Not quite sure what Crossfire/SLI Microstuttering looks like?

[stys]

It's not a myth. It's there and you can't explain it. Such is anamoly. Good frames, bad performance. I'Its actually macro-stuttering it's butt off. Benn telling you that for years now.

Maybe for you to get it, they should have pulled out the 480 and ran it again with a 5870 VS. 5970 to make you all ATI happy. Then you would have seen lower frames perform better.

[YCH]

For once, I can actually see what you're trying to point out!

Anyway, while it's blatantly obvious in that video at that time point (that's all I looked at), is there any side-by-side with both cards at 60fps+?

There are parts of the video where both cards are rendering more around 35- 40fps for a while and it's hard to tell the difference in the video (at around 4:00 into the video).

Also, when the scene spins around the floating objects (around 4:12) you can also see that the HD5970 is slowing down every once in a while then speeding up quickly.

[icemanchilled]

But the actual framerate fluctuates from start to finish on both, yet the stutter only shows up on one - what gives?

What gives is the video is showing the GTX 480 in a better light than a 5970, you should know CB cannot let that sort of thing go without trying to give some asinine reason for it, that's his point, nothing to do with the actual video pointing out the difference between a single and multi gpu configuration, just the same old overly defensive, knee-jerk reaction from CB. [rofl]

[Kaotika]

What gives is the video is showing the GTX 480 in a better light than a 5970, you should know CB cannot let that sort of thing go without trying to give some asinine reason for it, that's his point, nothing to do with the actual video pointing out the difference between a single and multi gpu configuration, just the same old overly defensive, knee-jerk reaction from CB. [rofl]

It wouldn't change anything if it was GTX480 SLI vs single HD5870 in this case - the video does not show "microstuttering" as seen on monitor (and the fluidness of the other video despite the FPS fluctuations is suspicious for sure)

Let's try explaining once more why it won't show microstuttering like it's seen on monitor
First reason is the most obvious - the video is played at 30 FPS, even if you could perfectly sync the actual capture and source refresh rates and keep solid FPS, exactly at the refresh rate through the whole benchmark, not a bit under or a bit less, you would end up encoding those ~60 frames per second into 30 frames per second - on a lossy codec which uses motion compensation among other technologies which change the way video looks (especially in motion)

Then, when we go a bit deeper to the issues present in this specific video
First, you have the actual test running on screen, in which the framerate fluctuates a lot during the course of the test.

Second, you have monitor, running at ~60Hz, which obviously means you might send more or less frames per second than the display can show, meaning it won't show you all the frames necessarily, and it might show some frames uneven times regardless of their actual rendering time.

Third, there is the actual camera, which depending on camera most likely captures at 24, 25, 29.94 or 30 frames per second, which again is different from both the framerate and how many times the display can refresh itself per second.
Even if the camera would capture lossless data, which it doesn't as far as I'm aware, it would do so at slower rate than the display can show images, meaning that not all the images would end up on the camera capture, and some can be shown longer than they should.

Now all the above is in the end encoded into a video with all the fancy features of the codec of your choice, at the FPS of your choice, which might be again different from the camera FPS especially if it isn't 30 on the camera.

And finally, YouTube will encode it to h.264, MPEG4 or MPEG2 depending possibly on source, but at least the resolution, and to 30 FPS regardless of what the FPS of the video you sent there is.

After all that, does anyone think the video on youtube is even close to what happens on screen?

[kitch9]

I've seen it with me 2x460's, and I saw it with my 4870x2.

It happens much less then I remember though.

[IgLoCSka]

Kaotika, that's all fine, we should see something much worse than we would on screen with those cards.

However, for some reason even with all those factors that are detrimental to the quality, one of the recording of one two cards manages to remain fluid, despite the fact that it's also varying a lot in FPS, whilst the other is as choppy as it gets.

Now don't get me wrong, I understand the need to synch the frames, but are you honestly suggesting that the GTX480 in that vid, fluctuating all over the place between 15 and 30 FPS always gets synced while the radeons fluctuating between 10 and 25 FPS never do?

It just doesn't make sense!

[Kaotika]

Kaotika, that's all fine, we should see something much worse than we would on screen with those cards.

However, for some reason even with all those factors that are detrimental to the quality, one of the recording of one two cards manages to remain fluid, despite the fact that it's also varying a lot in FPS, whilst the other is as choppy as it gets.

Now don't get me wrong, I understand the need to synch the frames, but are you honestly suggesting that the GTX480 in that vid, fluctuating all over the place between 15 and 30 FPS always gets synced while the radeons fluctuating between 10 and 25 FPS never do?

It just doesn't make sense!

No, I'm not suggesting that, but I'm suggesting that GTX480 part of the video is smoother than it is in reality, everyone knows that those sub-20 numbers are not fluid on the screen, while there's no difference in the video between those and the areas where the FPS is 30 FPS or higher

And the stuttering on the 5970 video is worse than microstuttering is on ones screen for unknown reason, that looks more like it should on any card at those sub-20 levels (well, a bit worse but still a lot closer to what any card should show at those FPS's than the GTX-part of the video)

That should be called more like "macrostuttering" like stys referred to it earlier, microstuttering isn't anywhere near that, and something that most people don't even notive (though most likely majority of enthusiasts do notice)

[IgLoCSka]

No, I'm not suggesting that, but I'm suggesting that GTX480 part of the video is smoother than it is in reality, everyone knows that those sub-20 numbers are not fluid on the screen, while there's no difference in the video between those and the areas where the FPS is 30 FPS or higher

And the stuttering on the 5970 video is worse than microstuttering is on ones screen for unknown reason, that looks more like it should on any card at those sub-20 levels (well, a bit worse but still a lot closer to what any card should show at those FPS's than the GTX-part of the video)

That should be called more like "macrostuttering" like stys referred to it earlier, microstuttering isn't anywhere near that, and something that most people don't even notive (though most likely majority of enthusiasts do notice)

So are you suggesting that the vids are plain simply doctored and fake? Because that I'd much rather believe than the camera, youtube, etc simply making one card look smooth and the other not...

[Kaotika]

So are you suggesting that the vids are plain simply doctored and fake? Because that I'd much rather believe than the camera, youtube, etc simply making one card look smooth and the other not...

No, I'm suggesting for whatever random reason, something went wrong somewhere and the resulting video on GTX is too smooth. The HD5970 looks just about like it should at low FPS's on any card more or less (and the video doesn't show microstuttering)
I don't think the video maker intentionally made it like that, though.

[icemanchilled]

No, I'm suggesting for whatever random reason, something went wrong somewhere and the resulting video on GTX is too smooth. The HD5970 looks just about like it should at low FPS's on any card more or less (and the video doesn't show microstuttering)
I don't think the video maker intentionally made it like that, though.

I think what you are talking about is too far fetched, you usually find that the most simple explanation is the correct one.

Saying that the codecs and encoding is so bad as to produce a different result each time is silly, that just wouldn't be accepted by the thousands of people that use these tools everyday, they need and want consistency in their work, so these tools have to produce a consistent result each time.

These are Trubitar videos too, which mean they have done this many times and will have a consistent setup when recording and encoding these videos, it simply shows that the 480 is much smoother in Heaven at low FPS, as we know, micro-stuttering is more exaggerated at low fps and so becomes much more noticeable, which is shown by the fact that the 5970s video actually gets much smoother when the fps increases.