MSI FX5600-VTDR128

• MSI KT4-Ultra BR (VIA KT400 chipset)
• AMD XP1800+ (overclocked to 166×10 = 1660MHz or “XP 2000+”)
• 512MB PC2700 RAM (333MHz DDR FSB)
• MSI FX5600-VTDR128
• Western Digital 120GB WD1200BB 2MB Cache
• Turtle Beach Santa Cruz soundcard
• LinkSys LNE-100TX v5 LAN adapter

Software Configuration:

• NVIDIA Detonators 44.03
  • Set to “Performance” except for UT2K3 benchmarks (“Quality”)
  • No texture sharpening
  • Coolbits installed for overclocking tests.
• Windows XP Pro with SP1 for Unreal Tournament 2003 benchmarks.
• Windows 2000 with SP3 for the rest.

Benchmarks:

• First-person Shooter – Unreal Tournament 2003 (v2199)
• Real-time Strategy – Age of Mythology
• Role Playing Game – The Elder Scrolls III: Morrowind
• Simulator/Action – Freelancer
• Theoretical Benchmarks – 3DMark 2001SE (build 330), 3DMark 2003 (build 3.1.3)

My Benchmark Selection
The best way to understand the limits of a video card is to test it with real-world games that are currently being played. I will not emphasize theoretical benchmarks or legacy benchmark games, as they often don’t translate into something useful for the average consumer. Being a moderator here at AMDForums, I have a sense of what is being played and what people are looking for when making purchasing decisions. People often upgrade based on performance of their current game – if their “Game X” is too slow, then it’s time for an upgrade. I’ve tried to take a game sample that is representative of that situation.

However, with that said, I have included 3DMark 2001SE and 3DMark 2003 benchmarks for those of you who want to compare performance against other systems.

Benchmark Numbers
Where applicable I have categorized benchmarks into a resolution group: 1024×768, 1280×1024, or 1600×1200. Each of these groups has a high and low detail test and each test includes combinations of anti-aliasing and anisotropic filtering options.

The average FPS is important as it tells us what we can typically expect from the card running at that level of detail at the given resolution. The minimum FPS number is also critical as it shows us how poorly the card can run. The maximum FPS isn’t all that important here since in a typical game you will not be staring at a wall, instead you will be in the middle of the action — therefore the lowest and average FPS are most useful to us.

Nomenclature

• AA, Anti-aliasing
  An algorithm that gives the illusion that an edge in 3D space is rendered with a finer grid than in reality (aka. “removes the jaggies”). AA is more important at lower resolutions since a rendered line has fewer pixels and therefore more jaggy. Higher the AA sampling level, the nicer and sharper edges appear, but usually at the cost of the hardware’s performance.

• ##xAA
  Notation used to indicate an antialiasing sampling level. i.e. 2xAA means “Two times antialiasing sampling level”.

• AF, Anisotropic Filtering
  The sharpening of textures as it recedes away from the viewer. Higher the AF sampling level, the nicer and sharper textures appear in the distance, but usually at the cost of the hardware’s performance.

• ##xAF
  Notation used to indicate an anisotropic filtering sampling level. i.e. 2xAF means “Two times anisotropic filtering level”.

• AAxAF
  Notation used to indicate an anti-aliasing/anisotropic filtering pair. Example, “2×4” means that anti-aliasing is set to 2x and anisotropic filtering set to 4x.

• FPS, Frames per second
  The number of frames rendered by the video hardware per second. Higher the number, the smoother things appear. Ideally you want this number to be as high as possible, but in some cases consumers prefer to have higher details for a nicer image and sacrifice some performance.