Review by Steve Angelly - January 16, 2006

GAME BENCHMARK RESULTS

F.E.A.R.

First up is the very popular game F.E.A.R. The in-game benchmark was used to gather results. Settings are listed below with the game identifying my test system as a MEDIUM performance system.

Noted as being very difficult stress-wise on video cards I am not giving the XFX GeForce 6800 GS XXX Edition any breathing room. The tests were run at 1024x768, 1280x1024, and at 1600x1200 resolutions. Full screen antialiasing (AA) and anisotropic filtering (AF) were applied at each resolution with no AA/AF, no AA/16xAF, 2xAA/16xAF, and 4xAA/16xAF.

F.E.A.R

Take a look at the 'Effects' and 'Graphics' settings below as selected by the game. Sure, a few settings are set to medium and Soft Shadows are turned off but look at what has been set to maximum.

The F.E.A.R. benchmark provides average and minimum frame rate as well as the percentage of frames that were less than 25 fps, between 25 and 40 fps, and greater than 40 fps. Resolution used, antialiasing and anisotropic filtering levels used, average frame rates and minimum frame rates results (average of 3 runs) in frames per second, and FPS <25%, 25-40%, and >60% were recorded.

The XFX GeForce 6800 GS XXX Edition performed well at 1024x768 with 2xAA/16xAF, and at 1280x960 with no AA/AF. Any higher resolutions or application of antialiasing brought the card to its knees in this game. A couple of odd results were at 1280x960 and at 1600x1200 when applying 16xAF. First, at 1280x960, applying 16xAF created a hit of 10 FPS. At 1600x1200, applying 16xAF created a 'zero' drop in FPS. Both of these settings may be playable but I suspect marginally so. I plan to test these settings in gameplay but unfortunately I ran out of time for including the results in this review.

With SLI the game is very playable up to 1280x960, 2xAA/16xAF, or 1600x1200, no AA/16xAF. I played at 4xAA/16xAF for a short stint at 1280x960 and the game seemed to run smooth but until I have put some more time with these settings I would not want to say much more. Same goes for 2xAA/16AF and 4xAA/16xAF. I suspect gameplay would be marginal at the higher application of full-screen antialiasing. Again, if results are better than I expect I will update the review.

BATTLEFIELD 2

I used FRAPS to provide the FPS results in this game. Using the in-game benchmark I activated FRAPS at approximately the same frame when the first benchmark screen opens and stopped at the last frame of the benchmark. Also, with this test system I had only 1 MB (2x512MB) of system memory at my disposal so the numbers may be a little low as opposed to using 2 MB of system memory.

BATTLEFIELD 2

Settings Custom Quality Terrain - High Effects - High Geometry - High Texture - High Lighting - Medium Dynamic Shadows - High Dynamic Light - High View Distance Scale - 100%

BF2 Test Results

XFX 6800 GS XXX Resolution 0AA/0AF 0AA/8AF 2AA/8AF 4AA/8AF 1024x768 73.38 73.36 64.26 56.69 1280x1024 67.50 57.91 49.89 43.27 1600x1200 53.93 45.84 38.94 32.61 6800 GS SLI Resolution 0AA/0AF 0AA/8AF 2AA/8AF 4AA/8AF 1024x768 75.92 75.48 72.38 71.35 1280x1024 62.18 59.75 58.90 54.17 1600x1200 49.73 46.62 46.08 39.48

Frame rate results were consistent across the board using the XFX GeForce 6800 GS XXX Edition with actual gameplay being good at 1280x1024, 2xAA/8xAF.  Application of antialiasing and higher resolutions did provide a significant hit when applied, but gameplay was enjoyable up 1280x960, 2xAA/8xAF and also at 1600x1200 with no antialiasing and anisotropic filtering set to 8x.

Checking the overclock (534/1.2) on the XFX GeForce 6800 GS XXX Edition in this game resulted in a net gain of an additional 5 FPS at 1280x1024, 4xAA/8xAF.

With SLI the game performance did not significantly improve and benchmark results did not scale well. Checking and double checking at different resolutions did not provide me with an adequate explanation other than performance was just sporadic. I did however enjoy stutter-free play at 1280x960 with 4xAA enabled on my P225f monitor. I tried the same settings on a 17-inch LCD, except using the native resolution of 1280x1024, and the experience was equally as good.

SLI provided a 16% increase in FPS at 1280x1024, 2xAA/8xAF and over 20% at 1280x1024, 4xAA/8xAF.

While I can recommend the XFX GeForce 6800 GS XXX Edition for this game I could not recommend purchasing an additional card other than just to run SLI or to be able to use 4xAA. A single 7800 GTX performs better for a little more cost and a 7800 GT would be a strong candidate for the same cost or less.

CHRONICLES OF RIDDICK - ESCAPE FROM BUTCHER BAY

This game came out of nowhere and proved to be more than worth the time of gamers everywhere. The movie may have not received high marks but the game has. COR-EFBB stresses high end systems but with just a little tweaking most systems will run the game with no problem and the game itself is a lot of fun with excellent graphics.

The in-game benchmark was used at resolutions of 1024x768, 1280x1024, and 1600x1200. The in-game selection of Shader 2.0 was selected and used in all benchmarks. Application of antialiasing and anisotropic filtering was applied as noted in the results table below.

Chronicles of Riddick - EFBB

Settings Shader Mode - Auto (2.0) Texture Quality - MAX Model Shadows - ON Model Quality - ON Pixel Aspect - Normal Vertical Sync - OFF Gore - ON Focus Frame - ON

Chronicles of Riddick - EFBB Test Results

XFX 6800 GS XXX Resolution 0AA/0AF 0AA/16AF 2AA/16AF 4AA/16AF 1024x768 57.05 55.95 42.05 28.58 1280x960 37.83 37.25 28.01 18.60 1600x1200 27.36 27.26 20.33 13.30 6800 GS SLI Resolution 0AA/0AF 0AA/16AF 2AA/16AF 4AA/16AF 1024x768 97.17 97.24 94.97 74.13 1280x960 97.32 97.11 92.30 71.51 1600x1200 97.31 97.08 89.78 68.68

With shadows on, the performance hit of about 8% was noted throughout the benchmarks. Actual gameplay did not seem to suffer much and I prefer playing this game with shadows and most other in-game features set to 'on' or 'maximum.'

Applying 2xAA initiated a more serious hit of 25% in the benchmark while 4xAA added another excessive 33% hit. This was experienced at all resolutions, further reducing playable resolutions to 1024x768. Dropping anisotropic filtering to 8x at 1024x768, 2xAA, enabled FPS to average above 40 for potentially acceptable gameplay.

Settings of 1280x960, with no AA and up to 16xAF, provided satisfactory gameplay. Upping the clock speeds did allow for use of 2xAA in actual gameplay but I would recommend backing off some on the in-game settings instead. Setting Shader mode to 2.0 or 'auto' is advisable as the use of 2.0++ should be avoided.  This in-game shader setting is very performance intensive and causes a severe hit in frame rates of 50% or more.

With a little tweaking this game, at moderate resolutions, runs good with the XFX GeForce 6800 GS XXX Edition.

COR-EFBB is a different game when running the GeForce 6800 GS in SLI with the game becoming CPU-bound. This is not a bad thing because in-game settings can be run at maximum at the highest resolutions. The only severe hit comes from running Shader 2.0++ (soft shadows) with FPS dropping an excessive 50%. However, even with Shader 2.0++ enabled frame rates stay in the 40s which still provide acceptable gaming performance.

I did not see any real benefit from application of Shader 2.0++ in the game. In fact, 1600x1200, 8xS/16xAF provided excellent gameplay with excellent graphics with a similar hit in performance while 4xAA/16AF provided better performance with near equal graphics.

As a side note, I did enjoy being able to play this game full throttle (above 70 FPS) at a resolution of 2048x1536. For me, that was a gaming experience with few rivals.

FAR CRY

This game has been around for a while now but, upgraded with HDR capability, is still a great benchmarking tool. XFX packaged the DVD full game version with the XFX GeForce 6800 GS XXX Edition so they have confidence that it will run well.

The test consists of the three runs of the Obisoft 'Regulator' benchmark made available by HardwareOC's version 1.4.1 Far Cry benchmark suite. Resolutions of 1024x768, 1280x1024, and 1600x1200 were used with no AA/AF, no AA/16AF, 4AA/16AF, and HDR with 16xAF applied at each resolution. In-game auto detect settings were used with Vertical Sync OFF.

Far Cry

Settings Texture Quality   High Shadow    High Particle Count    High Water       High Special Effects    High Lighting     High Environmental     High Music        High

Far Cry Test Results

XFX 6800 GS XXX Resolution 0AA/0AF 0AA/16AF 4xAA/16AF HDR/16AF 1024x768 84.44 85.36 85.34 58.72 1280x960 74.98 75.14 74.94 39.38 1600x1200 64.22 64.85 64.17 28.11 6800 GS SLI Resolution 0AA/0AF 0AA/16AF 4xAA/16AF HDR/16AF 1024x768 86.64 86.76 82.23 62.52 1280x960 85.59 86.09 81.05 41.37 1600x1200 83.27 79.62 79.59 30.08

This game is heavily CPU dependent at the lower resolutions for both the XFX GeForce 6800 GS XXX Edition single card and SLI setups. However, both performed well and the SLI setup still ran strong up to 1600x1200, 4xAA/16xAF. HDR provides a severe hit on frame rates but remains playable for both configurations through 1280x960.

XFX's confidence was not misguided as the XFX GeForce 6800 GS XXX Edition provides excellent gameplay at all resolutions with in-game values all set to 'high.'  I even found good game play with HDR at the native resolution of 1280x1024 on my Hitachi 17" LCD although I did trim the anisotropic filtering back to 8x.

In SLI the cards provided a 20% increase at 1600x1200, 4xAA/16xAF, over the single card.

DOOM3

Doom3

Settings High Quality Special Effects  -  ON Enable Shadows  -  ON Enable Specular  -  ON Enable Bump Maps  - ON Vertical Sync  -  OFF Doom3 Benchmark Results   XFX 6800 GS XXX 6800 GS SLI Resolution 0AA/0AF 0AA/16AF 4xAA/16AF 0AA/0AF 0AA/16AF 4xAA/16AF 1024x768 106.7 106.6 74.4 107.3 107.0 100.5 1280x1024 89.7 89.7 51.1 106.4 106.4 82.4 1600x1200 71.5 71.5 37.3 97.3 99.6 60.9

Like Far Cry, the CPU dependency is prevalent in Doom3 at lower resolution without antialiasing. Applying antialiasing creates a severe hit (43% at 1280x1024) on the single card. Still the game plays smoothly at 1280x1024 or 1280x960 with 4xAA. At 1600x1200, while gameplay experience is marginal with 4xAA, dropping down to 2xAA provides excellent gameplay. Frame rates using 2xAA yields a 32% increase (54.6 FPS vs. 37.3 FPS) over 4xAA using the single XFX card.  Still, the XFX GeForce 6800 GS XXX Edition did well except in the most active game scenarios using 4xAA.

Settings at each resolution tested with SLI functioned very well as identified by the results. Even 4xAA/16xAF at 1600x1200, providing almost a 40% increase over the single card, SLI performed smoothly during tests as well as in actual gameplay.