nvidia Jetson Nano? My guess is no because it has maxwell gpu

[bple2137]

If that's "alpha quality" and there's a room for optimization, in time the open driver could maaaybe even outperform the proprietary one... Even if it will be just on par, it looks promising!

[Myownfriend]

Really cool that nvidia is moving in the right direction. Unfortunately I have 10-series cards that now sit in the awkward spot between nouveau and this new driver. For my desktop I guess I can upgrade in a few years time when it becomes a problem. For my laptop though...

Same :-(

[sinepgib]

It looks rather good for an alpha release.

[pWe00Iri3e7Z9lHOX2Qx]

I assumed compute performance would essentially match the binary driver given the early data center focus. I expected 3D to be a bit of a mess. The fact that the new driver was not only completely stable in gaming tests, but was nearly even performance wise with the binary driver there as well is a bit of a shock.

[CochainComplex]

I wonder if performance of open part could be better if built with something like

-march=native -O3

I would also like to see those results. maybe the biinary is by default compiled with -o3 and/or x86vX