Microstuttering tests on ATI Radeon HD 4870 X2

[annihilat0r]

It may not be optimal, but what is the harm?

Input lag below 35FPS is intolerable. Plus, Vsync doesn't completely remove microstuttering

[Seraphiel]

I don't think Sampsa's tests will prove anything. And I do not believe this problem is solved or even understood, yet.

Plus, Vsync doesn't completely remove microstuttering

That is true.

[Cooper]

I don't think Sampsa's tests will prove anything. And I do not believe this problem is solved or even understood, yet.

Why it won't? You have a better methodology? Or this post ain't based on anything?

[L7R]

Damn ME if I ever buy a SLI or CF system again.

But like me, your only mistake you bought it at the wrong time.
As Sampsa has tested there's no issues anymore with new Ati cards. You can just go and crab a second 4870 and check if it's true.

[fellix_bg]

I don't think Sampsa's tests will prove anything.

It's not about much of a proving something, but to show the phenomenon in more wide detail set, IMHO, and hence a possible way to understand it after.

[Apoc@]

Vsync is not good in multi GPU due to the lack of triple buffering.

Could you please explain that?

[L7R]

Could you please explain that?

It doesn't work. That's that. Don't know about OpenGL but haven't see Tripplebuffering working on D3D with Crossfire. It works at 3Dmark06 rolling demo, though.

[Seraphiel]

Why it won't? You have a better methodology? Or this post ain't based on anything?

Because I have tested Grid myself. I can show benchmarks from that game with SLI, that gives the impression that there is no significant microstutter, and I can give others that will show the opposite.

No, I do not have a better method, and I am really curious about this phenomenon myself. However, taking a few games at certain settings, is not enough to conclude that microstutter is a thing of the past, nor does it prove anything at all in general.

I explore this, and I am thankful for Sampsa's results. However, it is too soon to tell, if the R700 solves this or not. Especially from these limited results.

EDIT:
It may not have been Sampsa's intend, but this particular quote made alot of people think the problem is solved:

Good news! I can confirm that based on my own tests microstuttering is gone on R700!

Sampsa's results may be true, and R700 have solved it all. But with so little tests done, I do not believe it wise to conclude anything yet.

I hope I am not misunderstood.

[TheVaLVe]

Could you please explain that?

The lack of triple buffering will give you input delay with Vsync on.

[Logos]

how does GTX280 SLI micro-stuttering perform compared to the results here ? was that already posted anywhere in the forum ?

[tiro_uspsss]

how does GTX280 SLI micro-stuttering perform compared to the results here ? was that already posted anywhere in the forum ?

GTX260 too for that matter!

[adamsleath]

i'd guess it would be no better than the 9800gx2 microslutter-wise.

ms does not discriminate between multi-gpu options from either company.

i still have the impression that it affects particular games more and others not much at all it's a bit of a worry; otherwise the card looks super, performance-wise according to the games fps graphs.

[Apoc@]

In OpenGL you can use Triple buffering, that's the little check box for in the Catalyst drivers. But that there isn't any Vsync in D3D available, i didn't know....

Shame on me, had to do more reading on that topic.

[kromosto]

well graph shows that in grid 4870x2 gives 57 to 66 fps in first 30 frames. difference is 9 fps so i dont think we can say ati solved this problem but they made it more acceptable.

[Eson]

No they do not, but they are pretty close to it. It all depends upon the scene and what is to be rendered.

So there is no clear definition about what is microstuttering and what is not?
Basically its an argument you can never win or lose with.

Just like the TFT vs. CRT debate, there will always me be people claiming they notice that small delay and stick to their CRTs. I think its a matter of learning, its not the tools that makes the craftman.

[Logos]

In OpenGL you can use Triple buffering, that's the little check box for in the Catalyst drivers. But that there isn't any Vsync in D3D available, i didn't know....

no VSync in D3D who said this ?

[Shintai]

how does GTX280 SLI micro-stuttering perform compared to the results here ? was that already posted anywhere in the forum ?

SLI and CF just shutter the same now. Both sucks.

[annihilat0r]

shutter, slutter... guys, it's "stutter"

[Cooper]

SLI and CF just shutter the same now. Both sucks.

Wrong thread dude

[Logos]

SLI and CF just shutter the same now. Both sucks.

OK then, and let's just dump our computers and get ourselves new typewriters

[adamsleath]

sampsa's graph is one way of measuring the effect of ms.
i'm interested to know whether higher fps makes the 'effect' worse or better?

...and i think to those who notice any sort of effect from multigpu vs a single gpu it can be very annoying, and others couldnt care less or dont even notice it.

OK then, and let's just dump our computers and get ourselves new typewriters

got an electric typewriter in my cupboard it's been redundant for quite a few years :|

[kromosto]

OK then, and let's just dump our computers and get ourselves new typewriters

does typewriters with double cartridges has microstuttering?

[Shintai]

Wrong thread dude

So this aint the thread with microshuttering test, including both X2 and GX2 cards along with regular CF? I be damned...

OK then, and let's just dump our computers and get ourselves new typewriters

Or just use single GPU cards. And have no worries about profiles jumping performance, microshutter and endless feature issues

[flopper]

So there is no clear definition about what is microstuttering and what is not?
Basically its an argument you can never win or lose with.

Just like the TFT vs. CRT debate, there will always me be people claiming they notice that small delay and stick to their CRTs. I think its a matter of learning, its not the tools that makes the craftman.

No, there is a felt and seen issue with lcd and crt.
The difference is BIG due to the aim gets way off with inputlag and the result is less k/D ratio

This microstutter issue seems to be a real issue, however, people cant replicate it consistently?
Whats that about?

If the sync only gets sense in peoples perceptions and not on hard data, the issue will not be resolved anytime soon.

[Apoc@]

no VSync in D3D who said this ?

On this page:

http://www.tweakguides.com/Graphics_9.html

The only thing they say is that if your pc is not fast enough you "could" have some kind of input lag. So i don't get it that people say it doesn't work on a MultiGPU setup.