Welcome Stranger to OCC!Login | Register

Serious Statistics Pt. 2: The Sync Encounter

   -   
» Discuss this article (2)

Fullscreen – V-sync and Triple Buffering:

The only reason I am looking at this combination of options is because I made the error those months ago in thinking that both options should be checked off. I made the incorrect assumption that triple buffering would not be a valid option without v-sync being enabled, as it is in fact a kind of v-sync and in other games is an option under v-sync. Now I know better. (Of course the data does not directly bare out that triple buffering is working, but my observations do.)

As a quick recap, I believe the v-sync option engages double buffering, while triple buffering is triple buffering. Combining these two technologies does appear to do something a little funky though.

 

 

 

 

 

GTX 1080 - DX11

Vega 64 - DX11

 

Looking at these we see similar behavior to when we only had v-sync on, with the frame times all collected around 16.667 ms. The display time graphs are much the same.

GTX 1080 - DX11

Vega 64 - DX11

 

It is the frame latency graphs where the funkiness comes in. When just using v-sync, the latency varied up and down, but was always under 16.667 ms. With triple buffering also activated, the latency is a solid line along 16.667 ms, so we are seeing the latency increase here, according to the data.

GTX 1080 - DX11

Vega 64 - DX11

 

It is possible there is an issue with triple buffering and PresentMon causing both this flat line and the flat line we saw with only triple buffering, but I think it is at least reasonable to believe there is increased latency here. If we take a quick look at another graph I can make that plots more of these data on the same graph, we can see just how high the MsUntilDisplayed and MsUntilRenderComplete values are. It is higher than the points just v-sync produces.

GTX 1080 - DX11

Vega 64 - DX11

 

Clearly it looks like combining v-sync and triple buffering is not what we want to do in Serious Sam Fusion 2017, but now we can move on to consider the frame rate limiter.




Related Products
Random Pic
© 2001-2018 Overclockers Club ® Privacy Policy
Elapsed: 0.2835187912   (xlweb1)