| Testing Methodology - Video Cards |
|
| Written by Chris Hunter | |||||
|
We want to present a real-world picture of what a graphics card will do for the end-user. BHFO uses gaming, synthetic, video, and desktop performance to cover the different facets of each card. Until we have multiple cards to compare against, we will focus on the end user experience and a wealth of information. We expose the strengths and weaknesses in different applications. Our analysis of what end-users would benefit the most from purchasing the graphics card is available at the end of the review. BHFO tests Vista (DX10) and XP (DX9) gaming performance by picking out the latest in software titles and finding reproducible game play areas. We run FPS data captures for 300 seconds, or five minutes of real-world game play utilizing fraps 2.93. This portrays a more realistic view of the end user experience. If possible, we plan our game run so actions or real-time cut-scenes occur in the same manner and time frame in each data set. BHFO tests 800x600, 1024x768, 1280x1024, and 1600x1200 resolutions due to a lack of official HD display capability. Each game runs at the maximum settings available unless otherwise noted. We then take the resulting data sets and compile a graph showing the gaming experiences at each resolution over time. We also replicate the same testing and graphs for all DX10 enabled games in Vista. If the video card will benefit from the use of AA and AF, we repeat testing to show performance hits experienced with different levels of filtering and anti-aliasing. BHFO performs an additional test run and records a video of the approximate game play experience to correspond with each set of graphs. This example allows users to see what the GPU is handling during the time-span of the graph. If a particular game does not have reproducibility or is repetitive in nature, such as UT3 or DIRT, BHFO shows half of the game run and then switches to extra game play footage to offer alternative coverage. We rely on 3DMark06 for synthetic benchmarking. BHFO runs the standard benchmark with all tests enabled. A full 3DMark06 test is run for significant and stable overclocks as well. Performance gains are then comparable to stock levels. We usually have a reference system on hand to offer additional comparison points. Unfortunately, the tremendous amount of time required to test and compile a full set of data does not fit within the time we have with the system. BHFO collects individual game test data in 3DMark 06 and presents the results for the reference system. This will offer an alternative performance viewpoint to compare to our main test bed. Thermal testing is part of our analysis. We run a full 3DMark test, excluding the CPU tests, to heat the GPU. After the 10min 51s of GPU load time, the system sits until the GPU returns to idle. The video card completes a thermal cycle when the card returns to idle temperature. Judgments on the efficiency of the cooler are based on the amount of time this test takes. We run the card through an additional thermal cycle to further test the GPU cooler if a stable overclock is possible. Lastly, BHFO analyzes high-definition playback and desktop performance. For HD content, we use BBC's "Planet Earth: The Complete Series" on HD DVD. The 81GB of video content and 20Mbps streams are a great way to judge video performance. Vista Ultimate with Aero is employed for desktop performance.
BHFO provides all of the content we generate
during testing in a torrent for distribution prior to the release of a review.
Each game has a dedicated tabbed Excel file, graph imagery, screenshots, and
higher quality videos. We also include the Excel files and images relating to
the thermal and 3DMark06 testing. Releases will be available on our torrents
page by logging in. The review torrents will only grow in quality and detail as
our community grows.
              
Powered by !JoomlaComment 3.12 Copyright (C) 2007 Alain Georgette / Copyright (C) 2006 Frantisek Hliva. All rights reserved. |
|||||
