• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

ATI HD3870XT - Single versus Crossfire

Soldato
Joined
15 Aug 2006
Posts
3,422
Location
127.0.0.1
Going to be doing some benches of various titles in single v crossfire mode, using game bench where possible and FRAPS when not.

CRYSIS

1920x1200 no AA/AF - Vista x64 - DX9 - Q6600 @ 3.8ghz - 850/1200 GPU - 7.12 cat

as you can see below the scaling is very good with 7.12 in crysis even without the coming 'multi gpu enhanced' patch from the devs.


Catalyst optimisations set to maximum quality

-------------------------------------------------------------------------------------------------------------------------------------------------------------

Crysis settings - mix of medium/high

Single HD3870 XT

24/12/2007 16:52:35 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: Custom
Custom Quality Values:
VolumetricEffects=Medium
Texture=High
ObjectDetail=High
Sound=Medium
Shadows=Medium
Water=High
Physics=High
Particles=Medium
Shading=High
PostProcessing=Medium
GameEffects=Medium
==============================================================
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 72.27s, Average FPS: 27.67
Min FPS: 23.13 at frame 1299, Max FPS: 35.14 at frame 987
Average Tri/Sec: -28736632, Tri/Frame: -1038380
Recorded/Played Tris ratio: -0.88
!TimeDemo Run 1 Finished.
Play Time: 65.48s, Average FPS: 30.55
Min FPS: 23.13 at frame 1299, Max FPS: 35.94 at frame 987
Average Tri/Sec: -31281100, Tri/Frame: -1024081
Recorded/Played Tris ratio: -0.90
!TimeDemo Run 2 Finished.
Play Time: 65.40s, Average FPS: 30.58
Min FPS: 23.13 at frame 1299, Max FPS: 35.94 at frame 987
Average Tri/Sec: -31329240, Tri/Frame: -1024450
Recorded/Played Tris ratio: -0.89
TimeDemo Play Ended, (3 Runs Performed)
==============================================================
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
24/12/2007 16:52:35 - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: Custom ~~ Overall Average FPS: 30.565


Crossfire HD3870 XT

24/12/2007 16:28:54 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: Custom
Custom Quality Values:
VolumetricEffects=Medium
Texture=High
ObjectDetail=High
Sound=Medium
Shadows=Medium
Water=High
Physics=High
Particles=Medium
Shading=High
PostProcessing=Medium
GameEffects=Medium
==============================================================
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 43.91s, Average FPS: 45.55
Min FPS: 34.47 at frame 1944, Max FPS: 64.09 at frame 898
Average Tri/Sec: -45062252, Tri/Frame: -989323
Recorded/Played Tris ratio: -0.93
!TimeDemo Run 1 Finished.
Play Time: 36.98s, Average FPS: 54.08
Min FPS: 34.47 at frame 1944, Max FPS: 64.58 at frame 946
Average Tri/Sec: -52660052, Tri/Frame: -973680
Recorded/Played Tris ratio: -0.94
!TimeDemo Run 2 Finished.
Play Time: 36.89s, Average FPS: 54.21
Min FPS: 34.47 at frame 1944, Max FPS: 65.20 at frame 901
Average Tri/Sec: -52821840, Tri/Frame: -974306
Recorded/Played Tris ratio: -0.94
TimeDemo Play Ended, (3 Runs Performed)
==============================================================
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
24/12/2007 16:28:54 - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: Custom ~~ Overall Average FPS: 54.145




Crysis settings - High

Single HD3870 XT

24/12/2007 17:05:01 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
==============================================================
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 98.25s, Average FPS: 20.36
Min FPS: 9.80 at frame 97, Max FPS: 24.30 at frame 1766
Average Tri/Sec: -19108060, Tri/Frame: -938721
Recorded/Played Tris ratio: -0.98
!TimeDemo Run 1 Finished.
Play Time: 90.41s, Average FPS: 22.12
Min FPS: 9.80 at frame 97, Max FPS: 25.03 at frame 996
Average Tri/Sec: -20515902, Tri/Frame: -927398
Recorded/Played Tris ratio: -0.99
!TimeDemo Run 2 Finished.
Play Time: 90.90s, Average FPS: 22.00
Min FPS: 9.80 at frame 97, Max FPS: 25.27 at frame 1001
Average Tri/Sec: -20398410, Tri/Frame: -927123
Recorded/Played Tris ratio: -0.99
TimeDemo Play Ended, (3 Runs Performed)
==============================================================
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
24/12/2007 17:05:01 - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 22.06



Crossfire HD3870 XT

24/12/2007 16:37:21 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX9 1900x1200, AA=No AA, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
==============================================================
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 59.15s, Average FPS: 33.81
Min FPS: 27.68 at frame 158, Max FPS: 43.95 at frame 880
Average Tri/Sec: -24553254, Tri/Frame: -726183
Recorded/Played Tris ratio: -1.26
!TimeDemo Run 1 Finished.
Play Time: 51.91s, Average FPS: 38.53
Min FPS: 27.68 at frame 158, Max FPS: 44.76 at frame 834
Average Tri/Sec: -27834230, Tri/Frame: -722435
Recorded/Played Tris ratio: -1.27
!TimeDemo Run 2 Finished.
Play Time: 51.78s, Average FPS: 38.62
Min FPS: 27.68 at frame 158, Max FPS: 46.34 at frame 1013
Average Tri/Sec: -27921360, Tri/Frame: -722883
Recorded/Played Tris ratio: -1.27
TimeDemo Play Ended, (3 Runs Performed)
==============================================================
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
24/12/2007 16:37:21 - Vista 64
Run #1- DX9 1900x1200 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 38.575
 
Last edited:
COD4

FPS captured with FRAPS in the first level about 5mins in.

1920x1200 4xAA/16xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1200 GPU - 7.12 cat - max settings in game


Single HD3870 XT

Frames, Time (ms), Min, Max, Avg
12915, 303314, 20, 81, 42.580


Crossfire HD3870 XT

Frames, Time (ms), Min, Max, Avg
24509, 295210, 38, 150, 83.022



Unreal Tournament 3

1920x1200 0xAA/16xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1200 GPU - 7.12 cat - max settings in game (insane or intense? lol)

i thought crossfire was knackered on this game until i moved Catalyst AI from standard (where it should be on crossfire) to advanced, was getting all kinds of crashes and screen flicker otherwise. quite a few ppl on rage3d have reported problems on crossfire with this game but it seems to be doing ok under 7.12 for me in vista 64 with the Cat AI adjustment. something to keep in mind if you have a similar setup and problems with UT3 :)

ShangriLa Level - 12 x Botmatch - 5 mins

Single HD3870 XT

Frames, Time (ms), Min, Max, Avg
15314, 240825, 38, 92, 63.590


Crossfire HD3870 XT

Frames, Time (ms), Min, Max, Avg
31375, 305864 , 54, 180, 102.578



Bioshock

1920x1200 0xAA/0xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1200 GPU - 7.12 cat - max settings in game

This test i really noticed the difference from my GTX as i really didn't enjoy the framerate with everything maxed out on the nvidia card. With the 3870 in single mode it seemed a little slower than the GTX and in crossfire noticably smoother than both, minimum 46fps and avg of 60 is perfect for this game, can you remove the 60fps cap in this like other UT3 games????

Single HD3870 XT

Frames, Time (ms), Min, Max, Avg
29161, 766324, 21, 62, 38.053

Crossfire HD3870 XT

Frames, Time (ms), Min, Max, Avg
45896, 766751, 46, 63, 59.858


F.E.A.R.

1920x1200 4xAA/16xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1261 GPU - 7.12 cat - max settings in game - soft shadows on


Single HD3870 XT

Min 20 Max 120 Avg 58


Crossfire HD3870 XT

Min 27 Max 415 Avg 108



HL2 : Lost Coast Benchmark

1920x1200 6xAA/16xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1261 GPU - 7.12 cat - max settings in game (water reflect all)

Single HD3870 XT

Frames, Time (ms), Min, Max, Avg
20015, 144664, 59, 274, 138.355

Crossfire HD3870 XT

Frames, Time (ms), Min, Max, Avg
24038, 144500, 53, 297, 166.353


8xAA/16xAF
Frames, Time (ms), Min, Max, Avg
20032, 144206, 48, 294, 138.912


16xAA/16xAF
Frames, Time (ms), Min, Max, Avg
13445, 144621, 45, 213, 92.967


16xAA/16xAF Adaptive AA Disabled
Frames, Time (ms), Min, Max, Avg
14255, 144415, 45, 212, 98.709



HL2:EP2

1920x1200 8xMSAA/16xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1261 GPU - 7.12 cat - max settings in game (water refect world)

The scores are from the first part of the game minus any cutscenes, basically from where that alien dood rescues alyx -> 'this vortal coil'.

Can only assume the minimum fps at zero is due to the section loading as there were no visible dips below 50-60fps, shame as it would have been nice to get that reading too. Considering the 8xAA filtering the game gets nice scores even in single card versus how sweet it looks.

Single HD3870 XT

Frames, Time (ms), Min, Max, Avg
55438, 752624, 0, 147, 73.660

Crossfire HD3870 XT

Frames, Time (ms), Min, Max, Avg
136720, 956881, 0, 273, 142.881



Medal Of Honor : Airborne

1920x1200 0xAA/0xAF Adaptive AA Enabled - Vista x64 - Q6600 @ 3.8ghz - 850/1261 GPU - 7.12 cat - max settings in game.

To be honest this is about 5 mins worth of play on the very first level but you can clearly see the impact that crossfire makes over the single card, the minimum fps make it seamless ...there's just pure action and no slowdown @ all.

Single HD3870 XT

Frames, Time (ms), Min, Max, Avg
30928, 477774, 39, 90, 64.734

Crossfire HD3870 XT

Frames, Time (ms), Min, Max, Avg
19756, 171818, 77, 172, 114.982
 
Last edited:
Crossfire seems very good, 100% boost there from single card.

We need someone to do the same with 8800GT SLi now with similar specs, 2 3870 look a really good choice of card.
 
i cant believe those scores wow, i was expecting much less as all i hear from nvidia owners is how SLI is a waste of money for the small gains.

this has me salivating at the mouth waiting for a pair of 3870 X2's
 
sorry for posting so much but question on that crysis test~

is that high test you listed the "very high" setting with all the direct x 10 bells and whistles?
 
it's high in DX9 in Vista 64.

Crysis in DX10 is pretty much unplayable on anything, i think i got about 23fps in crossfire in DX10.....i can run the test and post results if you like.
 
it's high in DX9 in Vista 64.

Crysis in DX10 is pretty much unplayable on anything, i think i got about 23fps in crossfire in DX10.....i can run the test and post results if you like.

that would be awesome thanks as im looking to get 2 of the HD3870 x2's, and if they perform the same as what your tests show, i can sort of estimate how crysis will run with the 4 cores.

you're a real star marscay
 
Seems like ATI are getting some great drivers out. Glad its one thing they still have that Nvidia doesn't.
 
that would be awesome thanks as im looking to get 2 of the HD3870 x2's, and if they perform the same as what your tests show, i can sort of estimate how crysis will run with the 4 cores.

you're a real star marscay

here you go m8, with 4 cores and more optimisation you might be looking at over 40fps @ 1900 res with 4 cores.


DX10 High settings 1680x1050

24/12/2007 22:57:55 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX10 1680x1050, AA=No AA, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
==============================================================
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 99.83s, Average FPS: 20.03
Min FPS: 11.48 at frame 1944, Max FPS: 26.16 at frame 1540
Average Tri/Sec: -15333111, Tri/Frame: -765320
Recorded/Played Tris ratio: -1.20
!TimeDemo Run 1 Finished.
Play Time: 90.10s, Average FPS: 22.20
Min FPS: 11.48 at frame 1944, Max FPS: 26.16 at frame 1540
Average Tri/Sec: -16948976, Tri/Frame: -763517
Recorded/Played Tris ratio: -1.20
!TimeDemo Run 2 Finished.
Play Time: 88.78s, Average FPS: 22.53
Min FPS: 11.48 at frame 1944, Max FPS: 26.16 at frame 1540
Average Tri/Sec: -17200500, Tri/Frame: -763540
Recorded/Played Tris ratio: -1.20
TimeDemo Play Ended, (3 Runs Performed)
==============================================================
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
24/12/2007 22:57:55 - Vista 64
Run #1- DX10 1680x1050 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 22.365


DX10 High settings 1920x1200

24/12/2007 22:47:41 - Vista 64
Beginning Run #1 on Map-island, Demo-benchmark_gpu
DX10 1900x1200, AA=No AA, Vsync=Disabled, 64 bit test, FullScreen
Demo Loops=3, Time Of Day= 9
Global Game Quality: High
==============================================================
TimeDemo Play Started , (Total Frames: 2000, Recorded Time: 111.86s)
!TimeDemo Run 0 Finished.
Play Time: 119.52s, Average FPS: 16.73
Min FPS: 11.12 at frame 1932, Max FPS: 21.28 at frame 1614
Average Tri/Sec: -12294512, Tri/Frame: -734710
Recorded/Played Tris ratio: -1.25
!TimeDemo Run 1 Finished.
Play Time: 106.21s, Average FPS: 18.83
Min FPS: 11.12 at frame 1932, Max FPS: 22.34 at frame 1547
Average Tri/Sec: -13802733, Tri/Frame: -733003
Recorded/Played Tris ratio: -1.25
!TimeDemo Run 2 Finished.
Play Time: 105.30s, Average FPS: 18.99
Min FPS: 11.12 at frame 1932, Max FPS: 22.34 at frame 1547
Average Tri/Sec: -13923636, Tri/Frame: -733073
Recorded/Played Tris ratio: -1.25
TimeDemo Play Ended, (3 Runs Performed)
==============================================================
Completed All Tests
<><><><><><><><><><><><><>>--SUMMARY--<<><><><><><><><><><><><><>
24/12/2007 22:47:41 - Vista 64
Run #1- DX10 1900x1200 AA=No AA, 64 bit test, Quality: High ~~ Overall Average FPS: 18.91
 
Very nice results there Marscay, glad to see you've been able to appreciate ATI when most people seem to be in the slagging them off as its in vogue party.

Some very nice results there really, its a shame ATI couldnt have brought the 3870XT to the party instead of the 2900XT first, as with crossfire performance like that for the price, they're exceedingly competetive against the GTX and Ultra, especially when you work in pricing for two cards fps:£, exceptional really up til recently. The 2900Pros which were slowly approaching the £150 were closer to the mark, but hit rocky ground over power demands.


As a note, I think sticking Catalyst AI to Advanced forces the card to use AFR dual card rendering mode, which generally works for all games, however isnt always perhaps quite as fast as specifically optimised dual card modes. Seems like a good way to get moderate performance improvements, with better compatability though, from your UT3 test :)

It's a real shame that ATI couldnt have had the 3890 or whatever its going to be called out before xmas, or some other performance part to compete with the Nvidia 8800 G92 series, as I've always prefered ATI drivers, but have gone Nvidia to last me better with its improved AA handling over a *single ATI card* etc, can't say Im unhappy with the card, but for example its resolution options arent as well customised out the box with Nvidia cards, just little things really, (not to mention my cpu etc is holding back the card a bit I feel, am going to do the siggy mentioned revamp soon) but then its my first REAL Nvidia card since the Geforce256 so perhaps I'll learn to like the drivers and control panel Nvidia use, as well as the ATI one.
 
Last edited:
good thread, have a 5 star for it

i think it would be a good idea to post a table or something in the opening post to make it easier to read and compare results
 
If ATI can pull this off with 4 cards aswell with 2 3890 and generally find a solution for future hardware that gives almost 100% from crossfire then they might have a nice weapon to fight back. Seems like getting a 3850 512mb wasn't a bad choice after all since i can get one later on and nearly double my performance in games.
 
here you go m8, with 4 cores and more optimisation you might be looking at over 40fps @ 1900 res with 4 cores.
[/SIZE][/SIZE]

ill be honest with you, my current monitor/rig runs games at 1280x1024, when you said 1900 just then my eyes widened like a 5 year old on xmas morning. i was more expecting to run it at 720p resolution with 40fps, now its more like a solid 60 at that res.

marscay, thanks a bunch for all your hard work!
 
Those are very impressive results indeed. Will definitely be going xfire sooner rather than later.
All depends on whether i can wait for the X2 version to save some space.
 
Back
Top Bottom