8800 Ultra vs. 2900XT
THE BATTLE OF THE SHADER BEASTS
E6600 @ 3.6GHz
Gigabyte P35 DSR3
Crucial Ballistix DDR2-800 @ 900MHz
Silverstone 850W Zeus
8800 Ultra (660|2300), ForceWare 163.69
2900XT 512MB (875|1950), Catalyst 7.9
Rarely does one come across such a golden opportunity to bench two of the strongest cards on the market back-to-back unless he works for a hardware review site. Well, seeing as I'm a recent 2900 owner who has switched back to the Nvidia camp, I found myself in a unique position. So this past weekend I did some extensive benching between my Ultra and 2900XT. I wanted to test a few of the latest and greatest, most demanding games on the market with two of the best cards. We hear so much of "You can't trust this/that review.." so here was a unique opportunity to lay all the doubts to rest.
The games I tested were Bioshock, Oblivion, Medal of Honor Airborne, Lord of the Rings: Online and Call of Juarez--arguably the five best looking games out at the moment. Each bench was conducted with utmost attention paid to locale (i.e. - all benches were ran in the same section of the game with identical surroundings.) Bioshock and Call of Juarez were benched in DirectX10 mode, and all titles were benched on Vista 32 @ 1680x1050 resolution. The Ultra was at stock clocks (660|2300) and the XT was overclocked (875|1950). I have had the XT all the way to 900MHz on the core in XP, but for some reason I cannot get it stable at that same speed in Vista. A real bummer considering that was one of the key elements that drew me toward it in the first place. Wizzard's ATI Tool is one of the best tweaking gadgets there is with its built in voltage adjustor.
All that aside, I settled for 875MHz on the XT core to do my benching with. At the core of the system is a Conroe E6600 @ 3.6GHz running on the P35 platform, equipped with 2GB of Crucial Ballistix DDR2-800. When swapping cards, the utmost care was taken with cleaning out old driver files before installing the new driver by booting into safe mode to remove old files.
So let's cut to the chase. The first on my list was Bioshock, so let's see how she did on the two cards.
Bioshock - "Plasmids Changed Everythin'"
(Stretch-run = From bathysphere to woman at baby carriage)
As the graphs show, it was just no contest. The Ultra handled it quite well, and the XT struggled in some places.. even with a considerable overclock. Considering Bioshock is a FPS, it makes play difficult when you get frames dropping as low as 24fps. Even with vsync off you get noticeable mouse lag.
Alright, next up was Oblivion. I conducted two test-runs: one without alpha AA and one with
alpha AA. Global settings were as follows--
Oblivion - Vista 32
In-game: Ultra-high, all sliders maxed except self-shadows.
Ini: Grass density set to 110
And away we go...
The location benched was at Fort Camala in the forests just outside of Chorrol. I've found this to be one of the most demanding places on the game with all of the grass, trees, and shadows all over.
The stretch-run was from the fort through the forest and to the top of the hill. I captured some shots showing the start and end of the bench:
Both cards run this game well, and from the benches you can see that the XT hangs with the Ultra pretty well although it's still behind. The bar graph is displaying the tests without alpha AA. The top frapslog is for the Ultra; the bottom frapslog window is for the XT. The 2nd score in the top window is the Ultra with alpha AA enabled (Multisampling Transparency). Amazingly, the Ultra takes only a .8fps hit with it enabled! I accidentally saved over the frapslog run for the XT that had the alpha AA run, but it wasn't too shabby either. It scored 41fps in the bench with alpha AA enabled (Adaptive Multisampling) and took ~3fps hit. Not bad at all, but still behind the Ultra--again.
Lord of the Rings: Online
You don't hear a lot of talk about this game in the benchmarking world, but in my opinion it has to be one of the best looking games out there. Not only does it have the most amazing sky I've ever seen, but it has some impressive lighting effects (post-processing and all that junk), and the draw distance is almost unparalleled. Basically, you can see as far as you ever wanted to. Those trees drawn out far away in the above shot are drawn in full detail at that distance.
In-game: Everything maxed, 16x Anisotropic
Silver Deep Mine to Frerin's Court
If you're wondering if that's a joke, it's not. I gave the XT a big fat 0 because it wouldn't even run the game at my native res. Every time I tried to enable 1680x1050 I would get an "Out of Range" message and my screen would go black.
Needless to say, the Ultra wins this one hands down.
Medal of Honor Airborne
Operation: Husky (entire campaign)
In-game: Maximum settings
Driver: 8x Anisotropic
This was the closest of the benchmarks. The Ultra eeked out a 1fps win over the XT. The XT seems to do fairly well in Unreal engine games as long as you don't run AA. Luckily, Airborne doesn't need much AA at all, if any. It's pretty clean as-is, although a little AA is always nice. I didn't try to force it on as the XT has had trouble enabling AA on these types of games.
Call of Juarez - DX10