Hi,
Just for a little information really. Decided to upgrade my cpu/board before I upgrade my GFX (waiting to see how things settle over the next couple of months), and did a few tests to see how this effected GFX performance....
Old rig is an AMD 939 X2 @ 2.7GHz, 4GB DDR mem on an asus A8R32-MVP crossfire board. (dual 16xPCIE 1).
New rig is an Intel Q9550 @ 2.83 (stock), 4GB DDR2 mem on Asus P5Q-Deluxe (Crossfire is 2 pcie2.0 at x8, same bandwidth as old board).
Using the same 2 3870XT's cards on both.
Upgraded as I was getting big performance drops when any MSAA was applied, and there didn't seem much difference when Crossfire was enabled on quite a few games. (also fancied a change to intel, and a quad core
).
Anyhow - obligatory debatable 3dmark06 results with crossfire enabled :-
AMD - got a score of 10437 (SM2 4315, SM3 5870, CPU 2066).
INTEL - got 16207 (SM2 6214, SM3 7690, CPU 4689).
Nice increase in all scores not just cpu.
Then tried the Devil May Cry 4 benchmark/demo. Running at 1920x1200, texture settings on "super high", no MSAA and using crossfire.
On AMD system, got 136.29, 88.79, 167.91, 91.53 ("S").
On INTEL system, got 143.41, 98.86. 189.07, 124.69 ("S").
A bit of an increase, but then it does apparantly "work great on intel quad cores" according to one of the loading screens.
On trying with 2xMSAA, this is where I had issues before.
On AMD system, got 51.30, 35.68, 68.25, 40.68 ("C").
On INTEL system, got 102.72, 75.34, 140.71, 81.71 ("S").
Wow - way better on new system now. Ran it a couple of times to make sure it wasn't an error. Next tried it at 8xMSAA :-
On AMD system, got 50.35, 33.34, 62.50, 40.09 ("C").
On INTEL system, got 100.95, 71.52, 120.82, 80.91 ("S").
Double the speed again, and hardly any difference between 2xmsaa and 8xmsaa scores on either system. I had thought the problem was memory/gfx bottleneck with MSAA applied, but doesn't look the case.
Tried with a single card in each system and got similar scores for each - reaching GPU performance bottleneck then.
Next tried GRID - but crossfire doesn't work with it and I get loads of screen flickering.
Tried WOW - and there is a noticable doubling of frame rate when crossfire enabled on intel system compared to the slight increase before on the amd one.
So - basically, where I thought the issue with MSAA at high res in crossfire was down to my cards/memory limit, it appears to have been cpu/chipset limited in my case. PCIE bandwidth is the same between the two as well, using the same cards etc. Both with fresh installs of XP and using catalyst 8.8.
This is before I even overclock my intel (amd is at max stable overclock) and makes a nice base for a decent high level new GFX card.
Hope I'm not just waffling and this may be of use to somebody.
Just for a little information really. Decided to upgrade my cpu/board before I upgrade my GFX (waiting to see how things settle over the next couple of months), and did a few tests to see how this effected GFX performance....
Old rig is an AMD 939 X2 @ 2.7GHz, 4GB DDR mem on an asus A8R32-MVP crossfire board. (dual 16xPCIE 1).
New rig is an Intel Q9550 @ 2.83 (stock), 4GB DDR2 mem on Asus P5Q-Deluxe (Crossfire is 2 pcie2.0 at x8, same bandwidth as old board).
Using the same 2 3870XT's cards on both.
Upgraded as I was getting big performance drops when any MSAA was applied, and there didn't seem much difference when Crossfire was enabled on quite a few games. (also fancied a change to intel, and a quad core
).Anyhow - obligatory debatable 3dmark06 results with crossfire enabled :-
AMD - got a score of 10437 (SM2 4315, SM3 5870, CPU 2066).
INTEL - got 16207 (SM2 6214, SM3 7690, CPU 4689).
Nice increase in all scores not just cpu.
Then tried the Devil May Cry 4 benchmark/demo. Running at 1920x1200, texture settings on "super high", no MSAA and using crossfire.
On AMD system, got 136.29, 88.79, 167.91, 91.53 ("S").
On INTEL system, got 143.41, 98.86. 189.07, 124.69 ("S").
A bit of an increase, but then it does apparantly "work great on intel quad cores" according to one of the loading screens.
On trying with 2xMSAA, this is where I had issues before.
On AMD system, got 51.30, 35.68, 68.25, 40.68 ("C").
On INTEL system, got 102.72, 75.34, 140.71, 81.71 ("S").
Wow - way better on new system now. Ran it a couple of times to make sure it wasn't an error.
Next tried it at 8xMSAA :-On AMD system, got 50.35, 33.34, 62.50, 40.09 ("C").
On INTEL system, got 100.95, 71.52, 120.82, 80.91 ("S").
Double the speed again, and hardly any difference between 2xmsaa and 8xmsaa scores on either system. I had thought the problem was memory/gfx bottleneck with MSAA applied, but doesn't look the case.
Tried with a single card in each system and got similar scores for each - reaching GPU performance bottleneck then.
Next tried GRID - but crossfire doesn't work with it and I get loads of screen flickering.
Tried WOW - and there is a noticable doubling of frame rate when crossfire enabled on intel system compared to the slight increase before on the amd one.
So - basically, where I thought the issue with MSAA at high res in crossfire was down to my cards/memory limit, it appears to have been cpu/chipset limited in my case. PCIE bandwidth is the same between the two as well, using the same cards etc. Both with fresh installs of XP and using catalyst 8.8.
This is before I even overclock my intel (amd is at max stable overclock) and makes a nice base for a decent high level new GFX card.
Hope I'm not just waffling and this may be of use to somebody.
. It's great when you really see the benefits of upgrading. Nice rig btw