oddity of 295x2 benchmark in DAI, need help

So I recently got a 295x2 at $550, upgraded my system as follows (no OC atm):

i7-4790K
12G DDR3 Ram (2x4GB+2x2GB)
EVGA GS1050 PSU
MSI Gaming 5 Mobo
295x2 (beta driver 15.5)
Windows 8.1 64bit
Acer LED 1920x1080

Click to expand...

Dragon Age: Inquisition (DAI) has an build-in benchmark tool, so I tried it.

The first run, it gave 110FPS (1080p), subsequent run, however, gave around 68 FPS.

I tried a few times, it seems I will only get 110 FPS if I reboot the computer and run the game fresh for the 1st time in that windows session. In any other situation (run benchmark 2nd, third, fourth times, log out then back into windows and run the game, etc), it gave about 68FPS.

PS. I get about 16k in 3dmark firestrike, and 94FPS in Dirt Rally 0.5 (used to get 120 FPS before its most recent update).

Any clue as to which number is true (110 or 68 FPS) and is there any problem with my system that causing 295x2 to miss its potential?

Thanks

Welcome to the world of crossfire my friend

Thanks, so you are saying this game uses two GPU only when system freshly booted and revert to one GPU quickly?

DAI is funny.

My CF system runs DAI fine, but it will crash in the benchmark only. Have close to 80 hours in the game no crash, never figured out why.

The razor software is incompatible with DAI also.

Running Mantle or DX11?

Make sure you have nothing running in the background. On an OSD put your CPU usage, both GPU usage, and GPU temps. Try to see if everything is running well.

I am running DX11, as mantle has some artifact showing.

What kind of OSD program would you recommend and what are the parameters I should be look for to see if everything is okay?

Thanks

okies, installed afterburner, here is some interesting thing I found.

When system freshly booted, DAI's benchmark uses both GPU, both were used at 100%, resulted in 110FPS.

Any subsequent benchmark within the same windows session, still uses both GPU, but NOT to 100%, more or less around 50%.

So 1: both GPU are used, xf is working; 2: only first benchmark after system reboot is truly a "benchmark", as subsequent ones don't use 100% of GPU.

Not sure why though.

the DA:I benchmark should be a far cry of how the cards perform since the benchmark is all of 1 minute long.
I would play the game with that shiny card you have.

oh no, its more like 10 seconds, but results were quite consistent and in line with independent benchmark using longer test period of the game.

I retested 0.5 version of Dirt Rally again, both GPU were not fully utilized all the time, it varies from 20% to 100%.

I guess I figured all gaming benchmarks will always push 100% GPU, apparently not true at all.