Benchmarks I
The lack of comparison systems makes comparisons somewhat limited. It would be remiss of us to compare the Scan system to a top-of-the-range XP3200-based PC, for the former isn't priced at truly high-end levels. Even so, we'll run through our gamut of benchmarks and try to spot any obvious performance failings. SiSoft SANDRA first.
An empty-ish graph doesn't help in deducing whether this is adequate or not. However if you note that this benchmark usually returns a result of 1420MB/s for an XP3200 Barton running at 200FSB with accompanying PC3200 memory in situ, the 1225MB/s, at a lower 166FSB, is just about where we'd hope and expect. The system was set up with 2-6-3-3 timings; that seems to reflect in the scores obtained here.
Previous XP2700-based systems have returned times of 187-189 seconds for our WAV-crunching benchmark. The test is to crunch over 600MB of U2 WAVs into 192kb/s MP3 format. Having a fast host processor will help one to convert their private collection into much-compressed MP3 form.
74.36 represents a reasonably good Pifast return for an XP2700 and PC2700 memory. The very fastest AMD systems, characterised by XP3200 and extremely low latency PC3200 memory, take a little over 63 seconds. Faster is always better, but 74.36s isn't too shabby for a system that's not at the top of Scan's proverbial tree.
And the speedy performance continues unabated with our 0.417 AR SETI Work Unit. A pre-built system that manages to knock out this average WU in under 2 hours and 30 minutes is more than capable of running everyday tasks. Systems are now at a stage where the usual PC uses, including e-mail, word processing, spreadsheets and basic image manipulation, barely tax a midrange system.
A simple HD Tach v2.61 sustained read speed test was run with generally decent results. The 120GB Maxtor SATA drive maintained an impressive 42.9MB/s average speed, and the graph was smooth and spike-free. However, the CPU utilisation of 48.7%, consistent over several runs, was a cause for concern. We'd expect a CPU utilisation of no more than 10%. A problem with NVIDIA's ATA driver or HD Tach ?. Evidence on the 'net seems to point to the former.