This was originally posted on 2020-11-24 on iRacing subreddit
I was fortunate enough to upgrade my CPU to AMD Ryzen 9 5950x, and I figured to do a frametime and utilization comparison in iRacing using fpsVR.
This was originally posted on 2019-03-31 on iRacing subreddit
TLDR: turning off HyperThreading in BIOS bumped up FPS in VR from ~60 during starts to solid 90.
I have just recently started iRacing and this week I was racing quite a bit in the Lime Rock Park with MX-5. Unfortunately I used to get around 60 FPS during starts and that is a very tense part of the race. Every frame counts during those moments. Since I have a pretty decent system, I though I should be able to get more FPS in VR. My relevant PC specs are:
Intel i7-4790K, 32GB Ram, Radeon VII and Samsung Odyssey+
First I turned down most of the graphics settings in iRacing. But even if turning off cubemaps and shadows would bump the FPS quite a bit, it was still far from solid 90. Then I noticed that some parts of the track are fine, but others are still choppy. Turning off Event Objects, Crowds and Grandstands helped with that a bit, but still no cigar.
Then after quite a bit of searching I noticed old forum threads regarding the multi-thread usage of iRacing almost three years ago, where setting CPU affinity of iRacing to a single core would help quite a bit. But that was acknowledged as a bug and was fixed in iRacing since. I tried setting affinity anyways, but I would get "Access Denied" (even when running with Admin rights) from Task Manager.
So finally today I took the plunge and rebooted to BIOS (or more correctly to UEFI F/W) and turned off HyperThreading and voila, 90 FPS throughout all the race I just did.
Here are the render stats from the excellent fpsVR of the last session before turning off HT and then with the HT turned off:
Frametimes Larger Size
Utilization Larger Size
| HT on | HT off | |
|---|---|---|
| Delivered fps | 84.47 | 88.76 |
| GPU Frametimes median | 7.7 ms | 7.8 ms |
| GPU Frametimes 99th percentile | 10.7 ms | 9.5 ms |
| GPU frametime <11.1ms(vsync) | 99.3% | 99.7% |
| CPU frametime median | 6.4 ms | 6.2 ms |
| CPU frametime 99th percentile | 10.1 ms | 8.3 ms |
| CPU frametime 99.9th percentile | 12 ms | 10.2 ms |
| CPU frametime <11.1ms(vsync) | 99.7% | 99.9% |
| Max. SteamVR SS | 152% | 154% |
| Render resolution per eye by SteamVR settings | 1758x2195 | 1770x2209 |
| Render resolution per eye recommended | 1426x1780 | 1426x1780 |
The median frametimes did not drop that much, but the 99.9th percentile (1 out of 1000) of CPU frametimes crucially did not reach 11ms - the amount of time available to render a frame, if GPU is pushing 90FPS.
Hope this is useful and helps to someone else as well.
