how can I prove triple buffering isnt being enabled in DX games?

[taltamir]

then maybe it is him who is misinterpreting the FRAPS results...
Would you mind looking at the fraps data dump and see the time to render frames and see if they are exactly 16.667ms (60 FPS) or 33.33ms (30 FPS) as I predicted and not values between the two? (values greater than 33.33 are possible and consistent with theory as well, those would be the 20FPS etc I mentioned)

 

[toyota]

where is this info?

 

[cusideabelincoln]

Or he could have triple buffering enforced via drivers or external utility. I've seen V-sync with no triple buffering break my performance in Chaos Theory. It would jump from 60 to 30. With triple buffering it never got below 48 or so.

 

[taltamir]

with latest version 3.2.7

in the 99 FPS tab you will see "Benchmark Settings" with the options of:
FPS
Frametimes
MinMaxAvg

Checkmark the Frametimes, when you get a benchmark it will also create a file with frametimes in the name.
Column 1 is frame number, Column 2 is amount of ms since the start of the test.
To get time to render each frame go to cell C3 and input:

=B3-B2

Then drag it down across column 3 all the way. This will give you the time to render each frame on column 3.

I apologize for the simplistic explanation, I know you are computer savvy enough to not need it, but I figured that other people are also reading.

Or he could have triple buffering enforced via drivers or external utility. I've seen V-sync with no triple buffering break my performance in Chaos Theory. It would jump from 60 to 30. With triple buffering it never got below 48 or so.

we will know once he examines the numbers.

 

[busydude]

where is this info?

You can find 3 files generated by FRAPS; fps, minmaxave and frametimes.

FPS -> shows the FPS(Calculated every second, obviously.)
frametimes -> Time(ms) taken to render a particular frame.

 

[toyota]

well I had to go back and re run two benchmarks since all I originally checked was min/max/avg.

taltamir, I see the frametimes but I have no idea what you are telling me to do.

 

[BFG10K]

There’s nothing wrong with Fraps except when trying to take a screenshot or video with the framerate overlay, and then it screws up. With the regular overlay it accurately displays what I’m seeing.

Here are photos of a similar section of the game showing Fraps FPS on a GTX480 with vsync forced from nVidia’s control panel. This is exactly what I see in-game:

http://img692.imageshack.us/i/20576169.jpg/
http://img703.imageshack.us/i/29879222.jpg/

I’ve tested the entire plateau section up until you slide towards Johns’ cabin, and all areas will settle on 60 FPS or 30 FPS no matter where you look, if you stand still long enough.

So Toyota’s system is either not running vsync, or triple buffering is being forced externally.

 

[taltamir]

I am saying that the frametimes listed are "time since test started" not "time since last frame"
I am saying to calculate time per frame rendered by subtracting one from the other.

 

[busydude]

taltamir, I see the frametimes but I have no idea what you are telling me to do.

Taltamir is asking you to calculate the delta time between successive frames.

 

[toyota]

There’s nothing wrong with Fraps except when trying to take a screenshot or video with the framerate overlay, and then it screws up. With the regular overlay it accurately displays what I’m seeing.

Here are photos of a similar section of the game showing Fraps FPS on a GTX480 with vsync forced from nVidia’s control panel. This is exactly what I see in-game:

http://img692.imageshack.us/i/20576169.jpg/
http://img703.imageshack.us/i/29879222.jpg/

I’ve tested the entire plateau section up until you slide towards Johns’ cabin, and all areas will settle on 60 FPS or 30 FPS no matter where you look, if you stand still long enough.

So Toyota’s system is either not running vsync, or triple buffering is being forced externally.

it was running vsync and I looked directly at the ground in that video so you could see it not go over 60fps and use that for an excuse. I also have NEVER forced triple buffering in any game.

 

[toyota]

I am saying that the frametimes listed are "time since test started" not "time since last frame"
I am saying to calculate time per frame rendered by subtracting one from the other.

I am too tired plus I am on my netbook now, lol. so here are the files if you want to do the math: http://www.megaupload.com/?d=YLS83ICS

 

[taltamir]

thanks, I processed them and here are the results:

https://spreadsheets.google.com/ccc?key=0ApcEvOA-L4zcdDJpZV9JNTgwc2V2UHZqZU9wRnI1M1E&hl=en
https://spreadsheets.google.com/ccc?key=0ApcEvOA-L4zcdEdldzhmbFdVbi1yam1FTjVjUW5UcWc&hl=en

You can see clearly on your vysnc off example the FPS hovers near 48... while with the vsync on, triple buffer off example the FPS jumps from ~60 to ~30 multiple times per second. (except for a few outliers). This matches theory, which says that should be the case unless triple buffering is turned on (in which case it should again hover around 48).

I am surprised at how far it can deviate though. occasionally you see a frame do 45FPS or 110FPS or something weird. And most frames deviate by up to +/-3 give or take from their supposed FPS.
I performed some tests on my computer and the same can be seen there. I don't know if its due to the driver implementation or FRAPS itself interfering / having measurement errors.
If it is the way vsync works then it seems to be loosely implemented in a "good enough" fashion.

 

[Voo]

I think it'll be much easier if anyone just understands how Vsync, double and triple buffering work and for me, this here explained it quite well..

 

[toyota]

I think it'll be much easier if anyone just understands how Vsync, double and triple buffering work and for me, this here explained it quite well..

well that was the anandtech article I was referring to in the op.

 

[BFG10K]

well that was the anandtech article I was referring to in the op.

Then why didn't you read it?

This can hurt performance even if the game doesn't run at 60 frames per second as there will still be artificial delays added to effect synchronization. Performance can be cut nearly in half cases where every frame takes just a little longer than 16.67 ms (1/60th of a second). In such a case, frame rate would drop to 30 FPS despite the fact that the game should run at just under 60 FPS.

Hell, an analysis of your own numbers shows I'm right because the vsync frametimes show they jump between 30 FPS and 60 FPS. If I had a video camera I'd upload a video of showing the exact same thing in-game with the Fraps counter.

But you still won't admit you're wrong, huh?

 

[toyota]

Then why didn't you read it?


Hell, an analysis of your own numbers shows I'm right because the vsync frametimes show they jump between 30 FPS and 60 FPS. If I had a video camera I'd upload a video of showing the exact same thing in-game with the Fraps counter.

But you still won't admit you're wrong, huh?

I am not arguing about how its technically supposed to work. you still have not gotten through your thick head that I am saying I have NEVER seen it do that in an actual game. you even gave me a specific game to try out and I showed you that it did not happen. please give me some more examples of games and I will show you that the framerate doesn't go straight to 30fps in those either.

 

[BFG10K]

you still have not gotten through your thick head that I am saying I have NEVER seen it do that in an actual game.

Uh, yes you have. Your Fraps runs were from an actual game, were they not?

Do you not understand the significance of the frametime calculations and what they show?

As for your video, I don’t think you were actually using vsync. Heck, your own Fraps benchmark with vsync proves the video isn’t accurate.

 

[toyota]

Uh, yes you have. Your Fraps runs were from an actual game, were they not?

Do you not understand the significance of the frametime calculations and what they show?

As for your video, I don’t think you were actually using vsync. Heck, your own Fraps benchmark with vsync proves the video isn’t accurate.

I am NOT denying or arguing about how it is supposed to work. if those files show it doing what it is supposed to then great. please get through your head that I am saying I have never seen the framerates drop straight to 30 fps in an actual game. you seem to use fraps as a tool when making your claims and so do I. using fraps you claimed in CoJ that it went straight from 60 to 30fps which it did NOT when I tested it. for the second time you have now said that I must not have had vsync on in that video. well I am uploading a video right now and would love to see you say that after you watch it.

 

[BFG10K]

Here are my Fraps FPS results with an in-game benchmark from a later game level in CoJ:

VSYNC
On Off
61 97
60 90
57 101
60 96
58 83
60 83
60 90
60 115
60 109
60 121
60 115
58 95
60 78
60 70
49 59
31 51
30 52
30 53
29 50
31 50
29 50
30 48
30 48
30 48
30 47
30 47
29 39
30 39
30 43
47 61
30 45
29 36
30 39
30 40
30 42
30 41
30 41
30 37
30 39
30 45
51 70
31 46
30 47
30 43
48 66
33 59
38 59
38 53
33 61
43 54
30 44
30 49
29 47
31 42
29 38
30 37
30 38
30 37
30 36
30 38

You can see the left column's results (vsync on) are overwhelmingly ~30 FPS or ~60 FPS, and those that aren't are just Fraps noise since we know it's not 100% accurate. Also in every case where the left column is ~30 FPS, the right column has an in-between value (i.e. between 30 FPS and 60 FPS).

Again, these are in-game results and exactly mirror what I see in every game level I've played in CoJ. So either Toyota's video wasn't using vsync, or he has a magic version of vsync that is quite special.

I am NOT denying or arguing about how it is supposed to work. if those files show it doing what it is supposed to then great.

You're flat-out trolling now, so stop it. Your own benchmark results show the framerate alternates between 30 FPS and 60 FPS with vsync, but you're still claiming it doesn't happen in game. If it's not happening in-game then what is Fraps measuring?

To put it perfectly bluntly, I don't believe you had vsync enabled in your video. More than likely you're genuinely confused, like you were when you mixed up the labels on the two benchmark results.

I also think you're genuinely confused as to what I mean when I say "it's either 30 FPS or 60 FPS". You also don't appear to understand what benchmarking noise is, nor do you understand how it relates to averages.

 

[toyota]

so for the THIRD time you are accusing me of having vsync off in that video? and you have the nerve to call me a troll? I am really sick of your little passive aggressive comments. I have said from the beginning that it does not show the framerate dropping straight to 30fps when watching the framerate in a game. you even said it will go to 30 fps on the screen after a few seconds and again that does NOT happen.


here is a video proving I had vsync on.

http://www.youtube.com/watch?v=T44vUa3vOsI

 

[BFG10K]

here is video proving I had vsync on.

Then explain to us how I can replicate the vsync phenomenon without fail in the same game, but you can't?

Explain to us how my Fraps results back what I’m seeing in-game, but yours don’t?

In-game with vsync Fraps either shows 60 FPS or 30 FPS in every level I’ve tested, excluding the in-between noise of course.

The only conclusion I can reach is that your system is not applying vsync properly when you enable it from nVidia’s control panel.

 

[toyota]

Then explain to us how I can replicate the vsync phenomenon without fail in the same game, but you can't?

Explain to us how my Fraps results back what I’m seeing in-game, but yours don’t?

In-game with vsync Fraps either shows 60 FPS or 30 FPS in every level I’ve tested, excluding the in-between noise of course.

The only conclusion I can reach is that your system is not applying vsync properly when you enable it from nVidia’s control panel.

I don't know how to explain it but vsync is clearly being applied. I am not playing around here and I am simply showing you what I have always witnessed when visually monitoring the framerate in a game.

you keep acting like I want to argue with you about how it works when again all I am telling you is what I have always seen in actual games is different.

 

[BFG10K]

http://forums.steampowered.com/forums/showthread.php?t=1286443

http://www.mmo-champion.com/threads/824589-Either-30-FPS-or-60-FPS.

http://forums.anandtech.com/showthread.php?t=1578076

http://forums.steamgames.com/forums/showthread.php?t=1368931

http://www.gamedev.net/topic/526384-frame-rate-and-vsync/

http://community.codemasters.com/forum/dirt-2-game-pc-1313/399480-wtf-vsync-30fps-cap.html

http://www.nvnews.net/vbulletin/showthread.php?t=41192

 

[toyota]

you can post all the links you want but I am telling you what I have seen for 8 years of gaming if I watched the framerate.

 

[BFG10K]

you can post all the links you want but I am telling you what I have seen for 8 years of gaming while watching the framerate.

Then your system has never been applying vsync properly, or you've been running games that have triple buffering.