Triple buffering and crossfire/sli

S

sadbuttrue

Gawd
Joined
Feb 16, 2008
Messages
519
I can't seem to find a concrete answer. Does this work with multiple cards using d3doverider in directx games? The anandtech article states categorically that it will not work, but then dozens of people on various forums seem to think it does. I guess it could simply be defaulting to vsync only and they assume it's working.

Personally I can tell the difference between vsync and triple buffering, even at 120hz. Anyone else out there with the same sensitivity and multiple cards able to shed light on this problem?
 
Hmm, can you think of an example where you'd need triple buffering? I personally hate Vsync, even triple-buffered, so I never use it, but I might be able to test if it I knew what I was looking for.
 
I can confirm that D3DOverider works with SLI. BUT, you must use SFR as it doesn't work with AFR/AFR2. Triple buffering is nice because it allows you to use VSync and lower input lag.
 
Vega said:
I can confirm that D3DOverider works with SLI. BUT, you must use SFR as it doesn't work with AFR/AFR2. Triple buffering is nice because it allows you to use VSync and lower input lag.
Click to expand...

Is this confirmed somehow? From my post in the input lag thread:

-Triple buffering reduces input lag with vsync per Anand:
http://www.anandtech.com/show/2794/4
But, nVidia says the opposite in their SLI user's guide:
http://forums.nvidia.com/index.php?showtopic=158727
"Triple buffering can reduce performance because of the increased memory consumption, and will offer no benefit if the frame output rate is higher than the refresh rate of the monitor. Additionally, triple buffering can introduce slightly higher input latencies because of prolonged storage times."

I always keep vsync on as I need it when I use 3DVision. I've tried triple buffering on and off with my SLI setup and can't say I notice a difference or any significant input lag.
 
Cheers for "kind of" confirmation :). Isn't SFR unsupported these days, being significantly slower than AFR?
 
DDuckMan said:
Is this confirmed somehow? From my post in the input lag thread:

-Triple buffering reduces input lag with vsync per Anand:
http://www.anandtech.com/show/2794/4
But, nVidia says the opposite in their SLI user's guide:
http://forums.nvidia.com/index.php?showtopic=158727
"Triple buffering can reduce performance because of the increased memory consumption, and will offer no benefit if the frame output rate is higher than the refresh rate of the monitor. Additionally, triple buffering can introduce slightly higher input latencies because of prolonged storage times."

I always keep vsync on as I need it when I use 3DVision. I've tried triple buffering on and off with my SLI setup and can't say I notice a difference or any significant input lag.
Click to expand...

That nVidia information is ancient. nVidia page says: Triple buffering forces the card to create a third buffer, in effect using an additional 50% of the available memory for frame storage." I don't see anywhere near 50% more memory required. More like 5%.

Anandtech does an entire article with detailed tests to show that VSync with triple buffering is almost as lag free as VSync off with double buffering. Compare that versus nVidia's one sentence: " Additionally, triple buffering can introduce slightly higher input latencies because of prolonged storage times", with no explanation.

I think I'll trust the Anantech article as they are usually spot on.


sadbuttrue said:
Cheers for "kind of" confirmation :). Isn't SFR unsupported these days, being significantly slower than AFR?
Click to expand...

Depends on the game. I use SFR in most of my games because a lot of the time it is just as fast as AFR/AFR2 and has the added benefit of allowing D3DOverider to work and enable triple buffering so you can turn on VSync without all the lag. Without VSync in portrait mode, I get tremendous screen tearing.
 
sadbuttrue said:
I can't seem to find a concrete answer. Does this work with multiple cards using d3doverider in directx games? The anandtech article states categorically that it will not work, but then dozens of people on various forums seem to think it does. I guess it could simply be defaulting to vsync only and they assume it's working.

Personally I can tell the difference between vsync and triple buffering, even at 120hz. Anyone else out there with the same sensitivity and multiple cards able to shed light on this problem?
Click to expand...


d3doverider works wit my games and my cards in crossfire.

Not sure on a 120hz monitor though, not sure if they behave different with those monitors.
 
Hi guys.

Go into l4d - the game that Anand mentions supporting triple buffering. Turn off vsync, move your mouse around. It will be nice and smooth.

Turn on triple buffering and do the same. Noticeable input lag.

I experimented with this when Anand first came out with the article. I have tried it with both SLI 285s and a 5870 with the same results.
 
The problem still seems unanswered tbh. Let me clarify once again - selecting triple buffering in d3doverider does not guarantee that it actually works in game. Heck, even on a single card it fails to work in some games.

digitaldurandal said:
Hi guys.

Go into l4d - the game that Anand mentions supporting triple buffering. Turn off vsync, move your mouse around. It will be nice and smooth.

Turn on triple buffering and do the same. Noticeable input lag.

I experimented with this when Anand first came out with the article. I have tried it with both SLI 285s and a 5870 with the same results.
Click to expand...

From my testing I'm almost certain triple buffering in L4D doesn't work. Try using d3doverider to force it and have L4D set to vsync disabled. Granted I run at 120hz with 200+fps, so triple buffering adds no discernible lag.
 
I know not everyone shares the same opinion, but in the vast majority of games I don't ever see any issues when not using VSync. Visual tearing seems mostly confined to a small handful of games. Given how problematic VSync can be [Dead Space anyone?] my policy is not to use it. I'd consider employing it if a game I played had considerable tearing, but of all the games I've played in the last few years I can't name a single example where that's been the case.
 
digitaldurandal said:
Hi guys.

Go into l4d - the game that Anand mentions supporting triple buffering. Turn off vsync, move your mouse around. It will be nice and smooth.

Turn on triple buffering and do the same. Noticeable input lag.

I experimented with this when Anand first came out with the article. I have tried it with both SLI 285s and a 5870 with the same results.
Click to expand...

Just to clarify, if you turn off VSync, its just using normal double buffering so it should be nice and smooth like you said. Then you say "turn on triple buffering and do the same". I hope you mean turn on VSync and triple buffering? Real triple buffering only works with Vsync ON. A lot of stupid game companies call frame queue flipping "triple buffering' which isn't triple buffering at all.

Remember, it is impossible to get VSync triple buffering to work in any other mode besides split frame rendering mode of SLI.

sadbuttrue said:
The problem still seems unanswered tbh. Let me clarify once again - selecting triple buffering in d3doverider does not guarantee that it actually works in game. Heck, even on a single card it fails to work in some games.

From my testing I'm almost certain triple buffering in L4D doesn't work. Try using d3doverider to force it and have L4D set to vsync disabled. Granted I run at 120hz with 200+fps, so triple buffering adds no discernible lag.
Click to expand...

I find that if I get the D3Doverider "beep" thats its actually detected the game and activated triple buffering, it works. I can notice a perceptible difference in input lag in the game. Regular input lag without triple buffering from Vsync on is just horrible and unplayable. Not only that, but I can see a visible rise in memory use of the Vram with triple buffering on versus off.
 
I forgot to mention the easiest way to see if triple buffering is working. Crank up your in-game settings to get a nominal frame per second reading of ~45 and watch your frames per second. If VSync is on without triple buffering, your FPS will go to half your screen refresh. For most people it will drop down to 30 FPS. With Triple buffering on, the FPS will rise to it's natural number of ~45 and will not be locked to 60, 30, 15 etc as you drop below 60 FPS.

This test only works if you can get your FPS below 60. If it's always above 60, VSync on with or without triple buffering will still produce 60 FPS.
 
If you have SLI and your framerate is always near or above the refreshrate, is there any harm in using triple buffering?
 
You must log in or register to reply here.
Back
Top