MP3, DVD, Kribi, Raytracing, HDTach
It should do better in the computational activity that is WAV-to-MP3 translation.
Here's one case where the otherwise excellent Athlon 64 3200+ struggles. All the bandwidth in the world isn't going to change the glaring fact that this is a MHz-driven activity. Still, the ST61G4 comes in at around 2s behind the SB62G2 Springdale i865G XPC. The test was to encode 607MB of U2 WAV files into 192kb/s MP3.
Realstorm's Raytracing benchmark is a benchmark that relies on memory bandwidth to a lesser degree than, say, Pifast. It has always favoured the Athlon CPUs; the Athlon 64 3200+ positively romps away here. We find ourselves repeating what's been said before. Lack of latency and bandwidth debilitate the ST61G4.
KribiBench, on the other hand, prefers the P4's way of doing things. It's not just a benchmark for the sake of it, either. The software renderer has projects with polygon counts that number in the billions. We use the JetShadow model with a realistic setting. You'll now appreciate why the ST61G4 is slower than the SB62G2.
The Silicon Image 3512 2-port controller's SATA ports almost begged to be exercised by dual W.D Raptors in RAID0. We duly obliged.
Hmmm. Great CPU utilisation and decent random access speed is overshadowed by the poor average transfer read and burst speed. There's no real explanation as to why it's so slow. It's as if the controller isn't running at peak capacity. Burst speeds should be far higher than the average read speed. The graph below shows what the two drives managed on a VIA VT8237's SATA ports.
Ignoring the high CPU usage for a second, note the far greater performance. We find it difficult to explain the shortcomings of the Sil3512 graph.
