I've read up on the subject of crossfire and the many issues that may occur, but no post has offered a solution to my CF issue. I hope your insights may shed some light on the matter, so here goes:
This is my setup:
Core i7 870 @ 3.65ghz HT on
8GB Kingston VR DDR3 @1333mhz
Gigabyte GA-P55-USB3 (x16x4)
2xOCZ Agility 3 in Raid0
2x MSI Cyclone HD6850 OC Edition (925/1100) CCC 12.8/CAP2
Al benchmarks show crossfire is operational and gain +70% in score with both cards enabled (despite having an x16x4 interface), however in just about every game, I get negative scaling. There's 2 games where I actually see an increase in performance: Two Worlds 2 (great scaling!) and Battlefield 3 (but the microstutter pretty much ruins that one). My most recent purchase, Guild Wars 2, loses about 35 fps with crossfire enabled, other players seem to be getting great scaling since the release of 12.8 CAP2.
What makes my CF issues bizarre: Whenever a game takes a performance hit with CF enabled (negative scaling) GPU-Z reports higher GPU usage on the second card than the first card (often a large difference, for instance GW2 reports GPU2: 100%, GPU1: 25%). Searching the net I haven't come across anyone who has had a bigger GPU usage on the slave card than on the master card, so I'm guessing something's wrong.
What I've tried so far:
-Replacing the crossfire bridge
-Re-installing/Rolling back drivers
-Remove second card, reinstall drivers, put slave card back in
-Testing both cards seperately
-Praying (hah hah hah, no.)
I'm running out of ideas, anyone got any suggestions? Thanks In advance!
Crossfire is a nightmare - there's no point in trying to fiddle with crossfire 6850s to get them to run. Sell them both and get a 7870. A factory overclocked 7870 will give you pretty close to 2x the performance of a 6850.
@BigMack70: You make a valid point, this post is literally my last attempt at getting my CF setup to run properly. If this doesn't yield any results it's back to a single GPU.
@ColGeek: I've only recently cranked up the clocks on my GPU & CPU, running at stock speed doesn't make any difference.
I don't think all setups have Microstutter, TriFire and TriSLI both seem to have less issues with it. Also, I'm pretty sensitive to Microstutter, I've noticed it my old HD 4850 crossfire rig, my brothers 6870 CFX rig and my friends GTX 560 Ti SLI rig, as well as my GTX 460 SLI rig. But with my old 6970's I didn't notice it at all.
You have to look at frametimes to know if you have or don't have microstutter. The only site I've seen do this is techreport, and every dual GPU setup they have looked at has microstutter to one degree or another. Sometimes it's minimal, sometimes it is horrible, but it's ALWAYS present to some degree - it's inherent to the technology.
I have not seen any tri- or quad- GPU analysis on this, maybe it does go away with extra GPUs, but I don't think there's been any good analysis done on that recently.
"The performance of three- and four-way CrossFire setups not only surprised us, but also managed to utterly convince us that micro-stuttering doesn't have to affect your multi-GPU configuration. For some reason, the third GPU almost always eliminates micro stuttering and has a less-pronounced effect on performance."
Yeah sorry I should have been more clear - I've seen that Toms article. By recently I meant 7xxx series and 6xx series analysis.
They are so different in 2-card configs (they're much better about microstutter than 6xxx series and 5xx series cards were) that I don't know how they are in triple card configs. Maybe they're even better, I dunno - don't think anyone has tested it.
So over the years Multi-GPU support has improved quite a bit, AMD still isn't up-to snuff at the level of NVIDIA though, multi-GPU supports still literally and directly remains the Achilles heel of ATI's Catalyst drivers. For years now we have been requesting user-based multi-GPU profiles. See when a new game comes out, multi-GPU support will not be supported by the Catalyst driver. ATI will always need to update this through drivers or downloadable profiles. What ATI needs to do is to allow the end user to make custom profiles per game. A small tip, renaming your game-executable towards a game title that is supported can, not always but often, help out." http://www.guru3d.com/articles_pages/radeon_hd_6850_amp...
Reviving an old thread.
Turns out the rumors saying there's no noticable difference between running a x4x16 or x8x8 multi-GPU setup are just plain wrong. At least when it comes to my personal experience.
I bought a new CPU, MoBo and Ram shortly after giving up on getting my old setup to run properly. Currently my setup looks as follows:
Core i7 3820 @ 4.5Ghz
Corsair H80i closed loop water cooler
Asus Sabertooth x79 (PCI-E x16x16x8)
8GB Corsair XMS 7-7-8-20 @ 1666Mhz
2xOCZ Agility 3 in Raid0
2xHD6850 Cyclone Power Edition
Upgrading my motherboard to a full bandwith x16x16 PCI-E beast unlocked the true potential of my 2 HD6850's. I can run anything maxed out without any hiccups, getting great scaling and beating a GTX580 is practically all tests.
So if you're getting unusual bad performance out of your x4x16 Crossfire setup, consider upgrading your motherboard to x8x8 or x16x16 capable one.