With a number of compelling games appearing on the market, a fast processor and a fast graphics card are must have items for the gaming enthusiast. Here I take a look at the recently released demo of Unreal Tournament 2003.
I probably wouldn't have been able to provide this information without the excellent work of Brent Justice and his HardOCP article on benchmarking the UT2003 demo.
You can view the benchmarking log file by clicking here. Most impressive is that the GeForce4 Ti 4600 maintained a minimum frame rate of over 30 frames per second in all of the tests, which are based on the high quality script.
There were 14 hours left before the article deadline and I'm glad I had time to spend with Bethesda's role playing game Morrowind: The Elder Scrolls III. This title should easily win RPG game of the year and will stress the highest-end systems. Prior to last night, the last time I played Morrowind was around the time I completed a preview of the nForce2 back in July. In that article, you'll find complete instructions on how I benchmarked Morrowind, but it was basically a walkthrough of the town of Seyda Neen while following checkpoints in a predetermined path.
This isn't intended to be an exact apples-to-apples comparison since there are a few variables that have come into play.
Results from the nForce based system were based on version 1.1.0605 of Morrowind, while the results from the nForce2 based system are from version 220.127.116.112. However, after reading through the list of updates, the majority of changes were bug fixes and I saw no references to performance enhancements.
Results from the nForce based system were based on Detonator driver version 29.42, while the results from the nForce2 based system are from version 30.82.
Results are from the nForce based system were based on hardware acceleration using a Sound Blaster Live, while the results from the nForce2 based system are from hardware acceleration with integrated audio.
One way to look at these results is that it's a comparison of playing Morrowind at two different points in time. In either case, this represents more of an overall system comparison as a faster processor and the nForce2 with integrated sound are coming into play. Results are provided with maximum, average, and minimum frame rates.
At a resolution of 1920x1440, the nForce2 based system achieved an average of 40, minimum of 26, and maximum of 48 frames per second. While it has served me well, it will be my pleasure to retire the processor hungry Sound Blaster Live.
Developed by Gas Powered Games in partnership with Microsoft, I first looked at Dungeon Siege in our GeForce4 Ti 4200 preview. Having spent time gathering feedback from fans of the graphically rich hack-and-slash RPG, I learned that frame rates in Dungeon Siege weren't as dependent on the capabilities of the graphics sub-system compared to some games.
Shortly thereafter, Gas Powered Games commented that frame rates in Dungeon Siege are more dependent on the central processor and memory sub-system than the graphics card. The processor takes on tasks which include artificial intelligence, line of sight calculations, physics and pathfinding. The CPU also handles the chores of loading and unloading objects in the background in Dungeon Siege's unique continuous world.
Using the benchmark provided by ExtremeTech, I tested performance in Dungeon Siege with sound enabled. However, upon installing the version 1.1 Dungeon Siege patch, the benchmark failed to write out a frame rate log file, although the playback feature functioned normally. This didn't turn out to be a problem, since I used FRAPS to capture the minimum, average, and maximum frame rate during playback. In the above screenshots, you'll find the internal frame rate counter in Dungeon Siege shown in the top left corner and the FRAPS frame rate counter in the top right corner. Although the frame rates are being calculated at different time intervals, they are generally consistent with one another.
The following results are based on 32-bit color using resolutions of 1024x768, 1280x1024, and 1600x1200, which aren't officially supported by Dungeon Siege. Sound was enabled with complex party shadows, trilinear filtering, and maximum object details.
Based on the last time I tested Dungeon Siege, these results show a significant increase in performance across the board - especially with EAX enabled. The minimum frame rate without EAX for all three resolutions was 31 frames per second, which occurred in the final scenes as shown in the bottom right screenshot. With EAX, the minimum frame rate ranged from 27 to 30 frames per second.
With EAX enabled at a resolution of 1024x768 and 4XS antialiasing, the minimum frame rate was 25 frames per second while the average frame rate was very close to 60 frames per second - 59.598 officially.
The thought of using Warcraft 3, which is Blizzard's first real-time strategy game to use three-dimensional graphics, as a benchmark may not sound as appealing as using the latest first person shooter. But a case can be made that it's equally demanding on a system's capabilities when using high quality graphics, 3D sound effects, and background music. With all of these factors coming into play, along with the intense battles that often take place, Warcraft 3 makes for a capable benchmark when measuring overall system performance.
It took me two days of searching for and playing back a number of replays before I left it to the pros and settled on a two against two clan match, which took place on the Tranquil Paths map. Unfortunately, the replay is based on version 1.2 of Warcraft as version 1.3 was released while I was close to completing this update.
The replay contains three distinct battles, beginning at the 6:30 mark and ending at the 8:30 mark, which qualify as "heavy-duty" with the third encounter being the most intense. The following are step-by-step instructions on how I used the replay as a benchmark.
After starting the replay, I selected player "aD-Fei[dC]" using the replay vision selection.
The replay speed was increased to 8X.
At the 6:10 mark, the replay speed was changed back to normal or 1X.
The hero Centrius (Level 1 Keeper of the Grove) was selected by clicking on his portrait, which is located at the top left of the screen. This causes the portrait to appear in the portrait window at the bottom of the screen.
I positioned the mouse over the portrait window and pressed and held the left mouse button. This causes the replay to be played back with the selected hero always appearing in the center of view. The left mouse must be held continuously during the benchmark.
At the 6:30 mark, I enabled frame rate logging in FRAPS.
At the 8:30 mark, frame rate logging was disabled, which concluded the benchmark.
The following scenes show the replay at various time intervals. The top left is the starting point, while the top right shows the first major battle, which takes place at the 6:47 mark. The bottom left scene is where the minimum frame rate for all the benchmark runs was recorded, which occurred when the battle group warped back to its home base. Although the frame rate shows 77 frames per second, I wasn't quick enough to capture the screenshot before the framerate counter was updated. The last scene is from the final, and most intense, battle during the benchmark and occurs around the 8:20 mark.
The benchmarks were run in 32-bit color using resolutions of 1024x768, 1280x1024, 1600x1200, 1920x1440, and 2048x1536 with all graphics options set to their highest level. Let's take a look at the Warcraft 3 documentation for the graphics options with the exception of shadows, which wasn't listed.
Resolution - This setting has the greatest impact on overall graphic performance. This option adjusts the resolution and color depth of the game display. The lower your resolution, the faster the game will perform as fill rate decreases.
Model Detail - This option adjusts the number of polygons per unit. There are three quality settings: LOW, MEDIUM, and HIGH. Setting this option to LOW reduces overall number of polygons and can increase performance.
Animation Quality - This option adjusts animation blending and animation smoothness. There are three quality settings: LOW, MEDIUM, and HIGH. Setting this option to LOW increases performance and reduces memory use. When running on a system with 128 MB's or less of system memory we recommend setting this option to LOW.
Texture Quality - This option adjusts texture size and texture color bit depth. There are three quality settings: LOW, MEDIUM, and HIGH. LOW reduces texture size by one step and uses 16-bit color depth. MEDIUM uses full size textures and 16-bit color depth. HIGH uses full size textures and 32-bit color depth. Setting this option to LOW increases performance on low memory graphics cards. We recommend ONLY using the HIGH texture quality setting on systems with 64 MB or better graphics cards.
Particles - This option adjusts the number of particles displayed. There are three quality settings: LOW, MEDIUM, and HIGH. LOW draws 30% of the total particles. MEDIUM draws 65% of the total particles. HIGH draws 100% of the total particles. We recommend reducing your particle setting if you are experiencing slow downs during large battles.
Lights - This option controls the number of lights rendered per object. There are three quality settings: LOW, MEDIUM, and HIGH. LOW uses one light. MEDIUM uses four lights. HIGH uses eight lights. If you have an older graphics card with 16 MB or less memory we recommend setting this option to LOW.
Here are the benchmark results with sound effects and music playback enabled using the Dolby Surround option.
I also tested using the highest quality of antialiasing available on the GeForce4, which is 4XS, at a resolution of 1024x768. That setup yielded an average frame rate of 39 frames per second with a minimum of 21 and maximum of 55 and is similar to the non-antialiased results at the 1920x1440 resolution.
An example of Warcraft 3 running at a resolution of 1920x1440 can be viewed by clicking the thumbnail on the left. Note that the full size image is 680KB. The downside of using antialiasing in Warcraft 3 is that text is somewhat blurred which can be seen by clicking the right thumbnail.
However, the "icing on the cake" in this exercise can be attributed to the Dolby Surround and EAX2 sound configuration options. Having played using a 4.1 analog speaker setup (Logitech Z-560) and with closed-ear headphones (Koss TD-61), both options offer convincing audio effects, which make it a difficult decision to select one over the other. However, those of with an existing digital speaker setup and are contemplating upgrading to nForce2 will be in for a audio treat as Warcraft 3 is one a a handful of games that support Doby Surround Sound. Another interesting and new technology recently introduced by Dolby is Dolby Headphone.
ABOUT DOLBY SURROUND
The following except was taken from the Dolby web site, which is a brief description of Dolby Surround Sound.
Dolby Surround Multimedia creates the Dolby Surround listening experience specifically for personal computer users. Spatial effects can be reproduced by controlled directivity surround speakers aimed out of the sides of the front speaker systems, or through "virtual surround" speakers created with DSP 3-D processing. Unlike the full multispeaker system needed for proper listening area coverage with a home theater, these new techniques are ideal for creating a personal surround experience.