facebook rss twitter

Review: MSI SiS645DX 645E MAX-LRU

by Tarinder Sandhu on 2 June 2002, 00:00

Tags: MSI

Quick Link: HEXUS.net/qals

Add to My Vault: x

Benchmarks I

We'll start of with Sisoft Sandra's memory benchmark. Although a purely synthetic benchmark, it does give us a broad idea of how each chipset handles streaming data.

First aspect to note here is that the two top benchmarks are not directly comparable to those below. They are both running at 2133MHz / 133FSB, whilst the others are at 1600MHz / 100FSB. They're simply included to illustrate the expected performance at 133FSB, as both chipsets are officially validated for Northwood B use.

The MSI 645DX outdistanced the other two DDR-based chipsets when run at 100FSB, the SiS645DX offers PC2700 support whilst the I845E and P4X266A are still only officially supporting PC2100 memory. The I850 is the leader of the pack at 100FSB. At 133FSB, however, the I845E can run memory at 177MHz (3:4 FSB:Memory ratio) whilst the MSI 645E Max-LRU can only run at 166MHz (4:5 FSB:Memory ratio), this is why the I845E streaks ahead. Note, however, streaming throughput as measured by Sandra increases for the MSI 645 Max-LRU when going from 100 to 133FSB even though the memory speed remains constant (166MHz).

Let's see how this throughput translates to practical applications as we turn to Pifast. Pifast simply calculates the constant Pi to X million places. It relies on memory throughput as well as sheer CPU power. We've set it to calculate to 10 million places using the fastest method.

The MSI continues to be the leader of the DDR pack at 100FSB, courtesy of its PC2700 utilisation, it still cannot overhaul RAMBUS, however. At 133FSB, the increased bandwidth on offer for the I845E sees it sneak a slight lead on the SiS645DX. The 645E Max-LRU puts up an impressive showing in this our first practical benchmark.

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.

We see the CPU-dependent nature of MP3 encoding, differing platforms offer no discernable advantage over one another, pure clock speed counts here. That's why we see the 2133MHz / 133FSB benchmarks take a substantial lead. If MP3 encoding interests you greatly, simply purchase the fastest processor possible.