PDA

View Full Version : My 4870x2 vs 4870 tests

Pages : [1] 2
LordJuanlo
08-18-08, 03:40 PM
Another benchmark run, I will use 90% of the games used in my previous (http://www.nvnews.net/vbulletin/showthread.php?t=115791) test. Games that have been excluded are:

- Assassin's Creed, because I have lost the savegame point
- Lord of the Rings Online, because testing was done in DX10 and after latest LOTRO patch DX10 is broken, it "kind of" works but it's unstable and the option for DX10 distant landscape has been disabled by developers, so comparison wouldn't be fair.

I have added a couple of new games that I didn't bench on the 8800GTX, but I have tested with the 4870 before removing it:

- Call of Juarez DX10 Benchmark
- Company of Heroes Opposing Fronts Benchmark
- Timeshift

The following settings are used:

- Resolution 1920x1200
- 4xMSAA (unless other specified)
- 16xAF
- Vsync disabled
- Default video card clocks
- 3 Ghz CPU speed (333x9)
- Vista Home Premium x64 SP1
- Catalyst 8.6 Hotfix for the 4870 / 8.8 Beta (8.52.6) for the x2
- AA and AF always from game setup, not forced via drivers

To do: Full Crysis benchmarks (very soon) and Lord of the Rings Online benchmarks (when Turbine fixes DX10)

On the left you get a graph with the FPS of the scene that you can see on the right in both cards. In games where I have used the built-in benchmark, you will get minimum, average and maximum.

Half Life 2 Episode 2

http://img145.imageshack.us/img145/1759/episode21cc0.png http://img379.imageshack.us/img379/3229/ep214xaaativi8.th.jpg (http://img379.imageshack.us/my.php?image=ep214xaaativi8.jpg)

http://img134.imageshack.us/img134/199/episode22wu3.png http://img59.imageshack.us/img59/2718/ep224xaaatijb1.th.jpg (http://img59.imageshack.us/my.php?image=ep224xaaatijb1.jpg)

http://img135.imageshack.us/img135/5936/episode23of8.png http://img379.imageshack.us/img379/7262/ep234xaaatiqa8.th.jpg (http://img379.imageshack.us/my.php?image=ep234xaaatiqa8.jpg)

http://img147.imageshack.us/img147/817/episode24dq0.png http://img375.imageshack.us/img375/6656/ep244xaaativd0.th.jpg (http://img375.imageshack.us/my.php?image=ep244xaaativd0.jpg)

http://img512.imageshack.us/img512/8656/episode25bv3.png http://img369.imageshack.us/img369/7632/ep254xaaatisb9.th.jpg (http://img369.imageshack.us/my.php?image=ep254xaaatisb9.jpg)

http://img134.imageshack.us/img134/199/episode26as6.png http://img59.imageshack.us/img59/342/ep264xaaaticu8.th.jpg (http://img59.imageshack.us/my.php?image=ep264xaaaticu8.jpg)

http://img135.imageshack.us/img135/4540/episode27is4.png http://img386.imageshack.us/img386/3818/ep274xaaativk3.th.jpg (http://img386.imageshack.us/my.php?image=ep274xaaativk3.jpg)

As you can see, Half Life 2 scales great on the first levels, but the last two show very similar performance. I don't know if it depends on the kind of scene, but it looks like Crossfire scaling is not constant in this game. I have re-run the last two tests many times, obtaining very similar numbers. I have noticed in those maps that the fan of the card does not spin as fast as in the others, it looks like the driver isn't enabling Crossfire, maybe it has something to do with Catalyst AI, or maybe is some kind of CPU bound. I would like that anybody with this card could test those maps too (just start a new game and go directly to chapters 6 and 7).

Bioshock

http://img390.imageshack.us/img390/4469/bioshockqg4.png http://img224.imageshack.us/img224/1787/bioshocknoaaatiiz6.th.jpg (http://img224.imageshack.us/my.php?image=bioshocknoaaatiiz6.jpg)

Fantastic performance here, close to double performance. But as in Half Life 2, I wonder whether it will be constant all the game. Sandly I didn't take other screenshots of other parts of the game with the 4870, but I will if I find out how to disable one of the cores and run with Crossfire disabled.

BloodRayne 2 (Walterman's FSAA patch)

http://img390.imageshack.us/img390/5940/bloodrayne2hi2.png http://img224.imageshack.us/img224/6234/bloodrayne24xaaatiex5.th.jpg (http://img224.imageshack.us/my.php?image=bloodrayne24xaaatiex5.jpg)

This old DX8 game, although greatly enhanced by Walterman's patch, shows a small performance decrease with Crossfire

Devil May Cry 4

http://img84.imageshack.us/img84/7163/devilmaycry4mr1.png http://img508.imageshack.us/img508/5260/devilmaycry4dx920080818dx8.th.jpg (http://img508.imageshack.us/my.php?image=devilmaycry4dx920080818dx8.jpg)

Another title that scales very good. Performance is insane.
LordJuanlo
08-18-08, 03:40 PM
Crysis 64bit (DX9 High no AA)

http://img353.imageshack.us/img353/3893/crysisdb5.png http://img398.imageshack.us/img398/2287/crysis64dx9noaaatick4.th.jpg (http://img398.imageshack.us/my.php?image=crysis64dx9noaaatick4.jpg)

Crysis performs really well on the 4870x2, AA is free, even at 8x in DX9. I did this bench without AA, but more Crysis benchmarks will be available soon with high AA modes

Gears of War (DX10)

http://img360.imageshack.us/img360/6/gearsofwar1oc3.png http://img375.imageshack.us/img375/8176/gow14xaaatiul4.th.jpg (http://img375.imageshack.us/my.php?image=gow14xaaatiul4.jpg)

http://img363.imageshack.us/img363/734/gearsofwar2wh9.png http://img370.imageshack.us/img370/9163/gow24xaaatioy1.th.jpg (http://img370.imageshack.us/my.php?image=gow24xaaatioy1.jpg)

http://img399.imageshack.us/img399/8437/gearsofwar3qh0.png http://img234.imageshack.us/img234/57/gow34xaaatiqr2.th.jpg (http://img234.imageshack.us/my.php?image=gow34xaaatiqr2.jpg)

http://img174.imageshack.us/img174/8084/gearsofwar4qv0.png http://img378.imageshack.us/img378/8711/gow44xaaatiba8.th.jpg (http://img378.imageshack.us/my.php?image=gow44xaaatiba8.jpg)

http://img508.imageshack.us/img508/3066/gearsofwar5mv4.png http://img370.imageshack.us/img370/8396/gow54xaaatiph6.th.jpg (http://img370.imageshack.us/my.php?image=gow54xaaatiph6.jpg)

Another title with irregular performance, the x2 reaches full 60 fps in some of the levels where the 4870 couldn't, however there are a couple of maps where it reaches the same FPS. Very strange.

Race Driver Grid
http://img353.imageshack.us/img353/8739/gridwt5.png http://img398.imageshack.us/img398/8255/grid4xaaatisu8.th.jpg (http://img398.imageshack.us/my.php?image=grid4xaaatisu8.jpg)

Almost double performance again, Grid really loves ATI 4000 cards

STALKER
http://img353.imageshack.us/img353/6868/stalkeruf2.png http://img398.imageshack.us/img398/6830/stalkermaxaaatilk5.th.jpg (http://img398.imageshack.us/my.php?image=stalkermaxaaatilk5.jpg)

I have no words for this one, it's amazing to watch Stalker with a high detail mod pumping 120 frames per second. Double performance, rock on!

Shadowgrounds Survivor
http://img102.imageshack.us/img102/4227/survivorlq6.png http://img398.imageshack.us/img398/7306/survivor4xaaatinx3.th.jpg (http://img398.imageshack.us/my.php?image=survivor4xaaatinx3.jpg)

This game also benefits from the dual 4870, nice performance boost.

Company of Heroes Opposing Fronts

http://img353.imageshack.us/img353/1341/companyofheroesoi9.png http://img353.imageshack.us/img353/78/cohcfgao0.th.jpg (http://img353.imageshack.us/my.php?image=cohcfgao0.jpg)
LordJuanlo
08-18-08, 03:41 PM
Clive Barker's Jericho

http://img161.imageshack.us/img161/3927/jericho1dp0.png http://img375.imageshack.us/img375/855/jericho14xaaatiss4.th.jpg (http://img375.imageshack.us/my.php?image=jericho14xaaatiss4.jpg)

http://img80.imageshack.us/img80/915/jericho2qa1.png http://img370.imageshack.us/img370/9738/jericho24xaaatias4.th.jpg (http://img370.imageshack.us/my.php?image=jericho24xaaatias4.jpg)

http://img294.imageshack.us/img294/4760/jericho3al6.png http://img234.imageshack.us/img234/1802/jericho34xaaatizk1.th.jpg (http://img234.imageshack.us/my.php?image=jericho34xaaatizk1.jpg)

http://img237.imageshack.us/img237/4306/jericho4bz6.png http://img378.imageshack.us/img378/4585/jericho44xaaatiqm5.th.jpg (http://img378.imageshack.us/my.php?image=jericho44xaaatiqm5.jpg)

http://img244.imageshack.us/img244/5278/jericho5dw8.png http://img370.imageshack.us/img370/9175/jericho54xaaatiiw3.th.jpg (http://img370.imageshack.us/my.php?image=jericho54xaaatiiw3.jpg)

This game has been very underrated IMO, I really like it and it's and another title that scales really well with Crossfire. BTW performance with 8xAA is the same than 4x.

The Witcher (2xAA)

http://img187.imageshack.us/img187/9830/thewitcherjt8.png http://img65.imageshack.us/img65/8710/thewitcher2xaaatint9.th.jpg (http://img65.imageshack.us/my.php?image=thewitcher2xaaatint9.jpg)

No Crossfire at all in the masterpiece of CD Projekt, I really hope future ATI drivers will fix it. If anyone got different results, please let me know.

Tomb Raider Legend

http://img187.imageshack.us/img187/4786/legendjy6.png http://img65.imageshack.us/img65/5733/tombraiderlegendfsaaatigl9.th.jpg (http://img65.imageshack.us/my.php?image=tombraiderlegendfsaaatigl9.jpg)

Very nice performance boost, this game has needed some graphics card generations to reach constant 60FPS with 4xAA, vSync and Next-Gen content at Full HD resolution. At last!

Two Worlds

http://img187.imageshack.us/img187/1597/twoworldswu3.png http://img65.imageshack.us/img65/1691/twoworlds4xaaatims7.th.jpg (http://img65.imageshack.us/my.php?image=twoworlds4xaaatims7.jpg)

Very small Crossfire benefits for that title

Rainbow Six Vegas 2

http://img187.imageshack.us/img187/6015/vegas2zz7.png http://img65.imageshack.us/img65/1475/vegas2quality4aaatiqx2.th.jpg (http://img65.imageshack.us/my.php?image=vegas2quality4aaatiqx2.jpg)

This UE3 powered game also shows a little gain in Crossfire, maybe in more demanding levels it will be more noticeable

Call of Juarez

http://img503.imageshack.us/img503/4881/callofjuarezmv5.png

This game also uses Crossfire nicely

Timeshift

http://img503.imageshack.us/img503/8413/timeshiftro5.png

Another one that loves dual cards, impressive boost

(sorry no picture for TimeShift and Call of Juarez, 20 pictures per post limit)
LordJuanlo
08-18-08, 03:42 PM
Reserved for Crysis Benchmarks
LordJuanlo
08-18-08, 03:42 PM
Reserved for Lord of the Rings Online Benchmarks
slaWter
08-18-08, 04:21 PM
LordJuanlo, thanks a lot for the tests!

AFAIK you need to disable Catalyst AI if you want to disable Crossfire.
HYBRID
08-18-08, 04:29 PM
Nice work man.
methimpikehoses
08-18-08, 05:56 PM
Wow, great work! :)
SH64
08-18-08, 06:07 PM
Good work!! .. awaiting the Crysis very high benchies!
DRen72
08-18-08, 09:59 PM
Yeah, I have to give you huge props too. Incredible reviews there.
sammy sung
08-18-08, 10:37 PM
Nice one :thumbsup:
Xion X2
08-18-08, 10:38 PM
Gears of War has a framerate cap at 62. That's why you're seeing similar results.

Supposedly there's a way to disable it, but it's tricky. I couldn't figure out how in the short time I messed with it.
MUYA
08-19-08, 06:15 AM
Thanks for the hard work dude!

Tempted to get a X2 but, its like USD$500+ right now in HKG....will wait for a few more weeks.
MUYA
08-19-08, 06:15 AM
Thanks for the hard work dude!

Tempted to get a X2 but, its like USD$500+ right now in HKG....will wait for a few more weeks.
zer0
08-19-08, 08:19 AM
@LordJuanlo

excellent review man!
slaWter
08-19-08, 09:27 AM
Good work!! .. awaiting the Crysis very high benchies!

Here are a few Crysis results:

Info: Vista x64 SP1, 177.83 for GTX 280, 8.52.2 for 4870X2

GPU Benchmark:

DX9 High:

4870 X2:
Run #1- DX9 1920x1080 AA=No AA, 64 bit test, Quality: High ~~ Last Average FPS: 58.51
Run #2- DX9 1920x1080 AA=2x, 64 bit test, Quality: High ~~ Last Average FPS: 51.84
Run #3- DX9 1920x1080 AA=4x, 64 bit test, Quality: High ~~ Last Average FPS: 50.01

GTX 280:
Run #1- DX9 1920x1080 AA=No AA, 64 bit test, Quality: High ~~ Last Average FPS: 48.13
Run #2- DX9 1920x1080 AA=2x, 64 bit test, Quality: High ~~ Last Average FPS: 42.47
Run #3- DX9 1920x1080 AA=4x, 64 bit test, Quality: High ~~ Last Average FPS: 37.96


DX10 Very High:

4870 X2:
Run #1- DX10 1920x1080 AA=No AA, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 32.36
Run #2- DX10 1920x1080 AA=2x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 29.91
Run #3- DX10 1920x1080 AA=4x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 28.59

GTX 280:
Run #1- DX10 1920x1080 AA=No AA, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 24.82
Run #2- DX10 1920x1080 AA=2x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 21.45
Run #3- DX10 1920x1080 AA=4x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 20.27


Harbor Benchmark:

DX9 High:

4870 X2:
Run #1- DX9 1920x1080 AA=No AA, 64 bit test, Quality: High ~~ Last Average FPS: 46.40
Run #2- DX9 1920x1080 AA=2x, 64 bit test, Quality: High ~~ Last Average FPS: 42.52
Run #3- DX9 1920x1080 AA=4x, 64 bit test, Quality: High ~~ Last Average FPS: 41.29

GTX 280:
Run #1- DX9 1920x1080 AA=No AA, 64 bit test, Quality: High ~~ Last Average FPS: 44.46
Run #2- DX9 1920x1080 AA=2x, 64 bit test, Quality: High ~~ Last Average FPS: 38.68
Run #3- DX9 1920x1080 AA=4x, 64 bit test, Quality: High ~~ Last Average FPS: 35.27


DX10 Very High:

4870 X2:
Run #1- DX10 1920x1080 AA=No AA, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 28.72
Run #2- DX10 1920x1080 AA=2x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 27.33
Run #3- DX10 1920x1080 AA=4x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 25.73

GTX 280:
Run #1- DX10 1920x1080 AA=No AA, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 25.71
Run #2- DX10 1920x1080 AA=2x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 22.41
Run #3- DX10 1920x1080 AA=4x, 64 bit test, Quality: VeryHigh ~~ Last Average FPS: 21.21


AA performance is very, very good, the performance drop is, most of the times, about the same as with the GTX 280 but the game remains playable on the X2.

Performance-wise I prefer the X2 but I still have a few strange problems:
-CCC 2.21 config on LVL 5 (DX9) runs slower than Very High settings, especially with 4x AA, LordJuanlo, can you confirm this?
-1080p 60Hz doesn't work on my Samsung LCD TV, it automatically switches to 1080i 50Hz which looks horrible. 1080p 60Hz works in DX9 on the Samsung an in DX10 on the Dell 3007, very, very strange :o
SH64
08-19-08, 10:00 AM
Thanks slawter!! :thumbsup:

if i'm to wait for Lord i would've got nothing :p
mullet
08-19-08, 11:06 AM
LordJuanlo & slaWter, now these are the benchmarks I have been looking for!!!! Great job guys!! :thumbsup: :thumbsup:
LordJuanlo
08-19-08, 01:35 PM
-CCC 2.21 config on LVL 5 (DX9) runs slower than Very High settings, especially with 4x AA, LordJuanlo, can you confirm this?

Nope, CCC 2.21 LVL 5 is faster, *MUCH* faster than Very High DX10. I'm installing it right now again and will post results. BTW thanks a lot for your numbers, it's a great comparison with the GTX 280
slaWter
08-19-08, 02:01 PM
Nope, CCC 2.21 LVL 5 is faster, *MUCH* faster than Very High DX10. I'm installing it right now again and will post results. BTW thanks a lot for your numbers, it's a great comparison with the GTX 280

No problem ;)

I just reinstalled Crysis and I'm going to test CCC 2.21 again.
slaWter
08-19-08, 02:37 PM
Nope, CCC 2.21 LVL 5 is faster, *MUCH* faster than Very High DX10. I'm installing it right now again and will post results. BTW thanks a lot for your numbers, it's a great comparison with the GTX 280

Didn't work. I still get crappy performance with AA in CCC 2.21. It cuts the frame rate in half :(
LordJuanlo
08-19-08, 02:46 PM
I have just finished a good bunch of Crysis benchs, and there is something wrong with DX10 Very High performance, it could be driver related because your DX10 numbers are much higher. I'm using the driver that came with the card, 8.52.6. What driver are you using Slawter?

My average FPS (with 4xAA) are around 40 FPS under DX9 high, 18 FPS under DX10 Very High and 33 FPS with CCC 2.21 (installed TOD and realistic shadows)
ASUSEN7900GTX
08-19-08, 02:51 PM
great job mate :)

seems the 4870X2 is a monster but the HD4870 ainīt all that bad either looking at those tests this time Nvidia gets itīs ass kicked around the globe a few times by ATI.


with that i mean ATI si back BIIIGGG

great benchies i need to figure how to make diagrams so i can do some benchies of my own but i am so noobish in Excel :D

great job keep those benchies comin
ASUSEN7900GTX
08-19-08, 02:53 PM
how do you bench the harbour as i canīt get a "auto" bench of it like hte island demo

i can choose the harbour but it wonīt benchmark it?
slaWter
08-19-08, 02:58 PM
I have just finished a good bunch of Crysis benchs, and there is something wrong with DX10 Very High performance, it could be driver related because your DX10 numbers are much higher. I'm using the driver that came with the card, 8.52.6. What driver are you using Slawter?

My average FPS (with 4xAA) are around 40 FPS under DX9 high, 18 FPS under DX10 Very High and 33 FPS with CCC 2.21 (installed TOD and realistic shadows)

8.52.2 for those benchmarks.

ASUSEN7900GTX, download the Crysis Benchmark Tool for the Harbor test.