Crossfire making no difference with dual 5870

[jtvang125]

Anandtech's review of the 5870 shows Resident Evil 5 doing 105fps @ 1920x1200. I'm getting 48fps on the fixed benchmark at 1920x1200. I know my cpu isn't as fast as a i7 (Q8400 @ 3.5ghz) but it shouldn't be that big of a bottleneck. Turning on CF doesn't do squat either, maybe 1 or 2 fps difference. The ati crossfire logo did come on when I was doing the tests. I should be destroying this game with a 5870CF setup.

The only mistake I can see was that I didn't uninstall and clean out all of the ati drivers before installing the card. However I did go back and do this but made little to no difference. Here's the specs of my rig.

Win 7 pro 64bit
Asus p5q pro turbo
q8400 @ 3.5ghz
2x 2gb ddr2-1066
2 XFX HD5870
antec 750w ps

 

[cusideabelincoln]

Have you checked anti aliasing settings (in CCC) and V-sync?

 

[Jacky60]

Don't know how well Resident Evil uses quad cores but this looks CPU bound. In some games an E8400 doesn't cut it any more as I found out when I replaced my E8400 at 4GHZ with i7 4ghz. In Arma 2 (a hog of a game the difference was a playable game vs a headache inducing slideshow. Try dropping resolution to 1024/768 then run benchmark with crossfire acive and deactivated-If performance jumps you're cpu bound.

 

[Jacky60]

Pretty sure it is cpu limitation see link-your fps is about right and won't double with 5870x fire. You really need a quad core cpu to match you 5870 xfire -maybe same generation quad ccore eg Q9550 as they're pretty cheap these days-and will fit mobo nop probs.


http://www.pcgameshardware.com/aid,...nom-strong-Update-Lynnfield-results/Practice/

 

[toyota]

Pretty sure it is cpu limitation see link-your fps is about right and won't double with 5870x fire. You really need a quad core cpu to match you 5870 xfire -maybe same generation quad ccore eg Q9550 as they're pretty cheap these days-and will fit mobo nop probs.


http://www.pcgameshardware.com/aid,...nom-strong-Update-Lynnfield-results/Practice/

he has a quad core. the Q8400 at 3.5 certainly would not be much of a limitation with a 5870 crossfire setup.

 

[NoM8s]

I get an average of a 38% increase in CrossFire with my VaporX HD5870's in BC2

 

[RussianSensation]

he has a quad core. the Q8400 at 3.5 certainly would not be much of a limitation with a 5870 crossfire setup.

I have personally tested Resident Evil 5 many times. This particular game/benchmark is very CPU dependent.

DX9 Fixed Bench:

1280x1024 4AA
Q6600 @ 3.4ghz + 4890 = 60.8 fps
Core i7 @ 3.9ghz + 4890 = 111.5 fps (+83&#37

1920x1080 4AA
Q6600 @ 3.4ghz + 4890 = 55.1 fps
Core i7 @ 3.9ghz + 4890 = 82.1 fps (+49%)

 

[betasub]

Nice data, Russian!

Looks like even 2x5870 aren't going to get spectacular results on a benchmark that is so CPU-limited (Russian's OC'd Q6600 should be somewhat comparable to OP's OC'd Q8400).

 

[lopri]

My numbers almost mirrored AT numbers.

http://www.anandtech.com/show/2848/10



E8400 + GTX 280



955BE + HD 5870



So from my testing i7-920 = E8400 = 955BE @2560x1600. But that aside, 48 FPS @1920x1200 using two 5870s can't be right regardless of his CPU. Let's help him fix his issue correctly. I'd first take one card out and check the number. When both cards are installed, make sure both cards are running at PCIe x8 using GPU-Z.

 

[Kenmitch]

OP you did install the cross-fire profiles also? Didn't see any mention of it one way or the other.

 

[Matrices]

Regardless of what the specific problem here might be, what is up with all these people going out and pairing $1,000 of GPU with $100 of CPU? We need to put out a PSA on all the geek forums: pro-tip, don't pair C2D, or even C2Q, with ATI 58xx or GTX 4xx, and certainly not multi-GPU.

 

[dug777]

Regardless of what the specific problem here might be, what is up with all these people going out and pairing $1,000 of GPU with $100 of CPU? We need to put out a PSA on all the geek forums: pro-tip, don't pair C2D, or even C2Q, with ATI 58xx or GTX 4xx, and certainly not multi-GPU.

I haven't seen many actual real world instances where my Q6600 at 3.2 or 3.4 wouldn't be enough to keep my 5850 well fed at 19x12...and even at 1000/1225 I see most agreeable and realisable improvements in actual gameplay.

Granted, I don't play GTA4 much, but that seems to just be a crap port in the scheme of things...

 

[hi87]

Regardless of what the specific problem here might be, what is up with all these people going out and pairing $1,000 of GPU with $100 of CPU? We need to put out a PSA on all the geek forums: pro-tip, don't pair C2D, or even C2Q, with ATI 58xx or GTX 4xx, and certainly not multi-GPU.

Even C2Qs huh? Even though Yorkfields are faster clock for clock than PIIs and clocks just as high. That means for multi GPUs we shouldn't use AMDs either.

 

[Matrices]

C2Q bottlenecks GTX 280 Tri-SLI, so it will definitely bottleneck GTX 4xx SLI or ATI 58xx CF.

See here: http://www.guru3d.com/article/core-i7-multigpu-sli-crossfire-game-performance-review/19

A single new gen card is probably fine with an OC'd C2Q, though.

 

[lopri]

Why is everyone so obsessed in asserting their opinions, instead of trying to help the original poster's problem?

 

[Matrices]

He hasn't even responded since his original post, so how do you propose we help - read his mind?

The obvious things to try are: make sure each card works fine by itself, and test CF scaling in other games.

 

[Jacky60]

As I said originally he's cpu bottlenecked. When my 5970/5870 trifire system has i7 920 at 2.66ghz Arma 2 benchmark =42fps. When at 4ghz its 57fps. C2Q especially Q8400 are poor performers compared to I7 in most games and certainly can't feed 5870 crossfire sufficiently given a really demanding game.

 

[lopri]

As I said originally he's cpu bottlenecked. When my 5970/5870 trifire system has i7 920 at 2.66ghz Arma 2 benchmark =42fps. When at 4ghz its 57fps. C2Q especially Q8400 are poor performers compared to I7 in most games and certainly can't feed 5870 crossfire sufficiently given a really demanding game.

Sigh..

 

[lopri]

How does that explain he's getting much less performance with 2 cards than with 1 card?

 

[TurtleCrusher]

Regardless of what the specific problem here might be, what is up with all these people going out and pairing $1,000 of GPU with $100 of CPU? We need to put out a PSA on all the geek forums: pro-tip, don't pair C2D, or even C2Q, with ATI 58xx or GTX 4xx, and certainly not multi-GPU.

You have no idea what you are talking about. Yorkfield C2Q's are far faster clock for clock. I had a thread here a LONG time ago comparing my Q8200 @ 3ghz to a Phenom2 940 @ 3.7ghz. They were comparable while both of us had a 4830 512mb.

Also, the Q8400 isn't a $100 processor. It came out at $219. It's still quite capable at stock speeds.

 

[Matrices]

You have no idea what you are talking about. Yorkfield C2Q's are far faster clock for clock. I had a thread here a LONG time ago comparing my Q8200 @ 3ghz to a Phenom2 940 @ 3.7ghz. They were comparable while both of us had a 4830 512mb.

Also, the Q8400 isn't a $100 processor. It came out at $219. It's still quite capable at stock speeds.

Keep telling yourself that.

Every day I see on [H] some poster complaining that his ancient C2D or C2Q processor isn't keeping up with his 5870 CF or 5970, or whatever other insanely expensive GPU setup.

Again, read the Guru3D link and weep. The C2D/Q the architecture is obsolete, and a setup with powerful GPUs is bottlenecked by that architecture. Period.

You just make yourself look silly when you sit there and bring up Phenom (I was talking about i5 and i7), an ATI 5830 (when we are talking about high-end multi-GPU setups), and state the MSRP of Q8400 when it first came out, as if that somehow contradicts my statement that you can get those CPUs for cheap here and now.

Anyway, none of this is immediately relevant to the OP's situation. Something is clearly broken with his CF setup.

 

[lopri]

I guess it's just the way it is. ^_^ Good luck, OP.

 

[aigomorla]

are u runninng benchmarks on full screen or widowed?

xfire ad sli does not work unless ur in full screen..

or is there a fix for this??? cuz i could never get them to work to either platform.

 

[happy medium]

Anandtech's review of the 5870 shows Resident Evil 5 doing 105fps @ 1920x1200. I'm getting 48fps on the fixed benchmark at 1920x1200. I know my cpu isn't as fast as a i7 (Q8400 @ 3.5ghz) but it shouldn't be that big of a bottleneck. Turning on CF doesn't do squat either, maybe 1 or 2 fps difference. The ati crossfire logo did come on when I was doing the tests. I should be destroying this game with a 5870CF setup.

The only mistake I can see was that I didn't uninstall and clean out all of the ati drivers before installing the card. However I did go back and do this but made little to no difference. Here's the specs of my rig.

Win 7 pro 64bit
Asus p5q pro turbo
q8400 @ 3.5ghz
2x 2gb ddr2-1066
2 XFX HD5870
antec 750w ps

There is no way a single 5870 is bottlenecked by a q8400 @ 3.5 .
Something is definitly wrong here.

 

[aigomorla]

There is no way a single 5870 is bottlenecked by a q8400 @ 3.5 .
Something is definitly wrong here.

yeah his cards arent scaling from what he said.

which means..

1. He's running the programs in windowed mode and not full screen mode.

or.

2. his drivers are so fubared, he might need to do a reformat to fix it.


Also his cpu is a major bottleneck.
Im sorry, as others said in the thread... u do not drop a monster gpu and XFIRE them on a cpu that has been outdated by 2 generations.

Yes u got outdated by 2 generations...
Nehalem and Westmere now... and in 5-6 months, one more will come out to which will be Sandy Bridge.

Keep telling yourself that.

Again, read the Guru3D link and weep. The C2D/Q the architecture is obsolete, and a setup with powerful GPUs is bottlenecked by that architecture. Period.

Agree'd...

OP... your crying over spilled milk.

For now see if its because ur running your programs under windowed conditions.