Run those number by me again
To help better determine how much responsibility the slow test system had for the low score, I decided to run some quick performance tests after having carried out various tweaks. Unfortunately the motherboard in the test system is probably the least flexible board I’ve used, which severely limited any over-clocking.
System changes - and their effects
* Reduce the memory bandwidth available by reducing the memory-speed from 166MHz to 133MHz (this resulted in an overall drop in score of two points (-5.4%)
* Increase the CPU speed and memory bandwidth slightly by increasing the FSB to 170MHz (across two runs the overall resulted, effectively, in no change)
* Increase the CPU speed and memory bandwidth slightly by increasing the FSB to 174MHz (across two runs the overall result was effectively the same - although the overall score fluctuated by two points during runs, the individual scores only fluctuated by <1%)
Based upon these results it became clear that although the results are effected by differing PC specs, this only becomes significant with large changes in the performance of the CPU/memory.
This was later confirmed when I attended the Intel benchmarks briefing. A slide was shown illustrated that only one portion of the testing processes involved a purely performance-limited test (any scenario that “Prepares video for a portable media player”).
This test is performed in scenarios 5, 7, 8, and 11 from the level 1 (SDTV tests), scenarios 5, 7, 8, and 10 from the level 2 (HDTV tests), and scenarios 4, and 6 from the level 3 (DLNA uPNP Media-server tests).
Therefore I would expect the scores to increase/decrease more substantially on a system that is able to fully run all the tests - and to depend on the performance of the components tested.
I later tried to improve the scores by trialing a DLNA-compliant media-server program on the test system. This should theoretically improve the capabilities of the machine and thus boost the overall score.
I used version 1.4.0.7 of Nero Media@Home from Nero7 and then tried Cyberlink PowerCinema 4 (4.0.1813). Both products are claimed to be uPnP DLNA-compliant media-servers.
However, the Intel DHCAT benchmark still wouldn’t run any of the Level-3 (Connected) tests. This leads me to believe that either (a) a software incompatibility on the test system is preventing the benchmark program from working or (b) like the recording tests, Intel DHCAT benchmark is very fussy about what products it recognises and so didn’t acknowledge either the Nero or Cyberlink servers as compatible solutions.
