Performance in games

Testbed configuration

• Intel Core2 Extreme QX9650 (3000 MHz) CPU
• Zotac 790i Ultra motherboard on NVIDIA nForce 790i Ultra
• 2GB DDR3 SDRAM Corsair 2000MHz (CAS (tCL)=5, RAS to CAS delay (tRCD)=5, Row Precharge (tRP)=5, tRAS=15)
• WD Caviar SE WD1600JD 160GB SATA hard drive
• Tagan TG900-BZ 900W PSU
• Windows Vista 32bit SP1, DirectX 10.1
• Dell 3007WFP 30-inch monitor
• ATI CATALYST 8.5; NVIDIA Forceware 175.16 (9xxx series) and 177.34 (GTX 280)
• VSync disabled

Benchmarks

• Call Of Juarez (Techland/Ubisoft) - DirectX 9.0, Shaders 3.0 (HDR), maximum quality settings; demo, batch file included.
• S.T.A.L.K.E.R. 1.003 (GSC Game World/THQ) - DirectX 9.0, maximum quality settings (dynamic lighting enabled); demo, copy files to the savegames folder, run the game, load level 'ixbt3', and type "demo_play ixbt3" in the console.
• 3DMark Vantage 1.00 (FutureMark) - DirectX 10.0, Shaders 4.0, multitexturing, 'Extreme' settings.
• CRYSIS 1.2 (Crytek/EA), DirectX 10.0, Shaders 4.0, 'Very High' settings, levels 'Rescue' and 'Harbor'; batch file, e-mail us to obtain the timedemo. We express gratitude to CRYTEK for creating a timedemo for iXBT.com / Digit-Life.
• Company Of Heroes Opposing Fronts (Relic Entertainment/THQ) - DirectX 10.0, Shaders 4.0, maximum quality settings; batch file, run the game, invoke graphics settings and click the test button.
• World In Conflict 1.007 (Massive Entertainment/Sierra) - DirectX 10.0, Shaders 4.0, 'Very High' settings with adjusted AA and AF; run the game, invoke graphics settings and click the test button.
• 3DMark06 1.10 (FutureMark) - DirectX 9.0c, multitexturing, 'Trilinear' settings.

Note that performance charts are located on a dedicated page, because they don't fit a reasonable resolution.

S.T.A.L.K.E.R.

Performance charts: S.T.A.L.K.E.R. (AA doesn't work in this game; AF only).

This game uses relatively old technologies. However, there are still many such games in the market, so its results are very interesting. Performance is apparently limited by system resources, by CPU, in low resolutions. So there are almost no differences there. As game resolution grows, we can see the total fill rate in the 9800 GX2 start to affect results. So the GTX 280 card is outperformed. However, the new card has an apparent advantage over the 9800 GTX. It's not outperformed by the 3870 X2 either (although it's because of the CrossFire problems in this game).

World In Conflict

Performance charts: World In Conflict

This game supports all latest innovations, including DX10. Firstly, the new card shoots forward on all fronts. Secondly, 512-bit bus provides really high bandwidth, and performance gains in AA+AF modes are fantastic. Especially relative to the 9800 GX2, which suffers from SLI problems in this game.

ATI cards are apparently left far behind.

CRYSIS, Rescue, DX10, Very High

Performance charts: CRYSIS, Rescue, DX10, Very High

Something strange happens here. The GTX 280 card is almost in between the 9800 GTX and the GX2. The new product leads in the 2560x1600 mode, where SLI fails (we already wrote about it). However, the game is too slow to play.

It's either the fill rate has a strong effect in this game (GX2 is much faster in this respect than the new product), or it's a problem with the drivers. You should also keep in mind that the average FPS of a SLI system may fail to reflect the real low instant FPS. That is the average FPS will be high, but it will be difficult to play because of lags.

CRYSIS, Harbor, DX10, Very High

Performance charts: CRYSIS, Harbor, DX10, Very High

The other scene in this game demonstrates the same layout of forces, so we have the same conclusions.

Call Of Juarez

Performance charts: CoJ

This game was written for DX9. It has high requirements to multitexturing speed, so the GTX 280 cannot outperform a more powerful card (from the point of view of fill rate). Or DX9 technologies and SLI are finetuned better and demonstrate much higher results than in DX10. The last conclusion seems right.

That's why the 3870 X2 is faster here.

Company Of Heroes

Performance charts: CoH

As soon as we run into a DX10 application, the new product shoots forward again. Especially with AA+AF.

What concerns the 3870 Х2, CrossFire fails here, and the card demonstrates low results.

3DMark Vantage: Graphics MARKS

Performance charts: 3DMark Vantage: Graphics MARKS

That's another DX10 application, plus optimizations for the new benchmark (groups of programmers always work on 3DMark optimizations both in NVIDIA and AMD), so the GTX 280 demonstrates phenomenal results.

3DMark06: SHADER 2.0 MARKS

Performance charts: 3DMark06 SM2.0 MARKS

That's an opposite example. As soon as shaders of the previous generation come into action, the new card slips up. So, the previous-generation cards shoot forward.

3DMark06: SHADER 3.0 MARKS

Performance charts: 3DMark06 SM3.0 MARKS

Ditto.