Welcome Stranger to OCC!Login | Register

Serious Statistics Pt. 2: The Sync Encounter

» Discuss this article (2)


I am not completely sure if a conclusion is really necessary for this article, but why not? As is perhaps not surprising, playing in a Borderless window under DirectX 11 results in the highest frame latency, because of omnipresent Desktop Window Manager's double buffering. When playing in Fullscreen or under Vulkan, as far as PresentMon/OCAT could tell only V-Sync of the in-game options actually introduced any latency. Of the driver-side options, it was again only the basic v-sync option to add latency, according to PresentMon/OCAT, with the Fast Sync, Enhanced Sync, and Frame Rate Target Control all appearing to send frames out just as quickly as they were available to send.

While we were getting flat lines at 0 ms for a lot of those tests, technically that should not be the case for a lot of them, as the processes within the driver or on the hardware itself will take some amount of time. This seems to be beyond PresentMon/OCAT's ability to capture, and is also beyond mine, so this is the best you are going to get. Still, some of these results are somewhat interesting I think. If nothing else, it gave me a good reason to dig in and make more improvements to the scripts I use for generating these graphs and processing the data.

If you are concerned about frame time, play in fullscreen and avoid the DWM. If you are concerned about screen tearing, either use triple buffering, a driver option (but not the basic v-sync option), or go borderless and accept DWM, and that I think would be the order I would recommend them in. However, as I long ago developed a concern of games crashing in fullscreen taking the computer with them, I will likely continue to play in a borderless window. I will just be sure to not use an in-game v-sync option, unless I know for sure the DWM double-buffering is disabled.

Related Products
Random Pic
© 2001-2018 Overclockers Club ® Privacy Policy
Elapsed: 0.0930809975   (xlweb1)