Eyefinity Crossfire Stuttering Resolved !

[Haldi]

nah, i did not ^^ But nice to see that he came to the same conclusion as me. means i'm not completly wrong 8)

[Eriol]

For those of you who want to dabble with frame times, there's a great program I bumped into that does nearly everything for you. Fraps Calc. Place it in a directory with a Fraps frame time .csv file and start it. What you get is something like this:



There's a config file for it too, unfortunately I haven't found any readme's and can't read any Russian so don't know the story behind this great piece of software. At least it's fairly self explanatory so you can just try different settings and see what effect they have.

Personally I ditched my 6950CF for a single 7970 so no micro stutter problems for me. I'd still like to see if 7970 CF does have these issues, so if anyone is willing to do some 7970 CF Eyefinity tests and post them here I'd certainly be interested to see them.

[Wijkert]

For those of you who want to dabble with frame times, there's a great program I bumped into that does nearly everything for you. Fraps Calc.

Awesome going to try this program out.

Personally I ditched my 6950CF for a single 7970 so no micro stutter problems for me. I'd still like to see if 7970 CF does have these issues, so if anyone is willing to do some 7970 CF Eyefinity tests and post them here I'd certainly be interested to see them.

Haldi already made some graphs of his 7970 cf setup and the ms seemed a little more pronounced than 6970 cf. Maybe it will get better when new drivers are released, but I don't think it will be gone completely. I will be buying a 7970 and selling my 6950 cf as well.

[Haldi]

There you go ! a Whole lot of unsorted Statistics about HD7970 Crossfire ;) Though one card is running PCI-E 3.0 8x ! and driver are 8.920.0

Attached files

CF PCI-E 8x.7z (4.5 MB)

[Eriol]

There you go ! a Whole lot of unsorted Statistics about HD7970 Crossfire ;) Though one card is running PCI-E 3.0 8x ! and driver are 8.920.0

Thanks. Looks like no improvement whatsoever, how unsurprising. :nudgenudge

Personally I got the best gameplay out of my 6950CF using a framerate cap that would force the frame rendering to be more even, but the problem with that was that I had to set it to minimum FPS, which kinda meant that the difference between single GPU and CF capped was just not worth it.

I wonder what's the state of that smart vsync mentioned in the techreport article
"impact of vsync on jitter is tough to predict; it adds another layer of timing complexity on top of several other such layers. More intriguing is another possibility Nalasco mentioned: a "smarter" version of vsync that presumably controls frame flips with an eye toward ensuring a user perception of fluid motion. We think that approach has potential, but Nalasco was talking only of a future prospect, not a currently implemented technology. He admitted AMD can't say it has "a water-tight solution yet."

I'm not sure how difficult it would be to code something like that to the CCC but it would certainly beat a static FPS cap. Currently vsync just doesn't cut it.

[Redraider89]

well. I cant argue that an i7 outperforms most games when it comes to FPS and when USING NVIDIA cards. did anyone ever notice that ALL the tests that show sandybridge running more frames than AMD all use nvidia GPUS????

Truth is. a) i dont care about hitting 150 fps. I want 60 fps on vsync so that my video is smooth ... like watching HD tv. 2) why dont they do some tests with a AMD system and AMD cards running the same benchmarks??

In answer to your question number 2, I've noticed time after time on sites like Tom's Hardware that they SPECIFICALLY, and EXPRESSLY DON'T use ATI/AMD combinations SPECIFICALLY and EXPRESSLY because that is what AMD recommends. So they intentionally don't do what AMD recommends, under the guise of being "objective" and not biased by doing what AMD recommends and claiming that doing what AMD recommends skews there review as being, supposedly, an AMD spokesperson instead of being "independent", and on and on.... That is why.

But they NEVER state what Intel or Nvidia recommends in their review and NEVER specifically state that they are not going to do what Intel or Nvidia recommends. It's ridiculous and you are a victim of that bias having wasted all that money on an i7 system.

But I'm glad you posted this because I came across your account of this experience trying to decide if I should go Crossfire, or one powerful GPU card and was afraid if I went Crossfire, I would not enjoy the results. Now I know, thanks to your post.

[BigMillerTime]

Hey guys,

First post, signed up to get in on the action! ;)

I also have the Asrock Extreme 7 Gen 3 board with an I52500k. I have 3 6950's unlocked to 6970's and I can't get one game to run right it seems.

I'm not sure what the best plan of attack would be... Ditch the motherboard or switch to NVidia. Perhaps the new 7xxx series would help. I'm confused on what route to take, but this is all B.S. I spent so much $$ to get a bitchin system and the thing performs like shit.

Any advice?

[Haldi]

jep. Install MSI Afterburner. Open RSST (MSI Afterburner symbol with this 60, in tray)
Click on "Settings" and set "Framerate Limit" to something that is between your Min FPS and Avg FPS (near min FPS might be better)

Let's see if that helps!

[Haldi]

News News ! Seems like i have to Redo my statistics with new Drivers :D







Source: http://techreport.com/articles.x/23217/8

[Wijkert]

News News ! Seems like i have to Redo my statistics with new Drivers :D

I am looking forward to your results, but I am sure you realize that the reviewer only used a single card for eyefinity/surround gaming and not two in crossfire/sli.

[Wijkert]

@ Haldi. Did you do any testing since that article has been released?

[Haldi]

Nope! Didn't have a lot time. And now MB is RMA and im running single 6970.

[Haldi]

Okay....

Here some new Results:
8.982.0.0 driver. (was 12.8 IIRC, with CAP2 12.8 installed)

Tried with 1000mhz core and 1425mhz Vram, and 1800mhz Vram!










Also tried with Vsync enabled at 1800mhz.


And with MSI Afterburner FrameLimiter set to 60

[Wijkert]

Okay....

Here some new Results:
8.982.0.0 driver. (was 12.8 IIRC, with CAP2 12.8 installed)

Tried with 1000mhz core and 1425mhz Vram, and 1800mhz Vram!

Also tried with Vsync enabled at 1800mhz.

And with MSI Afterburner FrameLimiter set to 60

Is it possible that you forgot to add the results themselves? Would like to take a look at them.

[Haldi]

Hooray, now i cant edit anymore... ^^ they really should fix this!


Here some new Results:
8.982.0.0 driver. (was 12.8 IIRC, with CAP2 12.8 installed)

Tried with 1000mhz core and 1425mhz Vram, and 1800mhz Vram!


No big difference.





Also tried with Vsync enabled at 1800mhz.




Fail!

And with MSI Afterburner FrameLimiter set to 60

https://www.wsgf.org/f/u/contrib/node_gallery_image/23203/Dirt39825Graph.png

https://www.wsgf.org/f/u/contrib/node_gallery_image/23203/Dirt39825.png


Seems like that Frame Limiter is the only way to prevent Micro Stuttering!
But the question is! Do you want limited 35 FPS all the time instead of sometimes 35fps with micro stutter but then again in another scene 50fps and no problems...

I Asked unwinder if he could modify his frame limiter, he refused. He does have enough to do with the rest of MSI Afterburner.
Not sure if he's allowed to or might agree to release the source code for the Frame limiter, nor do i have the knowlede to create a modified self adjusting version myself :(

[Wijkert]

Hmm, that indeed doesn't look promising. I have played with the frame limiter on and two things bugged me. When you drop below the frame limit, microstuttering seemed worse than without. Also even if the games stays at the max framerate all the time, gameplay isn't as smooth as with v-sync enabled. Until they resolve this issue, I will stay with one card what AMD is concerned. A friend of mine might be buying 2x 660Ti’s in the near future. He might let me borrow them to test how bad the microstuttering is in the green camp.

[Haldi]

phew.... RadeonPro really is a pain in the Ass!
1. Never got it working with any Dx11 game!
2. If you don't reach the set limit you get worse stuttering than witouth! (still have to do some research about that....)

But here you go:

Trine 2


Trine 2 With Dual Vsync 59FPS




Bulletstorm


BulletStorm Dual Vsync 59FPS




Alien versus Predator


Alien versus Predator Dual Vsync 59 FPS


Alien versus Predator Dual Vsync set to 110 FPS (which actually limited to 60 FPS!)



Borderlands 2


Borderlands 2 Dual Vsync 59FPS


Borderlands 2 Dual Vsync (Performance Problems stuck at ca 40 FPS.... maybe wrong Crossfire profile?)






So as you can see RadeonPro and Dual Vsync CAN bring a great improvement (in numbers.... about the real feel there wasn't so much difference... and i didn't really check it.)
but if it's not set correctly its more disastre than improvement!

[Haldi]

Freakin awesome webpage by Benchmark3d!

http://benchmark3d.com/micro-stutter-checker

http://benchmark3d.com/microstutter-cas ... tf-edition

Doing some Tests from Skyrim with only 1 CPU core...

[Wijkert]

Nice find! Had found it a couple of days ago, but forgot to post it here...

There is some discussion about if frametimes recorded via Fraps can be accurately when running multiple gpu's. Using it for a single gpu to plot frame latency (aka smoothness) is pretty much the best way though.
Interesting read: http://techreport.com/blog/24133/as-the ... ng-methods

[Haldi]

No big difference for me!

TESV Threads.png (62.57 KiB) Viewed 5507 times

TESV Single.png (40.03 KiB) Viewed 5495 times

TESV Multi.png (39.13 KiB) Viewed 5513 times