I did a great deal of Google searching to find some information on this subject and I came up empty handed. So in this topic I would like to reach out to the Oculus Rift community and discuss, with evidence preferably, the affect the Oculus Rift has on CPU performance for gaming.
In games like Arma 3 and GTA 4, CPU performance is a bottleneck long before the GPU is, in most high end systems today. I believe this is due to the massive amount of data the CPU has to compile and send to the GPU to render. This is evidenced by increasing draw distance in these games, and thus putting a significantly higher load on the CPU and watching GPU usage fall off dramatically.
My concern is, what happens when we suddenly have to render a second perspective? How does this scale against CPU performance? If a single frame on a 2D monitor is causing a CPU to bottleneck around say, 70 fps, what happens when we now introduce the Oculus Rift and must render an additional screen with slightly off perspective? Does it cut the frame-rate clear in half down to 35? Does the game engine somehow know to 'copy and paste' all the accumulated geometrical data and send it off to the GPU with a very slight hit?
If anyone out there with a Rift would be willing to do some testing, I think it would be a great contribution to the community and the world really, in determining what kind of performance users can expect to have on their computer when they pick up an Oculus Rift.
Testing for this is rather difficult as it requires a relatively controlled environment. This means no multiplayer games, and using the same scenario over and over again to get results. It also helps to narrow things down to specific games, games that are notorious for having CPU bottlenecks in particular as we are not concerned with the GPU usage in these tests. I will provide a simple setup using the two above mentioned games to show you how to test for CPU bottlenecks.
Required Software:
Third Party Stereoscopic 3D Driver to inject into game for Oculus Rift support
MSI Afterburner or EVGA PrecisionX
Arma 3 and/or GTA 4
Any other game that introduces a CPU bottleneck
Once you have all these items, the next step is to setup the GPU monitoring program (MSI/EVGA OSDs) to display GPU usage when in a game. This is done through RivaTuner Statistics Server which comes with both programs. Go into the Monitoring options for whichever program you chose, and make sure that GPU Usage and FPS are being monitored and that it is being displayed in the OSD. Test this by entering a 32-bit game and looking for the information in the top left of the screen.
Now that you have the diagnostics ready, it's time to setup the game scenario and create a CPU bottleneck. In either game, this is very simple to do. Here are the settings for each game to create a definitive CPU bottleneck no matter what GPU or CPU you currently own.
Arma 3:
Object Quality = Ultra
Terrain Quality = Ultra
Shadow Quality = Low
View Distance = 12,000m
Object Distance = 5,000m
SSAO = Disabled
Anti-aliasing = Disabled
GTA 4:
Shadow Quality = Low
View Distance = 100
Detail Distance = 100
Traffic Density = 100
With the above settings, you should be completely CPU bottlenecked no matter what setup you are running. To tell whether you are (and you definitely will be) what you should look for is in the OSD in the top left corner, GPU usage should be anything but 99%. Most likely it will be pathetically low, somewhere around 15-50% depending on how powerful your graphics card is. The faster it is, the lower the usage will be. But you should definitely be CPU bottlenecked and seeing your GPU working with a very light load.
Now that you have your CPU bottleneck, the next step is to create a scenario that you can return to with the Oculus Rift equipped and compare 2D performance to stereoscopic 3D. Log your 2D frame-rate numbers and GPU usage, quit the game, and reload it with the Oculus Rift enabled. Again, note the frame-rate and GPU usage.
Once you have some data, please post here with the results. The goal is to find out how much performance is lost when already in a CPU bottlenecked situation. This I believe will be the much larger issue when getting the 60 fps per eye target. You can nearly always scale GPU's in SLI/Crossfire to meet demand. You cannot increase CPU performance outside of overclocking and/or getting a faster chip, and we are already seeing CPU bottlenecking at the top end of the hardware spectrum in 2D mode..
If you already have some information to share on this topic, please provide it here as well. I am eager to see what the findings are. Thanks.