Benchmarks I
Starting off, as always, with SANDRA's synthetic unbuffered memory benchmark to see how our motherboards fare with bandwidth that takes no optimisations into account.
It's a little surprising that both i845-powered motherboards lag behind in this benchmark, especially as they're running the system RAM at the same speed as the P4X400 and SiS648. However, one benchmark a review does not make.
On to Pifast. It simply calculates the constant Pi to 10 million decimal places. Bandwidth, and lots of it, is required here.
Even though the MSI motherboard is running the exact timings as the ABIT IT7-MAX2 v2.0, being of the same chipset, it still falls around 0.4s behind. Meaningless in real terms, that kind of difference is only highlighted in benchmarks. No motherboard is poor here.
Next is DVD-to-DivX encoding. DivX4.12 with a 2-pass encoding of Gone in 60 seconds with an 1800kb/s bit rate. An average is calculated when the first VOB is complete. The times taken for each pass are different, and an average is calculated from both. DivX encoding is another benchmark that benefits from generous amounts of memory bandwidth.
Excluding the impressive benchmark laid down by PC1066 RAMBUS, you could throw the proverbial blanket over the DDR333-based motherboards, with 3 different chipsets making up the protagonists.
We next ran the OCUK SETI benchmark. The work unit sifts through data in the hope of finding some inkling of E.T existence, It takes a while to complete. You can download this benchmark from here. Simply click on the runbench.bat and wait a while. The result will be contained in the result.sah file after completion. It will give you a final time in seconds.
It appears that you can literally substitute one motherboard for the other and not notice the difference in performance terms, such is the similarity of results obtained. Chipset designers certainly seem to know exactly how to extract excellent bandwidth from their respective memory controllers. RAMBUS running at PC1066 speeds wipes the floor with DDR chipsets here. We've been measuring times in seconds, but now wee see a difference of almost 15 minutes in a single unit. If you want the best possible performance, at stock speeds, and irrespective of cost, go with super-fast RAMBUS.
