Benchmarks I
We'll start of with Sisoft Sandra's memory benchmark, this should be interesting with on-board video sharing main system memory.
You can see the devastating effect that the on-board video has to streaming memory performance. According to SANDRA, we lose about 20% of usable bandwidth by having to accommodate the on-board SiS315 graphics. The situation with the SS51G running a Ti 4200 is markedly different, however. It matches the similarly Ti4200-equipped ABIT i845E BD7II.
How do they fare in Pifast, a simple program that relies on bandwidth to calculate the constant Pi to X amount of places. I've set it to calculate 10 million decimal places by the fastest method possible. Seeing as the on-board graphics take a memory hit, I'd expect it to be the slowest here.
As expected, lower bandwidth directly translates into lower scores. So, having on-board graphics can make a distinct difference to non-3D applications.
Next we'll turn our attention to MP3 encoding. We're benchmarking by encoding a 638MB custom WAV file into 192kb/s MP3 using the LAME 3.92 encoder and Razor-Lame 1.15 front-end.
LAME encoding, although being a CPU affair, does require a reasonable amount of memory bandwidth to function properly. The fact that we are repeatedly two seconds slower with the on-board graphics is testament to this fact.
