Skip to main content

A Solid Multimedia Notebook From HP

3DMark05 Benchmarks

As its name implies, FutureMark's 3DMark05 test suite focuses on the graphics capabilities of computers. There are three basic benchmarks: Games tests, CPU tests and feature tests. Within the Games tests there are three subtests, each of which simulates different aspects and levels of complexity in various graphical gaming environments. Return to Proxycon is a classic shoot-em-up type game set in a futuristic environment. Fire Forest, the most complex graphically, involves a firefly moving through a dense and very ecologically diverse forest. In Canyon Flight, a powered sailing ship that flies through the air suspended under a balloon encounters a gigantic dragon while navigating though a graphically complex canyon. 3DMark05 reports an overall Games tests score, as well as the average number of frames per second achieved during each subtest.

The CPU tests measure CPU performance using specific segments from the Games tests. There are two CPU subtests, the results of which are reported as an overall score. Frames per second are also calculated for each of the two subtests.

The Features tests provide detailed measures in the areas of single- and multi-texturing fill rate, pixel shading and simple and complex vertex shading. These tests are valuable, but we feel that the Games and CPU tests adequately summarize the performance of a computer's graphics subsystem.

We report the summary scores for the Games (the 3DMark05 Score) and CPU tests. Additionally, we report the results for the three gaming scenarios in frames per second.

Because computer graphics systems and supporting components tend to respond differently depending on the resolution of the display being tested, we run the 3DMark05 Games and CPU tests at both the graphics system's highest resolution and a resolution of 1024x768. Because it takes less graphics power to handle a display set for 1024 x 768, than one running at the highest resolution, scores are generally higher at 1024x768 than at the computer's highest resolution. Also because highest resolution can differ for different computers a common resolution of 1024 x 768 provides a standard allowing for better comparison between different computers and their graphics controllers.

As with the PCMark05 tests, the 3DMark05 Score and CPU Score were scaled to reflect scores obtained by what were considered entry and high-end systems at the time 3DMark05 was developed. The range at that time was 1000 to 5000. Like PCMark05 scores, 3DMark05 scores have reached beyond 5000 as computer and graphics hardware has gotten better and faster. Scores lower than 1000 have also been measured. As with PCMark05, we will adjust the score range on our charts as we obtain test results beyond 5000.

Some mobile computers do not have graphics subsystems that can support the 3DMark05 tests. Others can support the tests, but perform poorly because their graphics subsystems and processors are slow. These often achieve 3DMark05 scores in the 200 - 600 range and frames per second rates that hover around 1. In this case we report but do not chart the overall 3DMark05 score and note that the particular computer under test is not suited to more demanding 3D graphics applications.

As with the other three test suites, we run 3DMark05 on Windows XP SP2. Like PCMark05 it requires Microsoft Media Player 10 and Microsoft Media Encoder 9. We use the same system and Windows settings as for SYSmark 2004 SE; you can find a reference to these in the section "Preparing for and Running SYSmark 2004 SE."

The following are 3DMark05 test result charts for the HP Pavilion DV4000 and Toshiba Qosmio G25. None of the four laptops is included in the test results. One of the four lacked a graphics controller adequate to run the 3DMark05 tests. The other three had the necessary graphics hardware. One of these lacked the display resolution to support the tests. The other two produced very low 3DMark05 scores and are thus excluded from the charts. I offer my conclusions after the charts.