• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

SLI/CF is microstutter an issue for you?

Caporegime
Joined
24 Sep 2008
Posts
38,284
Location
Essex innit!
I have just been doing some messing around with BF3 and not been overly happy with microstutter (I am hyper sensitive to these things) with my two Titans. I did some messing around with the NCP and had a play with V-sync whilst playing BF3. It seriously makes a big difference and I didn't notice any input lag.

I am curious to know if I am alone in spotting this and if anyone has a way of eliminating it?

If you want to know if you have any Microstutter, download Fraps from here:

http://www.fraps.com/download.php

Under the tab 99FPS, tick the box that has 'Frametimes' and in the box at the bottom which says 'Stop benchmark after []" place how ever long you want to do the test for. I usualy do a 5 min bench but 2 mins will be enough to show if you have any microstutter.

Once that test is complete, you will see the frames come back in the corner. Stop the game and download this little program called 'fraps bench viewer' from here:

http://sourceforge.net/projects/frafsbenchview/

Your bench results are generaly saved in C:\Fraps\Benchmarks and look for the bigger sized files. All you then do is drag and drop the bigger file to the fraps bench viewer program and you can then save that as an image by clicking the second tab which asks to 'save chart as image'

I would be very interested to see some results for certain systems.

vsync.jpg


Here is mine after messing around with V-Sync in BF3 and swapping to one card. Very smooth and very pleased with the results. These tests were at 1920x1080

A couple of stutters in Hitman but nothing to whinge about. Generaly pretty smooth. All the following tests were at 5760x1080.

HMA2013-03-0712-03-20-70-Time.png


Again with Sleeping Dogs, Very smooth with just the one noticable spike.

HKShip2013-03-0712-23-41-31-Time.png


FarCry 3 has been the pain in my side since release and when Nvidia released the 313.95 drivers, it was smooth as a babys bum but even when I switched from 680's to Titans, the more powerful cards struggle to keep this game smooth and this is one of those games that seriously shows the micro stutter. I know it isn't the game now because Nvidia released those 313.95 drivers and if it was smooth as silk in that, it should be for later drivers. Sadly not the case.

farcry3_d3d112013-03-0711-20-44-97-Time.png


Crysis 3 is just horrid as well. Even when I get flying frames per second, it still doesn't feel smooth.

Settings maxed
crysis32013-03-0711-09-00-87-Time.png


AA toned down a little
crysis3time.jpg


As you can see, no problems at all with Tomb Raider. Very smooth.

TombRaider2013-03-0619-09-50-86-Time.jpg


These are just a couple of tests and most are fine but some of these AAA Titles are shocking.

Like I said, I am interested in what you guys results are and if anyone else is getting/seeing microstutter. No fanboys, just good ol fashioned honesty :)
 
Good thread Gregos, will be interested to see the results.

If i win that Ares II expect a ton of AMD latency tests on all the current games.

I'm going to try and get tommy running a few benches using some specific settings.
 
Last edited:
I've been running 5870cf for a couple of years and only experienced stutter occasionally and even then it's not been a big deal in real world usage. It's one of those subjective things that affects some people more than others.
 
Unfortunatly don't have BF3 and took my 470 setup apart before getting my hands on FC3 or TR but its been the exception rather than the rule when I've had issues with microstutter on my 470s. I was doing frametime testing for my own personal interest a long time before there was mainstream awareness of it - example output from my own little app I made for testing it below:

Mafia2 w/ 470 SLI + GTX260 doing physics:

http://aten-hosted.com/images/mf2470p260.jpg

Which looks woefully inadequate compared to what you can do with fraps bench viewer now :D

Vast majority of AAA titles ran with no significant levels of microstutter on my setup but I've always aimed for a setup where one card is getting in the region of 50fps and the 2nd card is contributing to push it over 60fps for most of the time which generally is the optimal setup for avoiding the effects of microstutter with multi GPU.


EDIT: One thing that might be worth trying (I've not followed the Farcry 3 issues) is finding a game that generally works without microstutter issues and uses a similiar or the same engine and using that profile for the offending game, sometimes it can produce good results tho often can cause more issues.
 
Last edited:
The only game i really got terrible stutter was on FC3, and because of that game i never had CF enabled, had the second card for aesthetics :D

If and when i get it back i'll redo frame times for 5760, 1440 and 1080.
 
Good thread Gregos, will be interested to see the results.

If i win that Ares II expect a ton of AMD latency tests on all the current games.

I'm going to try and get tommy running a few benches using some specific settings.

It would be good to see if other cards like 680's/70's7970's/50's and earlier cards get it. I am not sure if it is a placebo effect because I can't remember noticing it before this bench viewer came out but I always felt it wasn't as smooth as could be after going SLI. Obviously V-Sync seriously helps but tech has moved on so much and we should not be seeing this in my opinion now. The 690 is hard modded to reduce it, so I would be very interested to see that Mr Geeman or any other owner ;)

Unfortunatly don't have BF3 and took my 470 setup apart before getting my hands on FC3 or TR but its been the exception rather than the rule when I've had issues with microstutter on my 470s. I was doing frametime testing for my own personal interest a long time before there was mainstream awareness of it - example output from my own little app I made for testing it below:

Mafia2 w/ 470 SLI + GTX260 doing physics:

http://aten-hosted.com/images/mf2470p260.jpg

Which looks woefully inadequate compared to what you can do with fraps bench viewer now :D

Vast majority of AAA titles ran with no significant levels of microstutter on my setup but I've always aimed for a setup where one card is getting in the region of 50fps and the 2nd card is contributing to push it over 60fps for most of the time which generally is the optimal setup for avoiding the effects of microstutter with multi GPU.

lol at that viewer :D How on earth did you know where it was sitting at in seconds? I guess it still gave a fair result, if spikes were evident, you could be pretty sure you was seeing a little stuttering.

I generaly find that anything under 50ms is smooth but when it goes over that, it starts to show.
 
Have you tried using max pre rendered frames at 1?

In FarCry3 I have but makes no odds. I followed a few ideas but completed the game twice and not fussed anymore. It would still be nice to have it smooth though and I will be hounding Nvidia till they give me drivers that replicate 313.95's. These were awesomes and previous 306.xx were real good. 310.xx were just a headache for SLI.

@Rossi, I think you have problems bud. Looked at your post on the 3DMark thread and something isn't right :(
 
lol at that viewer :D How on earth did you know where it was sitting at in seconds? I guess it still gave a fair result, if spikes were evident, you could be pretty sure you was seeing a little stuttering.

It was good for comparing the results from 2 different setups side by side, pretty useless for anything else, I gave up work on it when stuff like fraps bench viewer came along.

EDIT: Thats actually an output graph from the program, the actual program, the program itself looked like this http://aten-hosted.com/images/viz1.jpg with 2 rotating cubes, one rotating at the perfect frametime for the current calculated fps from the fraps csv file and the other using the frametime from the fraps csv file so you could see when the framerate variation was actually visually noticeable (tho lacked tactile feedback which can be more noticeable than visual), it had a lot of shortcomings but it was early days heh.
 
Last edited:
I generaly find that anything under 50ms is smooth but when it goes over that, it starts to show.

I'd agree but at lower than 50ms....

50ms is 20FPS so not really micro stutter, just poor frame rate.

I generally find If the fps is >60 microstutter is not an issue. A few ms deviation from 16ms (60fps) is better than most monitors can really show though my main screen is 120Hz I don't have 120Hz eyes.

I've had a pair of 2GB 560TI's for the last 18 months or so and as I'm only at 1080p and happily cut the detail to keep the FPS I haven't had any issues.

With your triple screens I can see how it easily becomes and issue with the less forgiving games.

AD
 
I'd agree but at lower than 50ms....

50ms is 20FPS so not really micro stutter, just poor frame rate.

I generally find If the fps is >60 microstutter is not an issue. A few ms deviation from 16ms (60fps) is better than most monitors can really show though my main screen is 120Hz I don't have 120Hz eyes.

I've had a pair of 2GB 560TI's for the last 18 months or so and as I'm only at 1080p and happily cut the detail to keep the FPS I haven't had any issues.

With your triple screens I can see how it easily becomes and issue with the less forgiving games.

AD

Some good points and for sure it doesn't help on triple screens. It just puzzles me that one set of drivers can deviate so much. FarCry 3 was awesome on just one set of drivers and regardless of frames, was so smooth and yet all the other drivers make it play like it has a twitch :(.

@ Davanius, that is real smooth. I bet there is no stutter when you are playing? even in MP :)

Thanks for doing that :)
 
I don't see any stutter with these drivers but I did with the 314.21.

Thanks for giving us something to do.

You have just taught me something new. Cheers. :)
 
RadeonPro with Flip Queue Size 1 seems to have solved most of my stuttering. Apart from Unigene and Firestrike benchmarks, just run terribly.

Are AMD working on resolving the crossfire stutter at the moment?
 
I don't get it Rossi. I did on 12.11s to some degree but the games I've been playing recently have been fine.

Which confuses me. There are people that say they do and some that don't.
Can you post Firestrike and Unigine graphs for me? I haven't been able to get an acceptable score on either because it's so bad.

As much as i feel i shouldn't, (because AMD's crossfire drivers/profiles are beyond rubbish in my experience) i have to resort to RadeonPro for games, which seems to clear most of it up, apart from these benchmarks.

I've been spending hours looking for the cause of this issue, from BIOS settings and limiting PCI-E to 2.0, to random applications, but to no avail.
 
Back
Top Bottom