4890 Crossfire not giving any noticeable improvement.

[Sekrab]

I have just started running 2 x 4890's in crossfire and I have to say I haven't noticed any difference at all in the fps in my games. I have tested this by disabling crossfire in the control center and then checking back only to see the exact same frame rates. It literally has zero effect on my games. The only difference I have noticed is a fairly substantial increase in my 3DMark06 score for the SM3.0/HDR test.

Someone suggested to me that I may have a bottleneck somewhere in my system, power or CPU or other... but really? to the extent that it entirely removes any gain through the second card? please, what do you guys think?

I am running LOTRO, Empire: Total War, and ArmA 2 in 1920x1200 x4 AA on the following platform, and not seeing *any* difference with or without the second card.

Intel Core 2 Duo E6750 2.66GHz
ASUS P5Q PRO Turbo Motherboard
4GB OCZ PC2-8500C5 1066MHz
2 x XFX ATI Radeon HD 4890 XTX XXX Edition 1024MB
Corsair 650W PSU

 

[Sekrab]

Thanks for the quick replies, my CPU isn't clocked, maybe I will take a look at that... but you say I need a quad CPU to really get the benefit of a dual-gfx setup?

I was under the impression that 2 out of the 4 cores in a quad did bugger all in the current generation of games... ?


magicroundabout: I am gaming at 1920 x 1200 wi 4xAA

 

[Sekrab]

Are you running with vsync? Get fraps out and make sure there's no improvement.

I am using FRAPS in a series of games some with vsync, others without. I am definitely seeing a difference of exactly ZERO frames with or without crossfire enabled.

I am a bit skeptical about needing a quad... I will have a look at o/c'ing my dual-core tonight.

 

[Sekrab]

Lotro is an odd one, I play it and I can't seem to get great performance out of it full stop, while it says it supports multicore, on a quad it never uses more than 25% of cpu power in task manager, which suggests it really only uses one core and thats why in the heavy populated area's, 21st hall these days, it grinds to a halt due to lack of extra cpu juice. Would be fantastic if they actually made it work with all 4 cores.

Come to think on it... I haven't seen a performance improvement since I was running with one 4870. Running on DX10 all Ultra it just chugs in some areas and sings in others. Maybe just the game like you say...

I would also say though that 2.66Ghz isn't that high, and considering 3-3.5Ghz is basically no sweat for most C2D's I'd overclock it to at least 3Ghz anyway, you can probably hit 3.3-3.5Ghz without too much of a problem and little to no extra voltage and at that point you'll find very very few games (bar the odd ultra detailed RTS/Sim, so Empire, Sup Com, but not C&C or Halo's RTS) cpu limited.

Thanks, that's some very helpful information

 

[Sekrab]

Ok, I overclocked to 3.2MHz last night and saw an immediate and substantial increase in fps. Happy.

However, when I disable or enable Crossfire I see no change in frames...

One thing I noticed was that in GPU-Z my Bus Interface for both adapters is only showing as PCI-E 2.0 x16 @ x8 2.0.

Surely this should read as PCI-E 2.0 x16 @ x16 2.0. as both the cards and the motherboard are rated to x16...

Does anyone know how I change this to x16 or what could be causing the trouble?

 

[Sekrab]

Is there a huge difference between x8 and x16 ?

 

[Sekrab]

I think you might be right about the chip, I was blown away by the effect of raising from 2.66 to 3.2. My mobo has a software o/c'ing utility that allows me to change the clocks on the fly, so I would alt+tab in and out of my game to see the effects the increased chip speed would have. I was getting a solid 20% - 25% increase!!!

I really didn't think CPU had much to do with games any more, not like in the old days where it really counted. Thought CPU muscle was for enthusiasts and it was more to do with bragging rights, etc... and thought that by far the biggest gains came from the graphics card. I was wrong.

I may well have to look at getting that as my next upgrade... very dissapointed overall with my 2 x 4890's so far.

Also; on a slightly unrelated note - what do you think to me selling both my 4890s and going for a single 5870? do we yet have any idea whether performance is likely to be competative? (and yeah, I do know there is a hundred page thread bobbing up and down here)

 

[Sekrab]

Fire-up GPU-Z and get it to keep logging the sensors page in the background....then play a game for a few minutes....then alt-tab back to GPU-Z and see how much each GPU was working.

I already used Rivatuner to confirm this... both cores are definitely active when I enable Crossfire.

i've 2 4870 512mb in crossfire and overclocked (820x1050) they hit over 27k in 3d mark06.

a single 5870 is hitting over 25k in the same benching tool.

unsure what your hitting with yours so maybe a comparison is needed?

Hmmm... I have something seriously wrong with my system, running with crossfire enabled and my cpu clocked to 3.2 I get a measly 15343 marks in 3D Mark 06.

SM 2.0 - 6513
SM 3.0 - 9256
CPU - 2721

Are all 3 scores terribly low or is it the CPU score that is letting me down?

 

[Sekrab]

What about the other 2 scores?

I just ran it again, this time with Crossfire turned off and these are my scores:-


14114 3DMarks

SM2.0Score: 6357

HDR/SM3.0 Score: 7497

CPU Score: 2757

I am getting a difference of 1229 marks for my £150 spent on the additional card. Ouch, just ouch...

 

[Sekrab]

--

 

[Sekrab]

Thanks but I've tried them already, I've actually rolled back to 9.7 after I found that both 9.8 and 9.9 cause my main game (LOTRO) to hang regularly. Apparently a fairly common problem with those revisions and that game...

http://forums.lotro.com/showthread.php?t=290745

I'm thinking of doing a full reinstall as it has been over a year since the last and my config has changed once or twice since then. Good excuse to run Windows 7 as well...

 

[Sekrab]

Ok, I have done a full format and reinstall of my system, and I'm now running the 9.9 Catalysts (the hell with LOTRO while I troubleshoot ) and my E6750 2.66MHz CPU is clocked to 3.2MHz...

So I run the 3DMark Vantage program as suggested by Randell.

...and here are my results:-


P10548
CPU 5424
GFX 15395


My question: Is that GFX score vaguely in line with a crossfired 4890 on a CPU limited system such as mine? and just how badly is my system hamstrung by this CPU score?

 

[Sekrab]

That looks a lot better.
To give you a comparison, my single 4870 was hitting about 8-9k with a quad running higher clocks (crossfire likes clock speed).
I'd say you're still a bit limited by the looks of it; but it seems your computer is now actually benefitting from the second card some, AFAIK a single 4890 usually hits around 9-10k GPU score, so you're actually getting something from the second card in vantage now

Whack on a game or two, and see whether your actual game results seem better

Right, I think I'm ok now, I bought me an aftermarket cooler and I'm now running the E6750 stably at 3.4MHz. I fired up Empire: Total War and I now get a difference of around 20fps with/without crossfire enabled. Thanks for all your help with this guys, much appreciated

One final question: I was reading some guys thread over at the ATi forums saying that a 650W PSU will not even cope with Crossfired 4870s and that I could well be looking at an expensive paperweight in a couple of months from now. Any truth to this?

 

[Sekrab]

Yeah, I think you're right about my game collection. Ironically, I just picked up Aion last night thinking "Ahhh great! this title utilises the CryEngine, a perfect opportunity to flex me new 4890s" ...only to discover that it has known issues with Crossfire!!! Not having much luck so far, am I? hehehe...

My PSU is this one ~ will it cut it?

 

[Sekrab]

Thanks, that's a comfort to know.