Benchmarks I
We'll start of with Sisoft Sandra's memory benchmark, one that highlights streaming data performance.
The first aspect to note here is the bottom three results are at 1600MHz (100FSB). The upper three with 133FSB are at 2133MHz (133FSB). This is solely to show how they handle both type of Northwoods, even though there is no 2.133GHz Northwood B. Quite expectedly, the MSI SiS645DX takes honours at 100FSB due to it supporting PC2700 memory out of the box, whilst the two I845E 'boards take it at 133FSB due to them being able to run memory at 1.33x FSB (177MHz) as opposed to the SiS' 1.25x FSB (166MHz).
We'll move on to our first practical benchmark, Pifast. Put simply, it calculates the constant Pi to x number of places. It requires an efficient memory controller for decent performance.
The 845E Max2 does well against the IT7 MAX, the 'board it shares a chipset with, at 100FSB. The major surprise is its relatively poor showing at 133FSB, getting on towards two seconds behind the IT7 MAX and even falling behind the 645DX. I've previously owned the MSI I845D Ultra ARU, the test 'board's immediate predecessor, and found the same result at 133FSB. It seems that MSI employ a rather strange divider system that negatively impacts on performance at 133FSB.
Next we'll turn our attention to MP3 encoding. We're benchmarking by encoding a 481MB custom WAV file into 128kb/s MP3 using the LAME 3.91 encoder and Razor-Lame front-end.
LAME encoding is an activity that is almost entirely CPU-bound, competing memory subsystems usually show little variance in results, this is the case here. Since our benchmarks measure to the nearest second, the results are very similar. Naturally, at 2133Mhz, the encoding time comes down significantly when compared to 1600MHz.