Welcome Stranger to OCC!Login | Register

Serious Statistics Pt. 2: The Sync Encounter

   -   

Fullscreen – Triple Buffering:

Unfortunately triple buffering is not available under Vulkan, so we have less data here, but it still serves its purpose. Also I find it odd triple buffering is its own option, separate from v-sync, and not identified as a kind of v-sync either.

As a quick reminder, triple buffering works by having two back buffers the engine can write to. This way one can hold a completed frame that can be flipped into the front buffer at any time, while the other back buffer can have a new frame written to it. This allows the engine to render as quickly as it wishes while the front buffer remains protected from screen tearing. We can see this in the frame time graphs, as they match what we saw without any v-sync:

 

 

 

 

 

GTX 1080 - DX11

Vega 64 - DX11

 

The display time graphs also look like they did without any v-sync, which is not exactly what we expected, since the front buffer should only be scanned out once every 16.667 ms. Also we are not seeing any frames being dropped, which would appear as a 0 ms for the display time. My guess is the way Serious Sam Fusion 2017 is handling the flipping between the back buffers and to the front buffer (which is when a frame actually starts to be displayed) is confusing PresentMon because there must be frames dropping and frames are only being scanned out every 16.667 ms. I did not observe any screen tearing while running these tests.

GTX 1080 - DX11

Vega 64 - DX11

 

Since the display times are not collected at the one-frame-later mark, here are the frequency plots:

GTX 1080 - DX11

Vega 64 - DX11

 

Finally the latency graphs, which show a flat line at zero:

GTX 1080 - DX11

Vega 64 - DX11

 

Technically there should be some latency here, but it would be minimal because of how quickly these frames are being rendered. The latency for one frame cannot be much longer than the frame time of the next frame, because of how triple buffering works. (After the next frame is rendered, that 'one' frame is going to be overwritten by the new next frame.)

Next is what happens when we have both v-sync and triple buffering enabled.



  1. Serious Statistics Pt.2: The Sync Encounter - Introduction
  2. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - No V-Sync
  3. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - V-Sync
  4. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - Triple Buffering
  5. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - V-Sync & Triple Buffering
  6. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - Frame Rate Limiter
  7. Serious Statistics Pt.2: The Sync Encounter - Borderless - No V-Sync
  8. Serious Statistics Pt.2: The Sync Encounter - Borderless - V-Sync
  9. Serious Statistics Pt.2: The Sync Encounter - Borderless - Triple Buffering
  10. Serious Statistics Pt.2: The Sync Encounter - Borderless - V-Sync & Triple Buffering
  11. Serious Statistics Pt.2: The Sync Encounter - Borderless - Frame Rate Limiter
  12. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - Driver V-Sync
  13. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - Fast & Enhanced Sync
  14. Serious Statistics Pt.2: The Sync Encounter - Fullscreen - Frame Rate Target Control
  15. Serious Statistics Pt.2: The Sync Encounter - Conclusion
Related Products
Random Pic
© 2001-2018 Overclockers Club ® Privacy Policy
Elapsed: 0.1442129612   (xlweb1)