Quote:
Originally Posted by
Sexparty 
It's easy, just get fraps and do a frame time benchmark.
But then the hard part is making something out of the infomation overload it gives you.
Not really, single GPUs don't suffer from micro stutter but dual GPUs do because it's hard for them to be in perfect sync. It's important enough that the GTX 690 comes with an imbedded controller to handle differences in frame times and smooth them out, hence the great performance it gives you over a traditional dual-GPU set-up and was the main reason I bought one.
AMD just need to arknowledge it's existence and improve it either at a driver level or a hardware level, whatever causes it and be done with it because they are the worst culprits. Ask anyone with an HD 6990 and they should suffer from it a lot, if they say they don't they don't know what they're talking about. Realistically it shouldn't be bad enough to bother you but in a lot of cases it is.
My understanding of micro-stutter is it represents the time between each frame is rendered. A constant 16.7ms interval between frames will equal 60 frames per second. What the HD 7970 cross-fire is doing is the first GPU will render the first frame in around 10ms and the second will render it in around 25ms meaning every second frame your pausing to recieve it on screen, recieving constant speed ups and slow downs and causes a jittery screen and flickering textures. With them not being sent in a consistent timing the screen can not look smooth, it's simply impossible. The extent to how severe it is, is however up to the set-up.
Apply that to something slower like a gun firing, every shot pausing a second then every second shot pause 2, then speed it up and it's going to look retarded. Do the same thing with shots being fired at the same 1 second interval and speed it up and it'll just look like a regular automatic weapon firing.
This must not effect all builds. I'm running a 6950 and an unlocked 6950 at different clocks on core/memory, and I have been for about a year (prior I was running them both as 6970's, but I reflashed the other one to a 6950 just incase the theory about long term effects of unlocking on VRAM were infact true.) Anyway, they do and always have run smooth as butter. If it does infact always take 25ms for one card to produce a frame while the other takes 15ms, it would be impossible for AMD cards in crossfire to exceed ~66FPS. This of coarse is not the case.
I'd suspect that the results vary depending on the games tested, and system configuration, and probably even the game settings. That makes me suspicious about the test method/pool. I know I haven't experienced it, personally. I've heard both AMD and nV users complain about microstutter, so I just find it hard to accept what you're using those graphs to propose.
Edit: I just did a two minute test in fraps with Skyrim at 2560 on Ultra and skimmed through the results. The frametimes were between 20ms and 22ms very consistantly. There was a single framerate pitfall that caused a dip below 10FPS and with it, a frame time dip from 10-61ms. I think I might make a thread for this so people can post their results and prove this ongoing "AMD stuttering" issue one way or the other for good.
Edited by Mookster - 6/17/12 at 5:54am
