Soltek Qbic EQ3702M Mini Barebone System Review - Page 4 Of 7
By Mike Chambers - January 5, 2004
DIRECTX GRAPHICS
Benchmarks in this review were limited to measuring integrated and add-in card graphics performance. The integrated graphics processor (IGP) in Soltek's Qbic EQ3702M is based on NVIDIA's GeForce4 MX and supports the DirectX and OpenGL Application Programming Interfaces (API). The GeForce4 MX is a DirectX 7-class fixed function graphics processor with hardware-assisted acceleration of transform and lighting calculations.
The DualDDR memory architecture and 128-bit memory bus is capable of delivering 6.4GB/sec. of memory bandwidth when outfitted with dual-channel DDR400 memory. Memory bandwidth plays a major role in determining antialiasing performance, which is based on a multisampling implementation. Advanced texture filtering is limited to 2X anisotropic on the GeForce MX.
Integrated Graphics Information
The GeForce4 MX lacks the programmable features offered by DirectX 8-based shaders, which will result in vertex shader operations falling back to the CPU while pixel shaders will not be supported. Even if some form of pixel shader support had been incorporated into the integrated GeForce4 MX, the performance hit would have certainly restricted its use.
Generally speaking, the integrated GeForce4 MX should provide acceptable gameplay performance at a resolution of 640x480 in newer games, although there will be cases where in-game graphics settings may have to be reduced. Older games can be run at 1024x768 and at 800x600 while the use of 2X antialiasing and anisotropic texture filtering will be on a case by case basis.
NVIDIA GeForce FX 5700 Ultra, 128MB, 475MHz/906MHz
NVIDIA Forceware Graphics Driver Version 53.03
Sapphire Radeon 9800 Pro, 256MB, 378MHz/351MHz
ATI Catalyst Driver Version 3.10
32-Bit Color / Vsync Disabled / 60Hz Refresh Rate
Windows XP Professional with Service Pack 1 / DirectX 9.0b
BIOS Settings
Aggressive Settings
Expert System Performance
133MHz Front Side Bus
Memory Timing 5-2-2
CAS Latency 2.0
Synthetic Benchmarks
3DMark2001 SE
Games Tested
Call of Duty - Version 1.1
Dungeon Siege - Version 1.11
Halo: Combat Evolved - Version 1.031
Jedi Knight 2: Jedi Outcast - Version 1.04
Max Payne Demo - Version 1.05
Wolfenstein Enemy Territory - Version 2.55
Notes
All applications and games tested were patched to their latest version.
Unless otherwise stated, 64MB of frame buffer memory was allocated to integrated graphics.
Unless otherwise stated, antialiasing and anisotropic filtering were configured using the driver control panel.
Game benchmarks and gameplay tests were conducted with sound enabled with the exception of the Halo benchmark.
FRAPS was used to capture minimum and average (in some cases) frame rates.
Halo does not support antialiasing.
3DMark2001 SE Performance
The Second Edition of Futuremarks's 3DMark2001 synthetic benchmark included support for DirectX 8.1 and pixel shader version 1.4. Due to the lack of hardware pixel shader support, the GeForce4 MX is unable to run the Game 4 Nature test, which accounts for a portion of the overall 3DMark score. I opted to use the Game 3 benchmark as it was developed with Remedy Entertainment's MAX-FX Technology, which was also used in the first person shooter Max Payne.
3DMark2001 SE - Game 3 Test Screenshot
The benchmarks were run at a resolution of 800x600 and include the results from the low and high detail tests. Various graphics settings were used in order to reveal the impact that antialiasing (referred to as AA) and anisotropic texture filtering (referred to as AF) have on performance.
3DMark2001 SE - Game 3 Performance
The results show that 2X AA causes a greater impact on performance than 2X AF. The 40% drop in performance when using 2X AA is steep, which will limit the use of antialiasing on the integrated GeForce4 MX. This system scored 3929 3DMarks at the default benchmark setting of 1024x768 with no AA and no AF.
Image Quality
Warning: This section contains two high quality images. Each image is around 250KB in size.
The image below is a comparison of no AA and no AF against 2X AA and 2X AF from the Game 2 test. The benefit of 2X AF is shown by comparing the ground, building, and roof textures. A closer examination is needed to see the benefits of 2X AA, but they are visible. With the integrated GeForce4 MX, memory bandwidth and performance considerations will limit AA to 2X or Quincunx.
3DMark2001 SE - Game 2 Image Quality Comparison
Before moving on to gameplay, it should be mentioned that there might be cases where the image quality that is being rendered may not match the intended image quality. For example, the comparison below from the Game 3 test is such an example as the lighting on the GeForce4 MX is quite different than the DirectX 8 reference image.
3DMark2001 SE - Game 3 Image Quality Comparison
Image quality discrepancies can be caused by a problem with the graphics driver, the application, or even the application programming interface (API). However, we also need to be reminded that driver "optimizations" have been discovered in popular game and synthetic benchmarks that can result in decreased image quality.
Dungeon Siege Performance
Dungeon Siege was released in April of 2002 and is an action based role playing game. The game has been a huge success for Gas Powered Games and Microsoft as an official expansion pack was released and fans are looking forward to 2004 when Dungeon Siege 2 is scheduled to debut.
What makes Dungeon Siege playable with a variety of graphics settings on the integrated GeForce4 MX is the object detail setting. The object detail setting is controlled with a slider and for the purposes of this benchmark I assigned a minimum, medium, and maximum setting, which is based on the position of the slider.
Minimum object detail allows the player to see enemies that would have been otherwise hidden at a higher object detail.
Dungeon Siege - Minimum Object Detail
Moving from minimum to medium object details had a noticable affect on performance.
Dungeon Siege - Medium Object Detail
While moving from medium to maximum object detail did not affect performance as much.
Dungeon Siege - Maximum Object Detail
I measured performance in Dungeon Siege by conducting a partial walkthrough of the introductory Farmlands level using FRAPS. The walkthrough takes about three minutes to complete and consists of following a path through the woods while battling enemies using a sword, bow and arrow, and magic. The walkthrough concludes near the first pond I encountered. Simple shadows and trilinear texture filtering were enabled along with EAX sound effects.
Dungeon Siege Performance
I took a different approach with some benchmarks by first determining, and then testing, settings I felt were playable. The results are ordered from the highest average frame rate to the lowest.
The top three results averaged over 30 frames per second and indicate that resolutions of 800x600, 1024x768, and 1280x1024 can be used by varying the object detail setting. Even 2X AA is an option at 800x600 knowing that extra performance can be achieved by disabling shadows or using lower quality sound.
