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

[Rroff]

edit nm

 

[tdream]

Perhaps he has sigs turned off.

perhaps

 

[kitch9]

perhaps

I always turn sigs off, as I post a lot from my phone.

So are you saying all your games stutter as per the video?

I'd have sold my cards a long time ago if that were the case. I've just come from a 5870 and got a nice boost with my pair of 460's, apart from one title (GTA4) everything is smooth in SLI, including Crysis. I'd like to think I'm sensitive to stuttering as I'm a guy who always chases the 60fps dragon.

 

[Jibber55]

I always turn sigs off, as I post a lot from my phone.

So are you saying all your games stutter as per the video?

I'd have sold my cards a long time ago if that were the case. I've just come from a 5870 and got a nice boost with my pair of 460's, apart from one title (GTA4) everything is smooth in SLI, including Crysis. I'd like to think I'm sensitive to stuttering as I'm a guy who always chases the 60fps dragon.

I'm not sure where you've got the idea this is a thread bashing multi-gpu setups or 460's have you read the start of the thread?
The guy has parted with, what I consider, valid information that is worth knowing about.
NOBODY is saying this makes sli/x-fire pointless or unusable at all but when I come to buy my next gfx card I will keep this info in-mind. It will not stop me buying multi-gpu if it's the right decsion for me.

 

[kitch9]

I'm not sure where you've got the idea this is a thread bashing multi-gpu setups or 460's have you read the start of the thread?
The guy has parted with, what I consider, valid information that is worth knowing about.
NOBODY is saying this makes sli/x-fire pointless or unusable but when I come to buy my next gfx card I will keep this info in-mind. It will not stop me buying multi-gpu if it's the right decsion for me.

Yes, I've read the start of the thread and the OP is claiming the stutter in the video is micro stutter, to which I'm saying it isn't, simple as that.

I'm also saying Crysis does not stutter like that on my 460's, I'd be taking a long hard look at my system if it did.

Have you seen the thread title?

 

[bfar]

Well, it can be measured. Data irregularity is a rather simple to quantify analytically (I wrote a program to do just this from FRAPS benchmarks).

You've been able to measure the variance/irregularity in the intervals at which frames are drawn (great idea!). And we can see there are bigger variances withe the dual card setups, for sure. But here's the thing - we still don't know why that affects our perception of fluidity in the animation. We still don't know for sure if it does! We could actually be missing something else.

As for the human-effect: Certainly it IS subjective, just like framerate (some people are happy with 40fps, some need 100fps to feel completely 'smooth'), but the way the brain processes data from the eye is well understood,

Agreed, but it's very difficult, if not impossible, to quantify in any meaningful way. The 60fps target has always been a good middle ground, not an actual law.

so a comparative framerate is easy to compute (comparative as in an equivalent framerate should there be no microstutter).

Comparative framerate? Its still drawing x frames/sec, albeit in an uneven way. "So my 60fps is better than your 60fps" Even single cards don't draw frames at perfect intervals.
We need a new variable to measure and average the variance/irregularity, and possibly another one to quantify the actual stutter effect. But before we get carried away, we need to be absolutely certain its the draw rate irregularity thats causing stutter - it could actually be a whole host of other things.

I agree that it's still a controversial beast, but this is mainly due to:
a) It being a subtle effect which is not immediately recognisable
b) It not being as straightforward and easy to understand as simple framerate measurements

Totally agreed.

c) People's attachment to a particular configuration (single or multi-GPU) and desire to believe either one way or the other that their chosen config is "the right way of doing things". Preconceptions cloud the mind, even if they are largely subconcious.

Mmm, that works both ways - I think this thread is proof!

 

[Jibber55]

Have you seen the thread title?

Yes, and I said in an earlier post that I thought it was "a little edgy" but as 460 price/performance makes it the obvious choice for some multi-gpu fun he thought it was a relavent title. I'm not sure if I would have done the same thing but I can see where he's coming from.
It's all in this thread if you choose to read it...

 

[Duff-Man]

Comparative framerate? Its still drawing x frames/sec, albeit in an uneven way. "So my 60fps is better than your 60fps" Even single cards don't draw frames at perfect intervals.

The human eye perceives the gap between frames far more strongly than the raw number of frames output. Given this, a comparative framerate (or effective framerate, or equivalent smoothness measure or whatever you want to call it) should be based on the largest local gap between frames, rather than the inverse of the average frametime.

Consider, as an example, the extreme case of two frames being output virtually simultaneously, with a 33.3ms gap thereafter. While a framerate counter would show a solid 60fps, the eye would obviously see a 30fps scene, being unable to distinguish between the two simultaneous frames. In this case it is clear that, as you put it, "my 60fps is better than yours". The actual framerate is 60fps, but the perceived smoothness is equivalent to a 30fps scene.

I concede that such an "equivalent framerate" index will not be perfect, but it is certainly a more appropriate measure of smoothness than raw FPS, which ignores irregularity entirely. It is true that single GPU setups also have some local frametime variation, but this is typically a 1-5% variation based on the tests I have performed. In this case the "true" framerate and "perceived smoothness" will be very similar, and the basic FPS count remains a valid measurement. With dual-GPU setups we typically see 10-35% irregularity in GPU limited scenarios, which drives the perceived smoothness further away from the measured framerate. In these cases the raw fps count is much less accurate.

 

[bfar]

Maybe a second is too long a unit of time. How about frames per deci, or centi second!!! Averaged out, those figures might support your theory.

Don't get me wrong, I think you're onto something. It's just that having gone from a decade of powerful single cards to a sli setup, I haven't seen any issues over a range of games/scenarios. The sli solution feels as fluid and faster than anything I've used before (and believe me, I have decade long habit of buying big loud expensive cards!). That leads me to believe that the irregularity issue isn't endemic, and is therefore possibly caused by something else, possibly preventable.

 

[An Exception]

dc

 

[An Exception]

delete

 

[rounddodger]

Um?