NVIDIA GTX 690 Official Launch "Performance Perfected"

[icemanchilled]

You missed a main fact - the whole article you are referring to self imploded in two words -not perceptible.

Here's something else to think about: How the extra latency presents itself to the user? For some reason CB automatically assumes this means input-lag, but i don't think it means it in that sense. The frame metering technology averages out the rendering times of the frames, these bnatural variances in the rendering time for each frame can present as jitter/stutter, (in multi-gpu setups). So, say one frame took only 5ms to render and the next took 15ms, how does this difference usually present itself? Usually, when a frame takes longer to render, it simply means lower fps overall, so the only effect the end user would see is lower fps. I don't think it has an impact on input-lag at all, unless the fps gets low enough, which would affect all cars then anyway.

[valtterieranen]

When frames take longer to render, there's more input lag. Pretty much everybody who's tried 120hz monitors are noticing and applauding a mere 8ms improvement in input lag against 60hz. Doubling the framerate from 30 to 60 cuts the lag down by 16ms.

"Tens of milliseconds" makes the difference between unnoticeable and barely playable between my 20ms and 50ms TV's.

"Tens of milliseconds" is much more than the extra 16ms of lag Triple Buffering induces at 60hz and many people complain about that. This "tens of milliseconds" is added on top of one extra frame of lag, 16ms at 60hz, for just having dual-GPU AFR rendering.

From a subjective point of view, the difference between SLI and no SLI in terms of input lag is very apparent to me. SLI feels considerably more floaty in terms of mouse response.

[Kaotika]

Valtteri already summed the most important things it seems

[icemanchilled]

When frames take longer to render, there's more input lag. Pretty much everybody who's tried 120hz monitors are noticing and applauding a mere 8ms improvement in input lag against 60hz. Doubling the framerate from 30 to 60 cuts the lag down by 16ms.

"Tens of milliseconds" makes the difference between unnoticeable and barely playable between my 20ms and 50ms TV's.

"Tens of milliseconds" is much more than the extra 16ms of lag Triple Buffering induces at 60hz and many people complain about that. This "tens of milliseconds" is added on top of one extra frame of lag, 16ms at 60hz, for just having dual-GPU AFR rendering.

From a subjective point of view, the difference between SLI and no SLI in terms of input lag is very apparent to me. SLI feels considerably more floaty in terms of mouse response.

I'm not disputing any of that. What i'm talking about is CB attributing it directly to the framerate metering, without actually having anything to back that up. You said yourself, in the other thread, you experienced the same sort of input-lag with Xfire 4800s, so that would indicate to me the metering is not the cause of the lag you were experiencing. I don't think it would add any significant lag at all, unless performance drops significantly, and then as the article at Techreport says, it could compound the issue, making it worse.

Valtteri already summed the most important things it seems

No. In our discussion, the most important thing is you providing proof that framerate metering adds a notable lag in comparison to hardware that doesn't.

[valtterieranen]

Fair enough. Yes, the actual amount from this metering remains open.

[Kaotika]

There is proof about the fact it adds lag, only thing without proof is how much, and nVidia at least didn't see need to correct TechReports "up to tens of milliseconds" estimate

[dragomirc]

Have a mercy CB, not again.

[icemanchilled]

There is proof about the fact it adds lag, only thing without proof is how much, and nVidia at least didn't see need to correct TechReports "up to tens of milliseconds" estimate

Why not put the whole quote CB? Let me help with that:

In most cases, we're talking about tens of milliseconds or less; that sort of contribution to lag probably isn't perceptible.

That's tens of milliseconds or less and it probably isn't noticeable.

[valtterieranen]

I think Techreport is insane for calling tens of milliseconds not noticeable.

[Kaotika]

Why not put the whole quote CB? Let me help with that:

That's tens of milliseconds or less and it probably isn't noticeable.

As Valtteri said, for him at least moving from switching from 30ms to 50ms TV makes the difference between "unnoticable" and "barely playable"

If the "tens" was 30 or even just 20, it probably wouldn't be noticeable - but when you account your monitor, AFR introduced lag and possibly mouse etc too, you're "quite a bit" higher than that 20.

Also, the metering IIRC is used regardless if you use SLI or not.