Video Card: Matrox Parhelia Review :: Performance

I have seen all the tests and interpretations from other reviews. My point of view is a bit different. How many frames per seconds do we need? 25 fps is enough for motion effect but important thing is with worst case scenario, it has to be over 25 fps. When some object blows up, or sun shines through trees, fps drops dramatically. With the best quality settings and with Anti-aliasing over 45-50 fps might be safe. Not always but it is OK. We will go through some tests and interpret these results on this basis.

First I would like to briefly explain why I use anti-aliasing for worst case scenario 3D Graphics. Anti-aliasing is a technique to get rid of the staircase effect on object edges. The System first calculates that graphics on one resolution higher and samples it to existing resolution. You need a lot of muscle power to do that. All graphics cards chips have more intelligent methods other than sampling to smooth the edges, but even with those methods, cards choke on anti-aliased scenes.

Base System

• Pentium 4 2.0A CPU
• MSI 845E MAX2 BLR Motherboard
• 256MB Crucial PC2700 DDR
• 40GB 7200 rpm IBM Deskstar IC35L040AVVA07 Harddisk
• Windows XP Professional
• Desktop resolution: 1027x768@32 Bit 85 Hz and 1600x1200@32 Bit 85 Hz

Video Card Drivers
• MATROX - 1.00.04.231
• ATI - Catalyst 2.2 and Control Panel 6.13.10.3017
• NVIDIA - Detonator XP 30.82
• V-sync always off - Maximum Quality settings

I ran more than 100 tests using over 10 different testing suites. I will highlight some of them. The NVIDIA drivers worked flawlessly on all tests. I ran into severe problems with the ATI Catalyst 2.2 drivers. I don't know what are they doing or if they are reading this but they have had driver problems for years and it seems they have done nothing about that. I had to cancel the Specviewperf 7 test because the system froze many times with the ATI Radeon. I also had unstable results with Matrox and Specviewperf 7.

Sebastian, from Matrox, told me they are preparing special drivers for CAD-CAM software. I ran Quaver test on Quake 3 to show how cards perform on today's OpenGL game engines. The Quaver demo applies more realistic load to the system.

Unreal II Demo will be the game for 2-3 years. It represents the new generation of games. I ran Antalus Demo on Unreal II Demo. 3DMark 2001SE is for our readers to compare with their results and for testing Direct3D performance. Sysmark2002 test shows 2D Applications performance.

Quake 3 Quaver Demo 1024x768@85 Hz 32 bit Max Quality Settings: Most of the PC's on the market come with 17" monitor so 1024x768 is the sweet spot today. Without Anti-aliasing and Anisotropic filtering, all cards are way above limits. Matrox doesn't look so good but over 100 fps isn't good enough?

Quake 3 Quaver Demo 1024x768@85 Hz 32 bit with Anti-aliasing and Anisotropic Filtering: This test reflects today's games with best quality possible. Other than GeForce 4 -MX440, all cards may supply playable performance. Now Matrox shows some performance here.

Unreal 2003 Antalus 1024x768@85 Hz Max Quality: This will be the game engine for next couple of years, so results are reflecting your investment. Other than ATI all cards looks good with quality settings. Matrox is still at good playable limits.

Unreal 2003 Antalus 1024x768@85 Hz with Anti-aliasing: This reflects the best quality settings for the engine that will be effective for next couple of years. Other than GeForce 4 -MX440 and ATI all cards looks good. Matrox is in the second place this time.

Unreal 2003 Antalus 1600x1200@85 Hz with Anti-Aliasing: This is the worst case scenario: Best quality graphics, highest resolution and future engine. Guess what? Matrox is leading.

Sysmark 2002 1024x768@85 Hz: Results are too close but ATI and Matrox are ahead of NVIDIA cards on 2D as usual.

3DMark 2001SE 1024x768@85 Hz : 3DMark became as the standard Direct3D test on all over the World. Unfortunately Matrox performed worse than a GeForce 3.