Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

FPS Problem Maybe?? #237

Open
ForbiddenEra opened this issue Jul 27, 2022 · 2 comments
Open

FPS Problem Maybe?? #237

ForbiddenEra opened this issue Jul 27, 2022 · 2 comments

Comments

@ForbiddenEra
Copy link

ForbiddenEra commented Jul 27, 2022

Hi,

I posted previously about a bug with FSR in 1.6, there was a pull to fix it seems, though I'm presently not setup for building and the latest release is still 1.6 AFAICT..

Anyway; unrelated, seeing a weird issue with OGL rendering; though I have two GPUs...

CPU: i7 9700
GPU0: RTX 2070 Super @ PCIe 3.0 x16
GPU1: RTX 3060 @ PCIe 3.0 x4
GPU2: Intel UHD630 (built in to CPU)

My main screen is a Samsung Neo G9 49" Widescreen; 5120x1440 resolution.

This screen is attached to the 3060 GPU. I haven't noticed any odd frame rate issues in other applications, though I doubt I use many OpenGL apps aside from Q2.

Was trying to figure out some FPS issues for a while and tweaking things. Today, I had thought there was perhaps a difference between 1.5 and 1.6, though my pre-post-testing ruled that out, however, that same testing showed me that when I select the 2070 Super as the primary OpenGL GPU that the FPS issue is gone..?

Q2 is and old game; I doubt the OGL code has been updated much and, barring some sort of super inefficient code somewhere (possible) I can't see it maxing out a PCIe 3.0 x4 link for rendering. I have in fact played v1.4 on an older system which was limited to PCIe 1.0 x16 for the GPU and achieved better FPS there.

With the setting set to 'Auto' or set to the '3060' which is again the card connected to the display I am using, I can achieve about 80 FPS full screen @ 5120x1440 and about 240 FPS at 1920x1080.

If I set it to the 2070 Super, which should be comparable/better/worse in some aspects than the 2070S, then it maxxes out at 1000 FPS in even 5120x1440.. So, somehow it's faster to render on the primary GPU and send 1000 FPS worth of 5120x1440 screen buffer from one card to another than simply rendering it on the card (that should easily be able to do 1000 FPS) that the display is connected to..

Any thoughts, ideas? Possible driver issue maybe?

This also makes me wonder if perhaps a similar issue is affecting me for RTX as I feel like my Q2RTX performance is perhaps lower than it should be as well; I think the 2070S did better on my old system at PCIe1.0 x16 and slower CPU (Mac Pro 3,1, dual quad xeon 3.0ghz) than my new system maybe for RTX... I've tried all the multi GPU options in Q2RTX but not sure if any apply for generic multigpu (non-SLI/non-matched-cards-kinda-stuff)

I only get about 50-60 FPS for RTX in 1080p with basically all settings at minimum w/no scaling. I know I've seen higher.

Any way to force which GPU it uses for rendering? Maybe I can get a similar result as with OGL if so?

@res2k
Copy link
Contributor

res2k commented Jul 28, 2022

Hm, you have a weird setup...
But first some back-of-the-envelope calculation... for 5120x1440 at 1000 FPS, assuming 32bpp color depth, you'd need 27 GiB/s of transfer rate. However PCIe 3.0 x4 is provides at most 3.938 GB/s (number taken from https://en.wikipedia.org/wiki/PCI_Express) -
that doesn't fit your description of GPU0 rendering and transferring everything to GPU1.
So maybe something doesn't do what you think it does...

In general, there are many "moving parts" that could introduce a bottleneck (possibly anything from OS and graphics driver down do the MoBo or graphics hardware). Not sure your issue is something that can be fixed in Q2RTX.

Maybe some general advice if you want to try to further narrow down the cause:

  • Check/compare with other OpenGL apps (eg Yamagi Quake II, vanilla Q2PRO, or benchmarks like Unigine Heaven)
  • Look at some performance metrics (maybe the "GPU" tab in task manager already tells you something, tools like RTSS and GPU-Z may also be helpful)
  • Not sure if you want to mess around with your hardware, but maybe swapping your graphics cars to the other slot may tell you something

@ForbiddenEra
Copy link
Author

I didn't think having multiple GPUs and multiple monitors was weird; especially for a developer.

And yes, I'm aware of the math which is why I'm confused; you'd almost expect the framerate loss if it did have to transfer the frame buffer; perhaps something is inverting the actual used card - I'll dig deeper next time I jam q2 and double check and verify as well as test other apps.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants