• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Underperforming(?) crossfire eyefinity rig - any suggestions?

W

wuLFe

n00b
Joined
Mar 7, 2011
Messages
8
Hey folks

I've been lurking on the [H] forums for a while, and was recently inspired to do a fairly serious upgrade - however, when checking benchmarks, etc, it seems that I'm getting a fraction of what I should be getting, and was hoping that someone could give me some suggestions on what the problem could be.

Firstly, the (relevant) parts of my setup:
PSU : 1000W Rosewill (forget the model number - bronze 80+ certified one)
CPU : 3.4 GHZ (stock) AMD Phenom X4 965 BE
RAM : 8GB DDR3
GPU : 2 * 6950, shader unlocked running @ 1.225V, 930 core, 1430 mem

OK - so I recently added the second unlocked 6950, and have just added two new 1920X1200 monitors (Asus VW266H), and I've been doing some benchmarks at 5782 * 1200 res - the numbers don't seem to be even in the same ballpark as other benchmarks I am seeing, and it is making me quite concerned, since, as y'all know, we're not talking chump change when we're looking at eyefinity/crossfire/etc.

The one that has me most concerned is Far Cry 2 - on the widescreengaming forum, at 5760 x 1080 with crossfire 6970s, they are pulling out what I assume to be an average of 85ish fps
http://www.widescreengamingforum.com/wiki/AMD_Radeon_6970_6950_-_Featured_Review_Page_6

When I benchmarked the same stage with the same settings last night, I got a minimum of 28, average 37, and a max of around 47 - so I'm less than half what I (assumedly) should be getting, and I'm tearing my hair out in frustration here.

if it helps, I'm using the latest official AMD drivers, have enabled crossfire in the CCC, and have overclocked the card using Afterburner v 2.1.0 beta 7 (I think - posting from work, so don't have all the relevant info available).

HELP!!! :(
 
Try to get benchmarks from a different program or game so you can be sure it isn't an error within FC2. Also try benchmarking with one monitor, as there will be more benchmarks for you to compare.
 
I'd be careful with that Rosewill unit. There is a loophole in the industry to get 80Plus certification on units that can't produce the maximum power required at all, let alone at good efficiency, so beware of that.

I wouldn't worry about those benches, they are very different to the norm for eyefinity results (far better than you would ever actually experience). min of 28, average of 37 is about right for that setup.
 
You are comparing apples to oranges.

Those benchmarks you linked are using a Intel i7 920 and you are using a x4 965 BE. You are severely bottlenecked I believe. Now I am not hating on AMD,I am running the exact same CPU. If I read your original post correctly you are running two 6950s bio flashed (unlocked to 6970). Basically a 6970 crossfire setup right? There are a few test/reviews floating around the net showing the 6970 xfire setup is HEAVILY bottlenecked on a x4 965 BE. My math (and memory) may be off but I believe your CPU is bottlenecking your GPU setup by about 75%. I won't link to the reviews on other sites as that is a little in poor taste generally. But google is your friend.

I hope this info helps.

Disclaimer: If any of my info is wrong I apologize. I am fairly certain the above info is in the ball park of answering the issue at hand though.

Better question though. Will two 6950s @ stock be bottlenecked by a x4 965? I believe the answer is yes but anyone with first hand experience please share. Even better question then the last. Will two 6950s @ stock bring up the min. FPS as to still allow for a better overall gaming experience through less dips in frame rate? We can call it a higher basement on FPS.Say the O.P. planned to upgrade to Zambezi and a AM3+ mobo later on would he be wasting his money to keep both cpus in xfire until later this year? Or should he just return one card,or both for a HD 6970 or gtx 570/580?

I know at that resolution (1920 x 1200) neither of the single enthusiast high end cards will be bottlenecked by the CPU.

And to O.P. I have the exact same LCD as well. I love it.
 
Well I am a dummy, I misread the O.P.s statement about using eyefinity but I am pretty certain he is still highly bottlenecked by the CPU. Just not as much at that high of a resolution. Maybe 50%?
 
Hey guys - thanks for the suggestions so far - I did some more testing when I got home from work, and here's what I came up with:

I ran a single monitor setup in 1920*1200 res, and used Afterburner to check the GPU utilization, and at no time did either of the GPUs pass 65% utilization - so I guess, at least on a low res, my CPU might be the bottleneck. I know it's not the greatest CPU, but I'm still surprised that it would result in a roughly 50% decrease in FPS vs an i7.

I'm not quite ready to upgrade yet (waiting for bulldozer to be released and then will decide between that or sandy bridge), but I'm still puzzled at these results. I'm at a friends house right now, but when I get back, I will try some other benchmarks to see if my results are more in keeping with the published results, and will post again then.

What is the best program to use to see what my CPU utilization is - perhaps using that in conjunction with afterburner for the GPU figures will shed some enlightenment on my predicament.
 
On a single display, yes the CPU will be a bottleneck, guaranteed.
Windows task manager is good enough to show CPU usage.

People don't really realise what the difference is between AMD CPUs and modern Intel CPUs.
My 3.4Ghz Q9550 was 10-15% faster than your stock 965, my 4.1Ghz i5 is 60-70% faster than it, and my friend's 4.9Ghz i7 2600K is more than twice as fast as your 965. CPUs have come on a long way since the Phenom II,or at least they have with Intel. AMD CPUs are great value, but they are cheap for a reason!
 
I'm also running three VW266H's in Eyefinity, my system (see sig) is a bit more modern, though I don't have my cards overclocked, just unlocked. I also have my CPU running at stock right now due to some pesky voltage issues I seem to be having that are BSOD'ing me.

I'll run some benches and see what I get.

EDIT: Here are a few from Just Cause 2 from last month, haven't had time to run anything new.

http://cloud.steampowered.com/ugc/594678513564339965/96FA1D469EB821DE6447437842AD127A2250F9CF/ (Concrete Jungle Benchmark)

http://cloud.steampowered.com/ugc/594678513564338073/D48DD7C64ACFAADEB65D59E0232DFA6488D5A28C/ (Desert Sunrise Benchmark)

http://cloud.steampowered.com/ugc/594678513564346719/41C5DE6BFBB60D679348F4DCAEF9A6B97FD08486/ (Dark Tower Bechmark)
 
Last edited:
OK - did some more benchmarks and found somebody else who was running roughly the same setup as me, and who was getting very similar numbers - as such, I guess my only options are
a) Live with it, or
b) Upgrade

I'm tending towards option b, and am looking into the i7 2600 - seems to be about the best CPU out at the moment. I had no idea that the CPU could make such a difference for gaming :(

Thanks for all the help - will post back if/when I have the new CPU running to compare the results.
 
if you really want to test if its the cpu bottlenecking the cards just raise your cpu voltage to 1.45v then change the multiplier to 18x(3.6Ghz) or 19x(3.8Ghz) should be able to hit either of those at 1.45v(if 19x fails try 1.5v) even if you are running the stock cooler the temps should be low enough to test it in a few games just to make sure its not the cpu. 3.6Ghz seems to be the cut off on phenom II's where they no longer become a bottleneck. even with eyefinity though unlikely it could still be bottlenecking the cards.

note do not use these voltages as a stable overclock. these are just voltage settings i know that will work with that overclock to test with but they are much higher voltages then they need to be for those given overclocks.
 
O.P you could also just repeat the original benchmarks you ran with 1 card. Compare the crossfire FPS to the single card FPS. Then you can see how much it is scaling. I decided to just run one card with my setup until Zambezi (bulldozer) comes out and then to crossfire later. But then again I am only running one monitor and not Eyefinity.

I have to say Sandybridge is a very tempting option as I have been looking at that as well. I still think I am going to wait for Zambezi and then decide though. Your decision is much tougher already invested into multi LCD setup with crossfire. I would suggest going with the i7 2600k if you just can't wait. You are in a tough situation and I don't envy you. You have to base your decision on how much expendable income you have and that is what it all comes down to.
 
The i5 2500K will do fine for gaming, but if you'll see the benefits of the 2600K then by all means go for it.
 
I did try running a single card and the crossfire, and it did scale, but not by much (we're talk maybe 70 fps avg at 1920 x 1200 w/ 1 gpu, and 90 fps with 2/ gpus). I know that my current mobo is only 8x/8x which might also be a slightly limiting factor, so I definitely think that an upgrade might not be a bad idea anyhow.

I will probably try overclocking the Phenom tonight just to see if that makes a difference .Haven't fiddled much with CPU overclocking since my Pentium II days, so it should be interesting if nothing else :)

Right now, I'm tempted to pick up th Gigabyte P67-UD7 board (future proof myself in case I ever want to add a third GPU), and the I7-2600 - quite a chunk of change, but I got a decent tax rebate back, and I really didn't need that unnecessary life-preserving surgery anyhow :cool:

I guess I have a few weeks to fiddle around with everything until those P67-UD7 B3 rev boards start showing up at retail - I was considering the i5 as well, but honestly, for $100 extra, I think I'll just go for broke (LITERALLY!)
 
the AMD is most definitely the bottleneck here. If you cant spare the extra $100 for a 2600k I would recommend a 2500k, I just picked one up for my brothers new computer (he is stepping up from a q6600)
 
wuLFe said:
I did try running a single card and the crossfire, and it did scale, but not by much (we're talk maybe 70 fps avg at 1920 x 1200 w/ 1 gpu, and 90 fps with 2/ gpus). I know that my current mobo is only 8x/8x which might also be a slightly limiting factor, so I definitely think that an upgrade might not be a bad idea anyhow.

I will probably try overclocking the Phenom tonight just to see if that makes a difference .Haven't fiddled much with CPU overclocking since my Pentium II days, so it should be interesting if nothing else :)

Right now, I'm tempted to pick up th Gigabyte P67-UD7 board (future proof myself in case I ever want to add a third GPU), and the I7-2600 - quite a chunk of change, but I got a decent tax rebate back, and I really didn't need that unnecessary life-preserving surgery anyhow :cool:

I guess I have a few weeks to fiddle around with everything until those P67-UD7 B3 rev boards start showing up at retail - I was considering the i5 as well, but honestly, for $100 extra, I think I'll just go for broke (LITERALLY!)
Click to expand...


Two 6970s should scale much better than 70 FPS (1 gpu) up to 90FPS (2 gpu). Crossfire scaling is superb in the 6900s. You should be scaling around 190%ish. In theory (perfect scenario) you should be getting about 133 FPS if there is no bottleneck. If my math is right (probably isn't) my original estimate of 75% bottleneck was in the ball park.
 
You must log in or register to reply here.
Back
Top