Comanche 4 represents one of the few DirectX 8.1 benchmarks available today. Although the game does include a benchmark utility of sorts, it does mimic actual gameplay. It should be noted that this benchmark is notorious for being very CPU dependant. Using an Athlon XP 2400+, the limitations of these cards should be evident without the CPU acting as a bottleneck.

From the start, I was somewhat surprised to see benchmark results of less than 42fps at 1024x768. However, a bit of research illustrated this to be normal. Despite my best efforts, the CPU was actually still the bottleneck for the system. This is depicted by the incredibly small variance in performance throughout each resolution.

Once we enable 2X antialiasing, we begin to see what a brute this benchmark can be. Here, the cards average roughly 35fps at 1024x768 and drop to a low of about 18fps at 1600x1200. Thus far, the 64MB card has kept pace with the 128MB card. As we enable maximum image quality settings, we begin to see a different picture. Throughout each resolution, we see a performance delta between the 64MB and 128MB cards which progressively increases as we raise resolution. At a resolution of 1600x1200, we find the similarly clocked stock EK card and overclocked Abit card to have a performance variance of roughly 3fps.

With Battlefield 1942, we are presented with an incredibly intensive game. This title has garnered the reputation of taxing even the highest performance systems. As such, I was looking forward to seeing how these "budget" cards would handle this application.

Performance throughout each resolution was impressive with default image quality settings. At a resolution of 1600x1200, the overclocked EK’s drastic advantage in core and memory frequencies allowed it to come within 9fps of the 60fps barrier.

Regardless of overclocked frequencies, each card had more than it could handle once enhanced image quality settings were enabled. Though gameplay was still somewhat fluid at a resolution of 1024x768, higher resolutions yielded framerates which were not high enough for solid gaming. As a result, I would have to recommend running this title at a high resolution without antialiasing or anisotropic filtering for the optimum combination of performance and image quality.

Perhaps the most anticipated title to be released this year, Unreal Tournament 2003 was used to indicate performance results in future titles. In this case, I opted to use [H]ard OCP’s benchmark utility for the UT2003 demo to provide accurate and thorough results. One aspect which I was looking to examine was the variance between the overall score and the minimum framerate achieved during testing.

With default image quality settings, every resolution yielded excellent results on either card. Though the Antalus map proved to be most taxing for these cards, framerates remained above 60fps throughout testing. Raising the image quality settings by enabling 2X AA, we see a dramatic change in performance for the 64MB card. Though not evident at a resolution of 1024x768, higher resolutions illustrate the need for additional memory. Even at 1280x1024, there are some cases where the additional memory accounts for an increase of almost 20fps! This resolution illustrates this scenario best as the card has yet to reach its limitations (with exception of the lack of additional memory for the 64MB part).

Enabling maximum image quality settings illustrates this fact to a smaller degree, due to the fact that we are reaching the card’s limitations. Again, using a resolution of 1024x768 indicates no difference between the two similarly clocked cards. However, raising the resolution depicts a different scenario. Yet, once the resolution is raised to 1600x1200 the 128MB card is stressed enough to be brought to the same performance level as the 64MB variant.

Looking at the benchmark results, it is clear that the best combination of speed and quality comes from using 2X antialiasing at a high resolution. Doing so yields stellar graphics at an acceptable frame rate for fluid gameplay.