rFactor 2 DevMode used to gain some insights on VR performance
Methodology
These benchmarks where created with an early implementation of an automated rFactor 2 specific benchmark tool. I've not yet published the code for it, because it is very specific to work with fpsVR and attempts to get Nvidia's Fcat VR to run on my system failed so far.
The future plan is to integrate the rF2 Settings Widget capabilities to run automated benchmarks with a variety of content. The current app relies on the DevMode executable which has limited content aboard. Therefore, you will not find track and car specific benchmarks like Circuit and Vehicle Detail.
The goal is not to have some numbers saying this system is capable or not, but to have an insight how much relative CPU or GPU frametime one would lose when activating certain settings.
CPU Single Thread Limit
First and foremost it needs to be said that the update rate of this Sim is limited by its physics engine. If you don't throw enough single thread performance at it, it will perform somewhat unpredictable whenever physics ticks can not be delivered in time.
One can read quite a lot of comments that game X is only utilizing one or two CPU cores. It is not that simple. Even today, it is not an easy task to calculate across multiple threads efficiently, if you need to consider concurrent results from other worker threads.
Now throw 4 smart engineers at this problem, with smart I mean expensive, and you will have a feature that people will take for granted but will not pay a single extra Euro for. So it is not surprising to find this behaviour in most non-multi-million dollar realtime engine productions.
rFactor 2 is no exception to that so this information is only really useful if the machine it is running on is beyond that single thread bottleneck.
Optimise CPU usage
In any case one can free the CPU from load by reducing the amount of draw calls and geometry. Thus setting Circuit, Vehicle and Opponent Details to the lowest possible setting.
First Benchmark Results
System
- i7-5820k @ 4,2GHz, 32GB DDR4-2888
- GeForce GTX 1080 Ti
- 1TB Samsung 860 EVO SSD (serving only the game not the OS)
Rain-Drops
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 13.870241 8.234471 NaN NaN
Low 14.440307 8.753541 4.109991 6.303620
Ultra 16.333206 8.787850 13.108447 0.391947
Rearview-Particles
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 13.802886 8.524134 NaN NaN
Low 14.258536 8.647248 3.301119 1.444302
Road-Reflection
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 13.823519 7.812769 NaN NaN
Low 13.870241 8.234471 0.337987 5.397606
Ultra 14.708351 8.574477 6.042506 4.129054
Environment-Reflection
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 13.968590 8.414098 NaN NaN
Low 13.870241 8.234471 -0.704076 -2.134825
High 14.341655 8.601035 3.398747 4.451571
Soft-Particles
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 13.870241 8.234471 NaN NaN
High 13.950171 8.446366 0.576269 2.57326
Special-Effects
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 14.204147 8.522965 NaN NaN
Low 14.230783 8.482059 0.187522 -0.479947
Ultra 14.330645 8.707881 0.701732 2.662349
# i9-10850k + RTX3090, Results are somewhat sluggish
Post-Effects
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 2.690852 4.291171 NaN NaN
Low 2.586419 4.095671 -3.881025 -4.555859
Medium 2.998736 4.559566 15.941603 11.326478
High 3.403225 4.346985 13.488658 -4.662305
Ultra 5.619619 4.366498 65.126272 0.448883
Shadow-Blur
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 2.756997 5.123894 NaN NaN
Low 2.795348 4.575426 1.391046 -10.704114
Medium 2.807449 4.579574 0.432884 0.090649
High 3.042394 5.216947 8.368650 13.917747
Shadows
GPU Frametime CPU Frametime GPU Percent CPU Percent
Off 2.724962 4.546595 NaN NaN
Low 2.687672 4.593158 -1.368449 1.024136
Medium 2.792317 4.598825 3.893516 0.123370
High 2.798490 4.700067 0.221047 2.201479
Ultra 3.052708 4.686365 9.084123 -0.291515
Process finished with exit code 0
Conclusion
It is quite interesting that you can gain 21% of GPU time alone by leaving Reflection settings at Low and Rain Drops set to Off. Other big GPU hits will be Multi Sample Anti Aliasing, super sampling factor, the amount of visible vehicles when headlights are active and Post Processing. These will have to wait for a later date.
I will try to update this page whenever I find the time to further develop the benchmark tool