Crossfire performance not up to par?

[Terronium 12]

So I'll jump right into it: I'm well aware that Crossfire doesn't actually double the performance (at least not in every game) but the performance I've been getting in some games doesn't seen to be at all acceptable. I ran the in-game benchmarks for F1 2011 and DiRT 3 both with crossfire disabled and enabled, details as high as they can possibly go, and here are the results:

CrossfireX Disabled - DCII clocked at 810/1250MHz

F1 2011 - Avg. 51 / Min. 45
DiRT 3 - Avg. 56 / Min. 47

DCII Clocked at 850/1350MHz

F1 2011 - Avg. 52 / Min. 47
DiRT 3 - Avg. 58 / Min. 48

For the next round it's enabled and both cores will now be running at 820/1250MHz as the Sapphire card runs 10MHz faster than the DCII at stock; both games running under their respective profiles, and as DiRT 3 seems to have a handful of them I used the dirt3*.exe profile.

CrossfireX Enabled - clocked at 820/1250MHz

F1 2011 - Avg. 73 / Min. 61
DiRT 3 - Avg. 104 / Min. 90

Clocked at 850/1350MHz

F1 2011 - Avg. 73 / Min. 61
DiRT 3 - Avg. 108/ Min. 92

F1 2011 annoys me the most because the performance gain isn't even as much as the gain seen in DiRT 3 which more or less does in fact double over using one 6950. Both games use the same engine but it seems DiRT 3 is a bit more optimized as far as the engine goes. I've more games on hand, most of which aren't installed but I'm wondering if this is at all normal.

I don't remember my performance with a single card in Crysis 2, but with Crossfire I'm seeing anywhere from 80-100+, and this is with mods.

 

[Solarstarshines]

Are you running the games with V-sync on if so there is your problem

 

[Terronium 12]

What am I, a noob?

Joking aside, V-sync is off. I always leave it off.

 

[smarteyeball]

Are you running the games with V-sync on if so there is your problem

F1 2011 - Avg. 73 / Min. 61
DiRT 3 - Avg. 104 / Min. 90

Clocked at 850/1350MHz

F1 2011 - Avg. 73 / Min. 61
DiRT 3 - Avg. 108/ Min. 92

I'd say no.

Performance looks pretty typical. Some games are simply better suited to multi GPU.

Driver optimizations also make a difference too.

You can use GPU-Z log file or Afterburner to check the GPU usage of each card. At your current CPU clock speed, it's certainly not a bottleneck.

When I had my 5970, for F1 2010 the CF 'fix' was to disable a core

 

[Solarstarshines]

then that is pretty odd

are you getting anything else besides low frames ? artifacts

 

[Solarstarshines]

Are you running the games with V-sync on if so there is your problem

F1 2011 - Avg. 73 / Min. 61
DiRT 3 - Avg. 104 / Min. 90

Clocked at 850/1350MHz

F1 2011 - Avg. 73 / Min. 61
DiRT 3 - Avg. 108/ Min. 92

I'd say no.

Performance looks pretty typical. Some games are simply better suited to multi GPU.

Driver optimizations also make a difference too.

You can use GPU-Z log file or Afterburner to check the GPU usage of each card. At your current CPU clock speed, it's certainly not a bottleneck.

I would say the same but the F1 is a very low power game they ran it on intel 3000 hd so im assuming he should be over 200fps with that game if the intel hd 3000 made it run smooth or look smooth at high settings

 

[Terronium 12]

No artifacts. Just doesn't really seem to like Crossfire all that much.

 

[Solarstarshines]

you can run Unigene and see if your cards are running correctly if your score is comparable to a previous score you can chalk it up to the game

 

[Terronium 12]

Performance there is pretty high, same goes for 3DM 11.

Although with the latter I was running bench-only clocks so it's not really comparable.

 

[smarteyeball]

you can run Unigene and see if your cards are running correctly if your score is comparable to a previous score you can chalk it up to the game

Good idea

F1 can be pretty light on resources - until you start cranking up the details

Performance there is pretty high, same goes for 3DM 11.

Although with the latter I was running bench-only clocks so it's not really comparable.

Bench only clocks aside, it does sound like it's more game/driver related.

(bench only clocks would only nab you an extra 5-10FPS or so anyway in real world usage).

That's the 'joy' of multi GPU setups. Mega boosts when all is well - bugger all difference if it's not properly supported.

Aside from the year I had the single 580, the rest of time I've had CF setups and I know the frustration when things don't perform as well as you'd like them too. You tend feel a little cheated.


At least the scores went up and not down

 

[Terronium 12]

Just ran Crysis for the lulz and while it's incredibly smooth it's only 50-60+ in most scenarios, and up to 130+ when I'm standing atop a hut or something. With the mods I just installed, it's knocked down to 30-40 in heavily populated areas (palm trees, water physics, etc, etc.) 50-70+ where it isn't, and up to 130+ under the same scenario (atop a hut or something).

Of course I wasn't expecting 80-90+ frames 24/7 but it's nice to know Crysis is going to be here nine years from now still not giving up such frames.

 

[smarteyeball]

Lol