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

[An Exception]

Which is what makes me laugh. At no point has ejizz considered that the ram is just to small for that resolution on the 460s. Also ati cards tends to be more efficient with their memory, so once again not a like for like comparison.

no one doubts microstutter exists, but its unrealistic. I do however understand the intention of ejizz which i hope is good, just not understanding the biblical obcession with it.

Biblical obsession? I'l have to bite my tongue on that one, as my patience wears a little thin...

As for the 1gb 460's being Vram limited I seriously, seriously seriously doubt that for various reasons, but like I'v said I'm worn down and have lost patience.

Oh well, ya can't feed Africa

 

[Rroff]

Why do people keep posting videos of a game stuttering? There is no way you can see micro stuttering in a video, and people posting the videos obviously don't really know what it is.

A ~25fps video confuses things even more... but microstutter basically is down to the dispersion of rendering updates within a set period of time. A consistant interval between rendering frames makes the rendering very smooth and close to or on the given framerate for the last second. A number of frames with a bigger gap between them can cause the "actual" fps as seen/felt by the player to be lower than that shown over the last second. So its perfectly possible for it to show in a video (tho a low framerate video makes it much harder) but you have to know what your looking for. Most of those videos are showing 1-2x 100-200ms pauses every 2-3 seconds which isn't microstutter.

Problem is all rendering systems single or multi currently have microstutter to a small degree... the problem is showing when it actually has a negative impact on the player's experience - a multi GPU solution might have say 4x more microstutter at the same framerate as a single GPU but be completely un-noticeable.


If SLI really stuttered as badly as those videos show compared to a single card no one would use it.

 

[kitch9]

Yeh...

I'm sorry dude, but from reading your posts I can only say you have no clue what so ever.

Micro stutter is subtle, very subtle to the point you almost have to train your brain to see it, once you see it you can't unsee it when it happens. There is no way it will show on a 30 fps video, I'm sorry but there isn't. I've seen it a couple of times on my pair of 460's, but it was only when I had v sync off, which I never tend to do.

The videos show normal stutter, which could be attributed to anything such as hdd.accessing, driver issues, lack of RAM etc as the latencies involved can cause the frames to stall like that.

 

[Rroff]

There is no way it will show on a 30 fps video, I'm sorry but there isn't.

While its much harder to detect it and especially doesn't work too well if your using a 25fps videocam to capture a game running at 60fps with a small to medium level of microstutter. Its not impossible for it to show up in a videocapture - remember that microstutter can quite feasibly reduce a shown 60fps to a percieved 30-40fps which would show up quite noticeably on a videocapture if you know what your looking for. Unfortunatly due to the low framerate of the videocam and the lack of synchronisation with game frame updates its mostly useless for showing microstutter with any real validity.

 

[kitch9]

While its much harder to detect it and especially doesn't work too well if your using a 25fps videocam to capture a game running at 60fps with a small to medium level of microstutter. Its not impossible for it to show up in a videocapture - remember that microstutter can quite feasibly reduce a shown 60fps to a percieved 30-40fps which would show up quite noticeably on a videocapture if you know what your looking for. Unfortunatly due to the low framerate of the videocam and the lack of synchronisation with game frame updates its mostly useless for showing microstutter with any real validity.

I've come from a single 5870 to 2 460's and I had a crossfire setup before that, and I can say there are a few in this thread talking out of their ass.

You are not one of them.

Micro stutter exists, only in certain scenarios ie not very often, its very subtle and its nothing like the videos.

If you use v sync I would go as far as to say its a non issue.

 

[Rroff]

Having used multiGPU setups since the 6800 (which was a bit naff) having owned a 7950GX2, 8800GT SLI, GTX260 SLI myself I'd agree that if it has any affect at all its very subtle. I never really noticed it at all and I've had plenty of potential to compare it equally i.e. 7950GX2 v single 8800GT, 8800GT SLI v single GTX260 and now GTX260 SLI v my new GTX470.

Only times I've really seen it was on early ATI 3800 series crossfire before ATI fixed the drivers and some 4870X2/9800GX2 setups but thats due to hardware issues on those 2 cards under certain configurations rather than the common cause of microstutter as such.

 

[kitch9]

Yeah, we are not talking about dropped or stalled frames which is what is shown in the videos. We are talking about minor variations in render time from one frame to the next. One frame may take 8ms to render and the next 16ms. This a subtle and barely perceptible. Enable V sync and the variation stops simple as that, because the monitor becomes the controlling factor when it comes to latencies from one frame to the next.

 

[Scougar]

Lost patience? You outright insulted people in this thread and you mention patience? lol.

Come on ejizz, you have to consider that lack of vram at those resolutions could possibly cause issues .

 

[tdream]

Yeah, we are not talking about dropped or stalled frames which is what is shown in the videos. We are talking about minor variations in render time from one frame to the next. One frame may take 8ms to render and the next 16ms. This a subtle and barely perceptible. Enable V sync and the variation stops simple as that, because the monitor becomes the controlling factor when it comes to latencies from one frame to the next.

Wrong

 

[Final8y]

Wrong

Explain then ?

 

[Scougar]

I dont disagree with ejizz, microstutter defo exists going on proff supplied from duff etc I agree with him in the general sense of it being a multi card issue, i defo wont argue that

Wonder if i should plug my second 260 in and measure it on mine.

 

[Duff-Man]

Wrong

Are you just trolling? Because all he did was state the definition of microstutter...


This thread has gone way downhill, but let me address or reiterate a few points:

1: - Regarding the crysis videos, that is clearly paging. Anyone who has run the benchmark will see this on the first flythrough. This is why the benchmark takes statistics only from runs 2+. On a GTX480 SLI setup it happens just the same as any other, but only for the first run. There is no hitching thereafter (I have a GTX480 SLI setup and plenty of experience with this - see my earlier numbers for quantification of microstutter in crysis benchmarks for SLI and single-GPU).

2: - As far as recording youtube videos to visualise microstutter - this only "works" if the framerate is extremely low (say sub-25fps), such that individual frames can be caught in the video. Given the limited frame capture rate, anything higher is masked by the video.

3: - Microstutter IS a subtle effect: It's not like tearing, or paging, where you can look at a game and say "hey - this is microstuttering!". At reasonable framerates, all it does is take away from the smoothness of the game scene. It will make (for example) 50fps as counted by FRAPS look like ~40fps.

4: - Adding a second GPU to your setup WILL STILL IMPROVE PERFORMANCE in almost every situation. Current dual-GPU scaling lies at around 70-90% efficiency, in GPU limited scenarios. You can expect microstutter to "eat away" 10 to 35% of performance. So you are still left with a sizeable improvement. Just be aware that the numbers you see in benchmarks are not fully representative of true performance increases and there is nothing else to worry about.

5: - In CPU-limited scenarios microstutter disappears, so when the GPU is not the limiting factor you lose NOTHING by adding a second GPU.

 

[kitch9]

Wrong

http://hardforum.com/showthread.php?t=1317582

Educate yourself.

 

[tdream]

Explain then ?

shift 2010-09-06 13-00-04-06 frametimes.xlsx

Need for Speed Shift frametimes @2560x1600 16AF no AA vsync on.

Should 16.67 ms all the way but clearly in the spreadsheet it isn't and it is noticeable in game too on panning shots.

Are you just trolling? Because all he did was state the definition of microstutter...


This thread has gone way downhill, but let me address or reiterate a few points:

1: - Regarding the crysis videos, that is clearly paging. Anyone who has run the benchmark will see this on the first flythrough. This is why the benchmark takes statistics only from runs 2+. On a GTX480 SLI setup it happens just the same as any other, but only for the first run. There is no hitching thereafter (I have a GTX480 SLI setup and plenty of experience with this - see my earlier numbers for quantification of microstutter in crysis benchmarks for SLI and single-GPU).

2: - As far as recording youtube videos to visualise microstutter - this only "works" if the framerate is extremely low (say sub-25fps), such that individual frames can be caught in the video. Given the limited frame capture rate, anything higher is masked by the video.

3: - Microstutter IS a subtle effect: It's not like tearing, or paging, where you can look at a game and say "hey - this is microstuttering!". At reasonable framerates, all it does is take away from the smoothness of the game scene. It will make (for example) 50fps as counted by FRAPS look like ~40fps.

4: - Adding a second GPU to your setup WILL STILL IMPROVE PERFORMANCE in almost every situation. Current dual-GPU scaling lies at around 70-90% efficiency, in GPU limited scenarios. You can expect microstutter to "eat away" 10 to 35% of performance. So you are still left with a sizeable improvement. Just be aware that the numbers you see in benchmarks are not fully representative of true performance increases and there is nothing else to worry about.

5: - In CPU-limited scenarios microstutter disappears, so when the GPU is not the limiting factor you lose NOTHING by adding a second GPU.

No he also said Enable V sync and the variation stops simple as that, which it doesn't as illustrated in the spreadsheet.

]

Regarding your point 3. A lot of people are invested in Sli financially and psychologically, they want the best performance for the price their paying or in my case bang for buck. Now if you don't notice the difference between microstutter @ 60fps average then I'm wiling to wager that you will be satisfied with 40fps too. Except fraps is telling you 40fps instead of a microstuttering 60fps which to objective viewers probably looks the same.

Point 4. You're putting a quantifiable percentage on what performance you expect to lose, except how can smoothness be a percentage? It either is or it isn't.

Point 5. The type of scenario which is polar opposite of what mutligpu configurations are meant for. Lower the res = more cpu limited = don't need such a powerful setup in the first place. The situation where someone would want to get a multigpu setup is for the latest games which uses the latest engines and technology @ higher/highest resolutions which means not cpu limited.

 

[tdream]

http://hardforum.com/showthread.php?t=1317582

Educate yourself.

Hardforum says so it must be true

 

[Rroff]

Bare in mind that most computer systems are limited in precision when it comes to timing especially due to rounding precision/error so your never going to completely erradicate "microstutter". But you can limit its affect so that its not noticeable.

 

[Final8y]

shift 2010-09-06 13-00-04-06 frametimes.xlsx

Need for Speed Shift frametimes @2560x1600 16AF no AA vsync on.

Should 16.67 ms all the way but clearly in the spreadsheet it isn't and it is noticeable in game too on panning shots.


No he also said Enable V sync and the variation stops simple as that, which it doesn't as illustrated in the spreadsheet.

]

Regarding your point 3. A lot of people are invested in Sli financially and psychologically, they want the best performance for the price their paying or in my case bang for buck. Now if you don't notice the difference between microstutter @ 60fps average then I'm wiling to wager that you will be satisfied with 40fps too. Except fraps is telling you 40fps instead of a microstuttering 60fps which to objective viewers probably looks the same.

Point 4. You're putting a quantifiable percentage on what performance you expect to lose, except how can smoothness be a percentage? It either is or it isn't.

Point 5. The type of scenario which is polar opposite of what mutligpu configurations are meant for. Lower the res = more cpu limited = don't need such a powerful setup in the first place. The situation where someone would want to get a multigpu setup is for the latest games which uses the latest engines and technology @ higher/highest resolutions which means not cpu limited.

NFS shift has its own issues so is a bad example.
http://www.xtremesystems.org/forums/showthread.php?t=250349&page=3

I can play it smooth & if i re start the same track a few times then it becomes a juddering mess.

 

[mame]

when the GPU is not the limiting factor you lose NOTHING by adding a second GPU.

 

[1LANCS]

Sick of reading this fanboy **** on every damn forum i go on, all i want is a yes or no, will i notice it, is it worth buying GTX460 SLI?

 

[Duff-Man]

shift 2010-09-06 13-00-04-06 frametimes.xlsx

Need for Speed Shift frametimes @2560x1600 16AF no AA vsync on.

Should 16.67 ms all the way but clearly in the spreadsheet it isn't and it is noticeable in game too on panning shots.

No he also said Enable V sync and the variation stops simple as that, which it doesn't as illustrated in the spreadsheet.

]

Regarding your point 3. A lot of people are invested in Sli financially and psychologically, they want the best performance for the price their paying or in my case bang for buck. Now if you don't notice the difference between microstutter @ 60fps average then I'm wiling to wager that you will be satisfied with 40fps too. Except fraps is telling you 40fps instead of a microstuttering 60fps which to objective viewers probably looks the same.

Point 4. You're putting a quantifiable percentage on what performance you expect to lose, except how can smoothness be a percentage? It either is or it isn't.

Point 5. The type of scenario which is polar opposite of what mutligpu configurations are meant for. Lower the res = more cpu limited = don't need such a powerful setup in the first place. The situation where someone would want to get a multigpu setup is for the latest games which uses the latest engines and technology @ higher/highest resolutions which means not cpu limited.

*sigh* How many times do I need to re-post the same information?


As I explained earlier in great detail, there are edge effects when you are close to the transition framerates with vsync enabled. Sufficiently close to these transitions you can still get irregular frame output. However, at framerates more than a couple of % above the monitor's maximum refresh rate, you will see no microstutter. I have plenty of quantitative evidence of this (which I'm not reporting AGAIN).

Regarding your "point 3", what exactly are you trying to say here? That entire schpeel is incomprehensible... Try to make concise points devoid of weak opinion.

pt4: I put a percentage on the apparent reduction in framerate. "Smoothness" is a subjective thing, which we normally measure in frames-per-second. Microstutter renders the traditional frames-per-second measure of smoothness inaccurate. Read back to where I describe the procedure for quantifying framerate irregularity used in my program. It is a well defined mathematical procedure, using very well known statistical processes.

pt5: This relates back to the "you don't lose by adding a second GPU" principle. Whether you have one GPU or two, there are inevitably places where you will encounter CPU limitation (this varies game by game and game-scene by game-scene). Because multi-GPU output syncs to the regular output of the CPU in situations where the GPU finishes its workload before the CPU, there will be no loss of performance from the addition of a second GPU in such circumstances.



... You sound like you think I'm trying to make a point about microstutter, or multi-GPU setups one way or the other. I'm not. I don't care whether people go for single or multi-GPU setups. I couldn't care less about persuading people to go for one setup or the other. ALL I'M TRYING TO DO is explain to people WHAT microstutter is, and HOW it affects their gameplay experience. This consists of analysing its effects, and as far as possible, quantifying them. By understanding what microstutter is and how it affects gameplay, people can make a more informed choice about what setup to spend their hard earned cash on. In addition, by raising awareness of microstutter as a real phenomenon that can effect gameplay (which it is) we improve the chances of nvidia or ATI actually doing something to reduce it within drivers.