Memory Tests
As always, memory-based benchmarks let us get a quick overview of the performance of the memory subsystem on our test system. In the case of the Albatron (the dark purple bar on the graphs), we're looking for competitive performance with the i865PE+PAT and SiS655TX (incase you haven't read my review of the ASUS board carrying that chipset, it's very fast), justifying the expense a Canterwood board carries over those cheaper rivals.Pifast first, to see what memory-based performance is like initially.
![](/mobile_image/?url=http://img.hexus.net/albapx875p/images/pifast.png)
Being slower than the i865PE+PAT and 655TX boards in Pifast is initally worrying. It's not much, but it's measurable. Let's see what Sciencemark tells us about available memory bandwidth and main memory access latency.
![](/mobile_image/?url=http://img.hexus.net/albapx875p/images/science_band.png)
A shade less measurable memory bandwidth goes some way to explaining the Pifast defeat, access latency will paint the rest of the picture.
![](/mobile_image/?url=http://img.hexus.net/albapx875p/images/science_latency.png)
It's only a millisecond or so of extra access latency, when doing Sciencemark's memory read task, but it's a fair bit slower than our new memory controller king on P4, SiS655TX. In fairness, the DFI's Springdale-PE + PAT combination is a pretty speedy Canterwood impersonator, so we're pretty much spot on in terms of memory performance with the Albatron, but I wasn't expecting it to place dead last when I was thinking about the results.
System level performance now with some CPU-heavy tasks.