Ok, probably not a lot of new info here, but may be of some use to people with older systems, thinking of getting new cards now that prices are getting competetive again. 
My setup - Asus A8R32MVP socket 939 board (ati 3200 chipset), AMD X2 4600+ at 2.76Ghz, 4GB memory (at 200Mhz, DDR1). Win Xp64 (so still directx 9
) using ati 8.02 drivers.
Got 2 HD3870XT's in crossfire.
Old setup - 2 1900XT's in crossfire. 3dmark06 score - 9314.
Single 3870 - 9289. So pretty much the same as my old crossfire setup, but using only 1 card!
Crossfire enabled 3870's - 10619.
All scores in 3dmark are slightly higher than with a single card, except for the HDR Canyon test which for some reason has a big increase from 44 to 74 fps. 
Tried a few games, benchmarked with fraps, all at 1920x1200.
UT3 (max settings!)
Single 3870 - min 18, max 63, avg 54 fps. (sync/smooth frames off too)
Crossfire - min 38, max 63, avg 58.
Bioshock (max setting, sync off)
Single 3870 - min 39, max 94, avg 63.
Crossfire - min 53, max 96, avg 72.
Crysis (demo version, max settings, but no AA)
Single 3870 - min 0 (?), max 93, avg 37.5.
Crossfire - min 0 (?), max 146, avg 57.
so... looks like basically like crossfire is pretty cpu limited (especially in the 3dmark06 tests!) even at this resolution. UT3 (and ut3 engines games) seem to more cpu dependant, hence why max fps hardly changes. Min framerate gets a nice boost though.
Crysis gets a nice boost in fps as I guess its more GPU dependant.
I guess the same probably applies for the 3870X2 as well in such a system. Could quite happily use a single 3870 in most cases, and at very reasonable frame rates.
I'm going to save up and get a nice cpu/board/mem upgrade in the future to make the most of these cards.
Sorry if this is just a bit of waffling, its late and I'm probably talking rubbish!
My setup - Asus A8R32MVP socket 939 board (ati 3200 chipset), AMD X2 4600+ at 2.76Ghz, 4GB memory (at 200Mhz, DDR1). Win Xp64 (so still directx 9
) using ati 8.02 drivers.Got 2 HD3870XT's in crossfire.
Old setup - 2 1900XT's in crossfire. 3dmark06 score - 9314.
Single 3870 - 9289. So pretty much the same as my old crossfire setup, but using only 1 card!
Crossfire enabled 3870's - 10619.
All scores in 3dmark are slightly higher than with a single card, except for the HDR Canyon test which for some reason has a big increase from 44 to 74 fps. Tried a few games, benchmarked with fraps, all at 1920x1200.
UT3 (max settings!)
Single 3870 - min 18, max 63, avg 54 fps. (sync/smooth frames off too)
Crossfire - min 38, max 63, avg 58.
Bioshock (max setting, sync off)
Single 3870 - min 39, max 94, avg 63.
Crossfire - min 53, max 96, avg 72.
Crysis (demo version, max settings, but no AA)
Single 3870 - min 0 (?), max 93, avg 37.5.
Crossfire - min 0 (?), max 146, avg 57.
so... looks like basically like crossfire is pretty cpu limited (especially in the 3dmark06 tests!) even at this resolution. UT3 (and ut3 engines games) seem to more cpu dependant, hence why max fps hardly changes. Min framerate gets a nice boost though.
Crysis gets a nice boost in fps as I guess its more GPU dependant.
I guess the same probably applies for the 3870X2 as well in such a system. Could quite happily use a single 3870 in most cases, and at very reasonable frame rates.
I'm going to save up and get a nice cpu/board/mem upgrade in the future to make the most of these cards.
Sorry if this is just a bit of waffling, its late and I'm probably talking rubbish!
Last edited:
