Why FPS fails

[Psi*]

I copied that subject line from this article. The real topic is micro-stuttering.

To be honest, I have never heard of this before and it wouldn't impact my typical activity with PCs, and I only drilled into this because I don't want to deal with spending a lot of $$ to get to Santa Clara (a geek fest) in a couple of weeks that I really need to do!

But I do game & I want to understand technology & the related issues outside of my day to day focus. I also enjoy nothing better than to raise issues when my geek (in my business) friends talk about PC tech that they haven't understood very well ... but like to sound like they do. This might be one of those issues.

This is inspired by a growing interest in iRacing with 3 monitors. I prefer FPS games, but I am a real racer & looks much less expensive. If I might make the jump to 3 monitors then I want to be well informed.

 

[RainMotorsports]

The article which I am very familiar with uses a tool that has had a feature thats been available for a very long time. Fraps in benchmark mode will do frame times which is how many milliseconds each frame took to render.

If you follow that 24 frames per second appears fluid motion to the human eye than that as your minimum (not average) framerate means this. Anytime a frame takes longer than 41.6ms to render you will perceive a stutter. Personally I would prefer a 30 fps minimum with an average of 60.

60 frames per second solid would be 16.6ms per frame, now in video games your not getting a solid frame rate. If your getting a solid average of 60 frames per second it means the average frame time is 16.6ms but not every frame is 16.6ms. If you have one frame that is say 50ms you will see a stutter despite getting an average of 60 fps and even a minimum above fluid motion.

It would also be argued by those who like 120 frames per second if you were getting mostly frames at 8.3ms (120 fps) and all of a sudden got a couple 30ms frames you would perceive a change in the fluidity even thought the frame times are below 41ms.

Triple monitor is a whole different complication over multi gpu. Multi gpu is a given for triple monitor. Your either going to do it or your not. If you look at alot of the tests the frametimes are very wild on paired low end cards but much tighter on high end cards. This goes for both brands interestingly enough. So a pair of 6970's or hopefully 7970's dont just perform better as a whole they perform better as a pair. Nvidia guy myself but were still waiting lol.

 

[Psi*]

thanks Rain, so is micro-stuttering such a bad thing? I suppose it depends how often the frame slips occur.

About the triple monitors you state that multiple GPUs is a requirement? I am guessing that this is hidden indirectly in some multi-GPU benchmarks ... but really? But a card with 3 DVI ports by itself would not work well for any kind of action game?

 

[Tharic-Nar]

As someone who has 2x AMD HD 5870s, microstutter is very perceptible in unoptimised conditions (new game/bad drivers). It will make 40-60 FPS look like 20 since the stutter is often regular. When the frame rate goes up much higher, it is less perceptible, but still a nuisance.

With multi-monitor rigs, multiple GPUs are only a requirement for NVIDIA cards. Realistically though, it's recommended for AMD too if 3D games are concerned. The other thing though, if you are going for 3 displays, one of them MUST use DisplayPort. So 1xDP and 2xDVI, or something like that. It's so that the daisy-chaining functions of DP can be used. Any additional connections must be through DP since the cards can only support up to 2x non-DP connectors, this includes passive DP adapters too, so you can't just buy 3x DP adapters and expect it to work.

http://www.amd.com/us/products/technologies/amd-eyefinity-technology/how-to/Pages/faqs.aspx

 

[RainMotorsports]

Psi I was just going into the basic concept of a perceivable stutter. It occurs on single cards as well but its not for the same reasons and its not as common. If you can not perceive it, might not be too bad. But to be safe I would say that a smoother performance in say an FPS game might increase reaction and aiming just enough even if you dont think your noticing it. Its kinda difficult to say, but if you see it, its annoying and probably will affect your gameplay.

This video shows an example of multi gpu micro stutter that may cause a very constant issue if percievable by the user. The video shows the setup experiencing 2 very fast frames followed by a long frame.
<iframe width="619" height="348" src="http://www.youtube.com/embed/zOtre2f4qZs" frameborder="0" allowfullscreen></iframe>

@Tharic yeah yeah yeah haha. If all you want todo is multi monitor web surfing there are low end nvidia solutions (450/460 range) that use an old trick with a chip for 4 monitor output. But even so not sure I would want to triple monitor anything other than portal with a single 5870,6970 etc. Yeah such a mess with connectors why cant we just get mini hdmi

@Psi You have to remember lets say we down the resolution to a managable 1280x720 spread across 3 monitors this is 3840x720. Thats 2,764,800 pixels per frame which is more than 1920x1080 at only 2.07 million. So at that resolution you might get away with a spanking new 7970 pretty well. But once you up the res to 1600x900 or 1920x1080 were talking a total resolution alone that a single card wouldnt handle at playable framerates.

1600x900 @ 3 monitors is 4.3 million which is a bit higher than the pixel count of 2560x1600 which would be a good struggle for serious gaming on any single card.

 

[Psi*]

LOL you guys have me thinking and re-reading your posts. This is almost like a foreign language to me and it shouldn't be. Definitely an area that I have not paid any attention to.