Here is the same test again with lower video card clocks.  This should prove that it isn't an issue of stability such as Freaky deduced.

I am left scratching my head.  I don't have 100% core utilization on one core such as you or SLI utilization yet the memory is still being shared among all three video cards.

I am using the same version of FurMark as you as well.



Compare the loads among the three, the core temperature, and the fan speeds.

Hmm...

"This should prove that it isn't an issue of stability such as Freaky deduced."
But everything that happened is telling me otherwise.
When i started up gpu-z i wasn't getting a load with both gpu's. But i was testing other applications before running gpuz that could of caused my one gpu from becoming unstable. It happens in games where the one gpu just downclocks, sometimes it freezes right up. When i hit control, alt, delete, and shut the application down, then restart the same application up, it stays stuck in 3dmode (400mhz) no higher. It doesn't go into 3d perfromance mode after the freeze up or downclock above until i restart my system.
So i restarted my system after running furmark the first time with no load on my one gpu, After the restart i had a load on both gpu's.
And since i've downclocked to 650 on the core i haven't had one freeze up or downclock in any application that it usually happens in.
The downclock only occurs after i exit the game or application like it should now.
I tried to up the core yesterday to see how far i could go before the downclock or freeze occurs and so far it starts at 670 on the core.
I'm not saying its a problem with everyone who can't run sli in gpuz but it was a problem i was having with downclocking in applications before this.
And changing my overclock has solved the problem.

ty_ger07

Thanks for the input.  I did not mean to doubt your finding Freakysqeeky.  Instead, I meant to say thing in my instance it isn't an issue of instability resulting in the lack of SLI capability in FurMark.

Very odd.

My GPUs all jump up to 3d load speed ("Extra" profile), but only the primary does any work.  The secondaries just sit there idle and share the GPU memory.

No problem
I was going to ask you if you changed the tabs at the bottom of each seperate gpuz for the each gpu's being used, but i can see your temps are all different so you must be reading all of your gpu's seperate with each gpuz application.
I did another test and i'm still using a full thread but no more on the cpu and both gpu's have a load.
I'm not sure what's going on with yours?

TEST: Resident Evil 5 (Benchmark App)

Settings were 1920x1200, 4xAA, Motion Blur ON.
System: (1) Q6600 @3.81GHz, (1) 295 in SLI Mode, and (1) 280 operating in dedicated PhysX mode.


Both sides of my 295 are working hard in this one...
280 is out of the picture...
All 4 cores of my CPU are working well together... Outstanding!

One of you i7 boys give this a run to see if all your cores get used. Check BatMan too if you have it.
I want to know if 4 cores is the max supported, or can it go even higher?

Test: Warmonger (PhysX Game)
1920x1200, all high settings.

System: (1) Q6600 @3.81GHz, (1) 295 in SLI Mode, and (1) 280 operating in dedicated PhysX mode.


Impressed with the system resources that this game uses.

Both sides of my 295 are heavy into the game...
280 PhysX processor has a constant workload... 
All 4 cores of my CPU are helping too.

Outstanding!! 

An old ageia Warmonger link with video: http://tv.hexus.net/show/...r_we_re-examine_PhysX/
I like it... 

A fun app to tinker with:  Voxel Plancton   (A CUDA App...)

http://forums.nvidia.com/.dex.php?showtopic=153257



It uses at least 4 cores on my CPU...
Very light use on 1/2 of my 295. (Under 7%)
High utilization on the second 1/2... (Especially with forcing V-Sync off in the Nvidia Control Panel.)
No processing is done on a PhysX GPU.

A charming little app... 

BTW - I think it's worth noting that in FluidMark, the Nvidia Fluid Demo, and PhysX Fluids by Kenneth Bugeja, that no matter what res I ran at, the PhysX workload remained the same. This is probably due to it moving the exact same number of particles around in the simulation, regardless of the screen resolution your running at.

So new workload lesson: A PhysX Simulation's workload on the system, is not resolution dependent. 

TEST: CUDA-Z's Perfprmance Test. (Version 0.5.95)
When pointing at 1/2 of my 295...


My 295 reports an intermittent 43% utilization.



When pointing at my 280:


My 280 reports an intermittent 55% utilization.

Look at CUDA-Z go, using all 4 Cores of my CPU just to check my GPU's performance!