A fast integrated L2 cache benefits the performance in business applications but 3D applications are utilizing data sets too large to fit into a 256 KB L2 cache. In fact, the original performance of the Thunderbird in Expendable was less than spectacular. Using a fresh installation of the operating system usually qualifies for being on the safe side, however, there are exceptions. In this case, I used Expendable (go.exe) already installed on a different partition only to be stuck with frame rates of around 85 fps. This is pitifully slow compared to the 105-115 fps obtained with an Athlon (classic) 900 overclocked to 968 MHz or the average of 105 MHz achieved with the PIII Coppermine (166 MHz FSB x 6, SDRAM at 166 MHz, CAS-2). However, a recent report at Planet Hardware showed about the same poor performance for the Thunderbird.
To make a long story short, the VIA KT 133 chipset and the Athlon abide by other rules than the Solano / PIII. Deleting the entire Expendable folder and reinstalling the benchmark including setup, to configure the game according to the VIA GART AGP drivers, yielded a more than 20% increase in the frame rates. And the morale of this story ...
Expendable frame rates at 1 GHz and 1070 MHz At first glance, these scores don't appear to be too overwhelming. Keep in mind though that they were achieved with only a standard GeForce SDR. The memory bus was set to 133 MHz/142 MHz, 2:2:2 latency settings.
Quake3 Arena Demo 001
Even though Quake 3 Arena is heavily SSE optimized, the Tbird and the ASUS A7V don’t need to shy from any comparison.
Quake3 Arena Demo 001 frame rates at 1 GHz and 1070 MHz using HSDRAM and VCM Something interesting happens here. At low resolution, VCM is markedly slower than stadard SDRAM, however, at high quality (32 bit color), VCM suddenly takes the lead. Even though the difference is only about 3-4%, the increase is remarkable since the MSI MS 8809 was never before able to reach these scores. The benchmark results indicate two things: First, there appears to be enough randomness in the locality of the data stored in memory to warrant a drop in performance when using VCM. Second, at high resolution, the integrated SRAM cache in the VCM DIMMs can take advantage of the larger, coherent data chunks resulting from 32 bit color depth
General disclaimer: This page only reflects the author's personal
opinion and assumes no responsibility whatsoever regarding any of the contents
or any damages that may occur explicitly or implicitly from reading the
contents of this site. All names and trademarks mentioned in this review are the exclusive property of the respective parent companies.
All contents of this site are protected by international copyright laws. Reproduction of the contents
even in parts is not allowed except after written permission by the author and referral to this site.
Copyright 2000 LostCircuits
