A system is said to have passed the benchmark when the average frame rate during the run meets or exceeds the target frame rate for the test.
| Target frame rate | |
|---|---|
| HMD mode | 88.9 FPS |
| Desktop mode | 109 FPS |
The benchmark workload is the same whether it is run on an HMD or on the desktop. The target frame rate for desktop mode, however, is higher than the target frame rate for HMD mode. The difference is explained by VR SDK overhead and HMD refresh rate.
VR SDK overhead
To achieve 90 FPS, the system must create and present each frame within 11 ms. When running on an HMD, however, there is additional CPU overhead. The VR SDK reserves 1–3 ms of time for each frame, depending on the HMD and the state. This time is used for reading the sensors, motion prediction, techniques like asynchronous time warp and frame reprojection, and for image manipulation to allow for lens distortion.
When running on the desktop, there is no SDK overhead, which enables the system to achieve a higher frame rate.
HMD refresh rate
When running on an HMD, the system is always locked to the refresh rate of the headset, which is 90 Hz for both the HTC Vive and Oculus Rift. This means that the frame rate is capped to a maximum of 90 FPS. The concept is similar to vertical sync on a desktop monitor, though the implementation in the VR SDK is more complex. The result is that even when the system is capable of rendering an individual frame faster, it must wait for the VR SDK before moving on to the next frame.
There is no such restriction when running in desktop mode. As soon as the system is finished with one frame, it can move on to the next. These savings are then reflected in a higher average frame rate for the complete run.