Tempted by some cheap GTX460 SLI action? Might want to check this out first before laying down your

Duff-Man · 9 Sep 2010 at 00:16

This is microstutter. AKA uneven frame output.

I've been trying to raise awareness of this issue for years, in the vain hope that it nvidia and ATI might start taking it seriously. The fact of the matter is that all multi-GPU setups suffer from this - the ATI setups are just as bad, if not worse, than the nvidia ones. It's a property of alternate frame rendering (AFR), and uneven frame output.

I wrote a small program to test the amount of microstutter in a game.

If you want more info on microstutter, the effect it has on games and apparent framerate, and how to minimise it etc, check out my thread on XS forums.

Randomface74 · 9 Sep 2010 at 00:17

Video proves nothing.

NZXT30 · 9 Sep 2010 at 00:21

i'm surprised that many people dont know that micro stuttering at 30 or below fps exists in every sli/crossfire setup. there is proof all over the internet about that with complete explanation.

An Exception · 9 Sep 2010 at 00:23

NZXT30 said:
I cant play any game with fps below 60 let alone below 30. if I was fine with sub 30 fps I would be a console gamer.

Who's talking about gaming below 30fps?

An Exception · 9 Sep 2010 at 00:25

billysielu said:
Video proves nothing.

If your not convinced about micro-stutter then you should be able to find enough evidence to convince your self otherwise should you do a little research.

An Exception · 9 Sep 2010 at 00:26

NZXT30 said:
i'm surprised that many people dont know that micro stuttering at 30 or even above below fps exists in every sli/crossfire setup. there is proof all over the internet about that with complete explanation.

Fixed. Micro-stutter that apparent doesn't just stop when you get to 31fps.

Duff-Man · 9 Sep 2010 at 00:27

NZXT30 said:
i'm surprised that many people dont know that micro stuttering at 30 or below fps exists in every sli/crossfire setup. there is proof all over the internet about that with complete explanation.

It exists at any framerate. It's only "noticable", in that it presents itself as a visible stuttering, at under ~30fps.

At higher framerates, the effect is to reduce the apparent framerate. This is because the human eye catches the maximum gap between frames, rather than the total number of frames output (as the FPS counter measures). A mild microstutter may make 60fps look like 50-55fps. A heavy microstutter would make it look like ~40fps. The theoretical worst-case scenario, for a 2-GPU setup, is for the effective framerate to be halved (i.e two frames are output simultaneously, with a double gap in between).

Bottom line: To get the same smoothness of gameplay, you need a HIGHER framerate with multi-GPU setups. The microstutter will "eat away" 10 to 35% of your performance, based on my analysis of a variety of configurations.

But, the silver lining: When you are CPU limited, the microstutter disappears entirely. The GPU output syncs to the highly-regular CPU output, since the GPUs are waiting for the CPU to finish its workload at every frame.

NZXT30 · 9 Sep 2010 at 00:29

Ejizz said:
Fixed. Micro-stutter that apparent doesn't just stop when you get to 31fps.

it does because at 30fps with micro stuttering dips you actually get less than 30 fps and everybody knows that when fps dip below 30 the difference is huge compared to for example 35 vs 31 fps. not hard to understand. at 60+ fps micro stuttering completely disappears from your eyes unless you own a 120 hz monitor.

An Exception · 9 Sep 2010 at 00:33

NZXT30 said:
it does because at 30fps with micro stuttering dips you actually get less than 30 fps and everybody knows that when fps dip below 30 the difference is huge compared to for example 35 vs 31 fps. not hard to understand. at 60+ fps micro stuttering completely disappears from your eyes unless you own a 120 hz monitor.

No, micro stutter is still noticeable over 30 fps and maybe even slightly at 60 fps as the micro stutter may give the apparent speed of 30fps while it's stuttering which would be a noticeable difference say 60fps to 30fps.
Of course the micro stutter becomes less apparent as you increase the FPS but it doesn't 'actually' disappear although it may become seriously unnoticeable with super-high FPS, but why then use a second GPU?.

NZXT30 · 9 Sep 2010 at 00:34

Ejizz said:
No, micro stutter is still noticeable overs 30 fps.

yes but hardly noticeable and disappears at 60+ fps. however thats not the point because with sli you get 60-80%+ more performance so for example 50fps with micro stuttering looks much smoother compared to 35 fps without micro stuttering. the video is flawed because the sli setup is running at higher resolution so there is no fps gain.

Duff-Man · 9 Sep 2010 at 00:34

NZXT30 said:
it does because at 30fps with micro stuttering dips you actually get less than 30 fps and everybody knows that when fps dip below 30 the difference is huge compared to for example 35 vs 31 fps. not hard to understand. at 60+ fps micro stuttering completely disappears from your eyes unless you own a 120 hz monitor.

Not really...

Read my post above.

NZXT30 · 9 Sep 2010 at 00:39

Duff-Man said:
Not really...

Read my post above.

I read it now take the effort to explain if you want.

Duff-Man · 9 Sep 2010 at 00:45

NZXT30 said:
I read it now take the effort to explain if you want.

I don't know how much more simply I can put it :confused:

Microstutter always acts to reduce the apparent framerate.

At framerates where you can catch the output of individual frames, the effect is observable as a a genuine "stutter" - you can see the difference in framerate regularity.

At higher framerates it simply reduces the smoothness of the scene. You can quite easily quantify the apparent framerate at any moment in time, based on the maximum gap between output frames. This gap will necessarily be larger than indicated by the framerate, and so the apparent framerate is necessarily less than indicated by the FPS counter.

If that was too complex for you then I'm not sure what else can be said. Perhaps someone else could translate?

NZXT30 · 9 Sep 2010 at 00:55

Duff-Man said:
I don't know how much more simply I can put it

Microstutter always acts to reduce the apparent framerate.

At framerates where you can catch the output of individual frames, the effect is observable as a a genuine "stutter" - you can see the difference in framerate regularity.

At higher framerates it simply reduces the smoothness of the scene. You can quite easily quantify the apparent framerate at any moment in time, based on the maximum gap between output frames. This gap will necessarily be larger than indicated by the framerate, and so the apparent framerate is necessarily less than indicated by the FPS counter.

If that was too complex for you then I'm not sure what else can be said. Perhaps someone else could translate?

if you get for example 70 fps with micro stutter dips at 64fps you dont notice the difference as your monitor can only display up to 60fps.

An Exception · 9 Sep 2010 at 00:55

Duff-Man said:
I don't know how much more simply I can put it

Microstutter always acts to reduce the apparent framerate.

At framerates where you can catch the output of individual frames, the effect is observable as a a genuine "stutter" - you can see the difference in framerate regularity.

At higher framerates it simply reduces the smoothness of the scene. You can quite easily quantify the apparent framerate at any moment in time, based on the maximum gap between output frames. This gap will necessarily be larger than indicated by the framerate, and so the apparent framerate is necessarily less than indicated by the FPS counter.

If that was too complex for you then I'm not sure what else can be said. Perhaps someone else could translate?

I think Captian Slow probably get's it now...

One thing I don't get, why have you got 480's when you know soo much about micro stutter?

Final8y · 9 Sep 2010 at 00:56

Duff-Man said:
This is microstutter. AKA uneven frame output.

I've been trying to raise awareness of this issue for years, in the vain hope that it nvidia and ATI might start taking it seriously. The fact of the matter is that all multi-GPU setups suffer from this - the ATI setups are just as bad, if not worse, than the nvidia ones. It's a property of alternate frame rendering (AFR), and uneven frame output.

I wrote a small program to test the amount of microstutter in a game.

If you want more info on microstutter, the effect it has on games and apparent framerate, and how to minimise it etc, check out my thread on XS forumshttp://www.xtremesystems.org/forums/showthread.php?t=258433.

Damn good thread to on that forum.

Rofflay · 9 Sep 2010 at 00:56

i can see a little micro stutter on the single 5870 or is that just becuase its only 30fps? either way neither are completely smooth to my eyes and oh yes i see every little stutter/hitch.

sli looks horrible in that video tho!

NZXT30 · 9 Sep 2010 at 00:58

Rofflay said:
i can see a little micro stutter on the single 5870 or is that just becuase its only 30fps? either way neither are completely smooth to my eyes and oh yes i see every little stutter/hitch.

sli looks horrible in that video tho!

yeah thats not micro stutter thats 30fps which is just not smooth enough.

An Exception · 9 Sep 2010 at 00:58

NZXT30 said:
if you get for example 70 fps with micro stutter dips at 64fps you dont notice the difference as your monitor can only display up to 60fps.

That would be the case yes. But if the apparent FPS dropped to 45 which it could do, then you will suffer from some perceptible micro stutter but not quite as bad as witness in the video.

neo-omega · 9 Sep 2010 at 00:58

I'm surprised no-one picked up on him using a triplehead2go rather than the native eyefinity of the 5870. Surely if he wanted an accurate reprisentation he'd have used eyefinity rather than 3h2g, unless it was for control purposes when gonig from ATI to Nvid or something?