Jason Dunn
09-18-2007, 09:30 PM
In order to draw accurate comparisons when testing hardware, I wanted to outline two of the tests I've devised to benchmark CPUs, RAM, hard drives, and other hardware. All tests are run on Windows Vista.
Video Transcoding Test
Using the trial version of Nero Recode (http://www.nero.com/nero7/eng/Nero_Recode_2.html) 2.5.5.0 - because it's a fast, multi-core optimized program - I transcode a decrypted VOB file pulled from Lord of the Rings: The Fellowship of the Ring. The file is 2 hours 58 minutes in length and is 6.98 GB in size - it's the entire movie put into a single VOB file, and represents a typical task of ripping a DVD and converting it into a useful format with the variable of the optical drive removed from the equation. Using Nero Recode, I selected the Nero Digital Category > Nero Digital Profile: Standard, set to 2 mbps. The default of Real-time encoding (1-pass). The processing time set to ABOVE NORMAL and the video preview is disabled. No other programs are running.
RAW Photo Encoding to JPEG
Using the demo version of DXO Optics Pro v.4.2, I select a test batch of 20 RAW images. I use a trial version of DXO Optics Pro to process the RAW images in batch mode using the DXO default settings: each image is processed for white balance levels, sharpness, color correction, contrast, noise reduction, optics distortion, and several other things. The default settings of "Minimum processing power for DxO Optics Pro" is left alone - it still uses 100% of the CPU resources. This is a real-world test because DXO is set up to allow you to quickly adjust each RAW file with exactly the changes each image needs, then to process them all to JPEG, TIFF, or DNG in batch mode. This is a brutal test and it’s almost entirely CPU based. I've done some experiments with this test running at different clock speeds on the same system, and if I scaled the CPU and RAM speed back by 5%, I saw the benchmark take 5% longer to complete. Further, DXO Optics Pro is multi-threaded and runs beautifully on dual-core systems (I don’t have a quad-core CPU so I’m unable to test that scenario).
I also use other benchmark tools such as 3D Mark 2006, PC Mark 2005, HD Tach, and PassMark Benchmark.
Video Transcoding Test
Using the trial version of Nero Recode (http://www.nero.com/nero7/eng/Nero_Recode_2.html) 2.5.5.0 - because it's a fast, multi-core optimized program - I transcode a decrypted VOB file pulled from Lord of the Rings: The Fellowship of the Ring. The file is 2 hours 58 minutes in length and is 6.98 GB in size - it's the entire movie put into a single VOB file, and represents a typical task of ripping a DVD and converting it into a useful format with the variable of the optical drive removed from the equation. Using Nero Recode, I selected the Nero Digital Category > Nero Digital Profile: Standard, set to 2 mbps. The default of Real-time encoding (1-pass). The processing time set to ABOVE NORMAL and the video preview is disabled. No other programs are running.
RAW Photo Encoding to JPEG
Using the demo version of DXO Optics Pro v.4.2, I select a test batch of 20 RAW images. I use a trial version of DXO Optics Pro to process the RAW images in batch mode using the DXO default settings: each image is processed for white balance levels, sharpness, color correction, contrast, noise reduction, optics distortion, and several other things. The default settings of "Minimum processing power for DxO Optics Pro" is left alone - it still uses 100% of the CPU resources. This is a real-world test because DXO is set up to allow you to quickly adjust each RAW file with exactly the changes each image needs, then to process them all to JPEG, TIFF, or DNG in batch mode. This is a brutal test and it’s almost entirely CPU based. I've done some experiments with this test running at different clock speeds on the same system, and if I scaled the CPU and RAM speed back by 5%, I saw the benchmark take 5% longer to complete. Further, DXO Optics Pro is multi-threaded and runs beautifully on dual-core systems (I don’t have a quad-core CPU so I’m unable to test that scenario).
I also use other benchmark tools such as 3D Mark 2006, PC Mark 2005, HD Tach, and PassMark Benchmark.