Benchmarks I
ScienceMark 2.0 provides some interesting memory bandwidth and latency information. Both of these factors are crucial in determining performance with respect to memory-intensive applications and games. Let's have a look.
The SiS655FX starts off pretty well in the memory bandwidth examination test. It's just a hair's breadth behind the Intel dual-channel duo, albeit with a slight clock advantage. The single-channel Athlon 64 3200+ puts a damn good showing for a chipset running with only half the available bandwidth.
An on-die memory controller does have some practical use, as ScienceMark 2.0 graphically points out. The 655FX produces another decent showing, suggesting that the ensuing benchmarks should be relatively close between the Intel-based trio.
Seems that way, doesn't it ?. Pifast is our resident, real-world memory test that calculates the constant up to 10m places. Version 4.1 is being used to keep parity with older results. Just over a second behind the Springdale and 1.7s behind the ABIT Canterwood. The 655FX benchmarks Pifast in a time analogous to a non-PAT-enabled Springdale.
The Realstorm Raytracing benchmark is an AMD benefit, however a greater amount of usable bandwidth always makes a difference, no matter how small. That's why the Intel trio are separated by no more than 0.4 FPS. SiS' implementation seems to be holding up well thus far.
