• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Tri-Fire, Microstutter, and Bus Speed

USMCGrunt

2[H]4U
Joined
Mar 19, 2010
Messages
3,103
So I'm currently running x3 6970s in crossfire and if my frame rate drops below 60fps I start getting some wicked microstutter. Right now, I'm running in 16x8x8 mode, I'm wondering if anyone knows if running in x8x8x16 mode would reduce the stutter? The motherboard I'm using has slot 1 and 3 sharing lanes and 5 and 7 sharing lanes. If I put a card in 1, 3 and 7 I could accomplish this. As it stands they're in 1, 5, and 7.

This microstutter keeps me from running eyefinity resolution at max settings on some games..which sucks :(

Of course, if no one knows I'll do some testing to see...
 
Unlikely, microstutter below 60fps is inherent to all Crossfire/SLI to some extent, you just have to keep the frame rate over 60.
 
No Microstutter here when I dip below 60fps.

Of course im running 1 6990 and 1 6970....not sure if its because you are running 3 cards.
 
Is the card the monitors are connected to in the X16 slot? I wonder if people who see microstutter in 3 card configs could be experiencing a limitation in the crossfire bridges. Seems people who run trifire using a 6990 and 6970 etc.. don't experience the same as say someone running 3 cards of some kind.

What I'd do first is try moving the monitors to another card if it's in the x16 slot, See if there is any difference.
 
running 3 6970's in TRI-FIRE here and have no microstutter, the only problem i have is all my games run great with 3 cards except Black Ops ( i get flickering with 3 cards i have to disable 1 card ) what drivers are you running?
 
running 3 6970's in TRI-FIRE here and have no microstutter, the only problem i have is all my games run great with 3 cards except Black Ops ( i get flickering with 3 cards i have to disable 1 card ) what drivers are you running?

Hmm I dont get that, im running 11.10 preview V2.

I also when i play black ops. Have it set to super sampling, 8x with edge detect on
 
Running the 11.10 V2 preview drivers (still same microstutter on 11.10 V1 and 11.9 too). The card running at x16 IS the card driving the monitors. Im wondering if the x16 card having all the extra lanes ends up waiting around for work more then the other two causing the microstutter. My idea is that maybe having the "bottleneck" of an x8 card as the monitor driver will reduce/eliminate the stutter. Maybe ill have time tonight to do some fraps video captures and comparisons to see if there is a difference.

Something else to take into account is that I might be feeling the effects of having an NF200 in the chain that others may not.
 
I was wondering this too. I'm just not sold on the idea of the nF200 personally, especially given how much of a pain they used to be.
 
I was wondering this too. I'm just not sold on the idea of the nF200 personally, especially given how much of a pain they used to be.

Well, this is my first step into the world of Intel and I was under the assumption that the P67 and Z68 chipsets were unable to run x8 across 3 cards minimum without an NF200...believe i was wrong though.
 
Well aint this a bitch...I took a video capture using fraps and and it was clearly stuttering for me but in the video it was silky smooth...guess it makes sense though....
 
Switched the cards around last night and had some playtime....didnt make a damn bit of difference...
 
Switched the cards around last night and had some playtime....didnt make a damn bit of difference...

MAybe try replacing the bridges? Iv seen a few bridges that are brand new, not work out of the box?

Maybe something is funky between them.

Also have you tried just 2 cards to see if you notice microstutter?
 
I was running two cards for a long time and there WAS a bit of stutter but it wasnt until I was under 30 fps or thereabouts...so I never worried about it then.
 
Are you sure the stuttering is because of Tri-Fire and not just because you're under the refresh rate of your monitor?

Try using only one card and tweak things around to get the same FPS you notice stuttering at. If it still has "micro-stuttering" then it's just a refresh rate issue and you should probably turn down your settings to maintain more than 60 FPS.
 
I get micro-stutter sometimes on my 6990 (had it on my 5970 prior also), but alt-tabbing out and back into the game once fixed it for me.
 
I get micro-stutter sometimes on my 6990 (had it on my 5970 prior also), but alt-tabbing out and back into the game once fixed it for me.

Maybe alt tabbing disables crossfire? Can you check to be sure?
 
So I'm currently running x3 6970s in crossfire and if my frame rate drops below 60fps I start getting some wicked microstutter. Right now, I'm running in 16x8x8 mode, I'm wondering if anyone knows if running in x8x8x16 mode would reduce the stutter? The motherboard I'm using has slot 1 and 3 sharing lanes and 5 and 7 sharing lanes. If I put a card in 1, 3 and 7 I could accomplish this. As it stands they're in 1, 5, and 7.

This microstutter keeps me from running eyefinity resolution at max settings on some games..which sucks :(

Of course, if no one knows I'll do some testing to see...

Use a framelimiter like the one included in the video capture software Dxtory, set it to your min-fps and your microstutter will be gone for the most part.
See one of my tests here (580 SLI, but still):


Blue is without limiter, red is with.
 
Are you sure the stuttering is because of Tri-Fire and not just because you're under the refresh rate of your monitor?

Try using only one card and tweak things around to get the same FPS you notice stuttering at. If it still has "micro-stuttering" then it's just a refresh rate issue and you should probably turn down your settings to maintain more than 60 FPS.

How do you figure its refresh rate? I've played plenty of games at a frame rate lower then my refresh and not had this. Im not jumping the bandwagon here and trying to say its something its not.

Use a framelimiter like the one included in the video capture software Dxtory, set it to your min-fps and your microstutter will be gone for the most part.
See one of my tests here (580 SLI, but still):


Blue is without limiter, red is with.

How exactly is a frame limiter going to help? Microstutter is the video cards delivering the frames unevenly. Basic example is Card A takes 1 sec, Card B takes 1.2 sec, Card C takes 1.1 seconds. At least thats my understanding of it from reading this article --> http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-3.html
 
A few quick questions I came up with;

1. Are you forcing +20 powertune in the CCC, AMD overdrive settings?
2. Are your cards overclocked at all? Have you tried at stock speeds?
3. Have you checked your bios settings to make sure primary video is on pcie and everything else is ok there?

I would advise against overclocking the cards and not forcing the powertune setting to some kind of increase. If you exceed the driver controlled TDP limit the card will throttle gpu clocks in no organized fashion. You could run furmark or Kombuster by msi and monitor with an on screen display and see the fluctuations happen on all 3 gpu's and to odd frequencies that are not near eachother or allow the cards to sync. It's a theory of mine that by having all 3 gpu's frequencies jump around it could keep them out of sync and thus introduce microstutter when they are being stressed beyond their driver controlled tdp limit. Please look into this and post back.
 
How do you figure its refresh rate? I've played plenty of games at a frame rate lower then my refresh and not had this. Im not jumping the bandwagon here and trying to say its something its not.



How exactly is a frame limiter going to help? Microstutter is the video cards delivering the frames unevenly. Basic example is Card A takes 1 sec, Card B takes 1.2 sec, Card C takes 1.1 seconds. At least thats my understanding of it from reading this article --> http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-3.html

I don't know what it does exactly, but it definitely works. It probably buffers the frames and forwards them to the display in a timed manner. This is clearly evident when measuring frametimes. I know of dozens of people using Dxtory to get rid (for the most part) of microstuttering. I have also tried the limiters included with fraps (only working in capturing mode) and nthusim and they work, too. It could be, that you introduce additional input lag, though. It's not a perfect solution obviously, but it is better than nothing and worth to try.

The good thing with dxtory is, that the trial has no restrictions, is free and that all the settings are saved in application profiles. I bought the software for 32 bucks - best money I spent in a while. I wouldn't be too happy with my SLI-rig without it.
 
How do you figure its refresh rate? I've played plenty of games at a frame rate lower then my refresh and not had this. Im not jumping the bandwagon here and trying to say its something its not.

How exactly is a frame limiter going to help? Microstutter is the video cards delivering the frames unevenly. Basic example is Card A takes 1 sec, Card B takes 1.2 sec, Card C takes 1.1 seconds. At least thats my understanding of it from reading this article --> http://www.tomshardware.com/reviews/radeon-geforce-stutter-crossfire,2995-3.html

For example: 45 fps means your monitor gets a new frame every 22.22 msec but it's trying to draw a frame every 16.66 msec.

Code:
Refresh  MsecElapsed  FrameNumber  TimeFrameWasReady  TimeDelayBeforeDraw
0        0            A            0                  0
1        16.66        A            0                  0
2        33.33        B            22.22              11.11
3        50.0         C            44.44              5.55
4        66.66        D            66.66              0
5        83.33        D            Repeated Frame
6        100          E            88.88              11.11
7        116.66       F            111.11             5.55
8        133.33       G            133.33             0
9        150.0        G            Repeated Frame

You end up with a pattern that presents some frames for longer than others and has a variable time delay from when the frame is generated until when it is displayed. This is stuttering.

Having variable delays on top of this where each video card takes a different amount of time to generate the frame would make it worse, but the problem is already there.
 
Last edited:
1. No
2. No
3. Yes

I'm assuming this was for the questions I asked, so no your cards are not overclocked, no you haven't forced +20 in powertune and yes you checked the bios settings are cool.

For shits and giggles can you force +20 powertune for each card in amd overdrive and see if it helps/hurts things. Also I listed a link for the new beta of kombuster, give that a shot and see if you have any voltage or gpu throttling going on. A good on screen display helps when monitoring this. The new beta kombuster has some good stuff on screen like gpu utilization bars for each gpu, temps etc.. The data may be helpful to you.

I don't think the problem has anything to do with the bus choking the cards as from hardocp's findings I'd be surprised if these cards even used the pcie 2.0 8x bus to maximum capacity let alone the x16. The data flowing should fit nicely in the slots allowed bus, hell maybe 2x over.
 
So if we apply microstutter to that, we should get something like this:

Code:
Refresh	msElapsed	Frame#	FrameReadyAt	Delay
0	0		0	N/A		N/A
1	16.66		0	N/A		N/A
2	33.33		1	20		13.33
3	50		1	RepeatFrame	
4	66.666		2	60		6.67
5	83.333		3	80		3.33
6	100		3	RepeatFrame	
7	116.666		3	RepeatFrame	
8	133.333		4	120		13.33
9	150		5	140		10
10	166.667		5	RepeatFrame	
11	183.333		6	180		3.33
12	200		7	200		0
13	216.666		7	RepeatFrame	
14	233.333		7	RepeatFrame	
15	250		8	240		10
16	266.666		9	260		6.66
17	283.333		9	RepeatFrame	
18	300		10	300		0
19	316.666		10	RepeatFrame	
20	333.333		11	320		13.33


Frame rate without microstutter: 50fps
Frame rate with microstutter: 33.3fps
Frames drawn in 20 refreshes: 11
Frames repeated in 20 refreshes: 8
Longest time to display new frame: 50ms - effective frame rate for this frame: 20fps
 
Makes sense, but how do I add a custom refresh rate in CCC? Under Desktop Management> Desktop Properties, it allows me to select a couple different refresh resolutions (57, 60, 65, 70, 72, 75) but when I try something other then 60, it changes the resolution and only allows 60hz at bezel corrected 5760x1200.
 
I'm assuming this was for the questions I asked, so no your cards are not overclocked, no you haven't forced +20 in powertune and yes you checked the bios settings are cool.

For shits and giggles can you force +20 powertune for each card in amd overdrive and see if it helps/hurts things. Also I listed a link for the new beta of kombuster, give that a shot and see if you have any voltage or gpu throttling going on. A good on screen display helps when monitoring this. The new beta kombuster has some good stuff on screen like gpu utilization bars for each gpu, temps etc.. The data may be helpful to you.

I don't think the problem has anything to do with the bus choking the cards as from hardocp's findings I'd be surprised if these cards even used the pcie 2.0 8x bus to maximum capacity let alone the x16. The data flowing should fit nicely in the slots allowed bus, hell maybe 2x over.

Downloading now, will try the +20 on powertune as well. What am I looking at/how do I identify voltage/gpu throttling?
 
Do your monitors actually support more than 60Hz at their native res? Most don't.

Max supported is 75hz at 1920x1200 reported by CCC under My Digital Flat Panels> Properties

Asus Website has this listed under specs (whatever it means): Digital Signal Frequency: 28~83 KHz(H)/ 50~85 Hz(V)
 
Downloading now, will try the +20 on powertune as well. What am I looking at/how do I identify voltage/gpu throttling?

Cool, Your 3d voltage can be easily monitored with msi afterburner or any program that provides a line graph. Make sure your not getting any insane vdroop.
 
Cool, Your 3d voltage can be easily monitored with msi afterburner or any program that provides a line graph. Make sure your not getting any insane vdroop.

ran it for like 20 minutes earlier today and all 3 cards stayed a consistant...1.175v I believe...
 
ran it for like 20 minutes earlier today and all 3 cards stayed a consistant...1.175v I believe...

Did your gpu speeds fluctuate? Also did you try forcing +20 powertune. These are related. Without +20 or at least some kind of increase your cards gpu frequencies can/will fluctuate to keep the driver controlled TDP limit. If your not overclocking then your cards shouldn't fluctuate and should stay at a constant 860mhz, unless there is some kind of bios issue with one or more of your cards. I'd double check this to be sure.

Try running the extreme burn in at the highest quality and resolution settings your system can handle. BTW dont let the warning scare you, it's perfectly safe.
 
Yes, I do have them set to +20 power setting and they all stayed running at 860 the entire time.
 
Yes, I do have them set to +20 power setting and they all stayed running at 860 the entire time.

You should be stable even without +20 at stock speeds. If not there is a problem. Anyway, has the +20 changed anything with the microstuttering in game?
 
Whew, that would definately be some extra coin to drop on 3 120hz monitors lol.

I'm wondering because I have 3x 120hz monitors and experience 0 microstutter. I'm wondering if I would have been SOL if I bought 3 IPS 60hz panels instead. Anyway 120hz is the fluidity I need now. I don't think i'll ever step backwards.
 
Are you using different display connector types to connect to your monitors? e.g. 2 DPs and 1 DVI, 1DP 1 HDMI 1 DVI? That sort of thing?

I saw someone say in another thread that doing this, especially when combining DP and DVI on three different monitors was causing a problem like this. Something about the signal timings between the two and that DVI was somehow slower than DP. Their fix IIRC was to connect their center display on the DVI and the two outside displays on DP, which the theory was that now the DVI would be the master of the three and control the overall timing.

I dunno, it could have been complete bullshit but I thought they said it fixed their problem.
 
Back
Top