Benchmarks I
On the face of it we'd be surprised if there were large discrepancies between the Canterwoods' performance. We'd expect a little variation based on how well tweaked the boards' BIOSes were and variations arising from differing running speeds, but one Canterwood should perform much like another.
A quick look at the SiSoft SANDRA buffered benchmark. A dual memory channel and quad-pumped FSB should see the Canterwood clear in this test.
Pretty much what one would expect. Don't be fooled by the comparison XP3000's performance. The ensuing benchmarks will mark out its prodigious potential. The Canterwoods seem to bunch at around the 5GB/s level; ~20% below their theoretical maximum. Taking away prefetching showed the Canterwoods in an equally impressive light, with all three managing ~ 2500MB/s.
Running Pifast v41 next. It simply calculates the constant Pi to the desired number of places. 10 million is chosen for this test. If you want to run it for yourself, click here for the benchmark standings and download link. Just unzip and click on the .bat file.
4 chipsets and 3 processors are separated by less than 5 seconds. The Asus, by virtue of its higher operating speed, relatively low latency timings, and comparatively mature BIOS takes the opening benchmark. The MSI, with considerably higher latencies, does well to finish within 1.5 seconds behind Asus' time. Even an 'untweaked' Canterwood is faster than a fully tuned i845PE. The reference board is surprisingly quick in this test, a fact that's even more impressive considering it's running at exactly 3000MHz. I'd expect subsequent BIOS releases to generally speed-up retail Canterwood boards.
Now MP3 encoding. We're benchmarking by encoding a 610MB custom WAV file (U2's Pop album, incidentally) into 192kb/s MP3 using the LAME 3.91 encoder and Razor-Lame 1.15 front-end.
Very little differences between the Intel protagonists here. The XP3000, running at 2.17GHz, just doesn't quite have the horsepower to compete with the 3GHz P4 in MP3 encoding. A higher work per clock cycle ethos only gets you so far if you're almost 50% behind in pure MHz stakes.
SETI, though, loves the AthlonXP. Running the Hexus SETI benchmark with a 0.417WU.
3 minutes separates the Canterwoods, but the i845PE is almost 15 minutes behind the Asus P4C800D. It seems as if SETI really appreciates the extra bandwidth that dual-channel 200FSB memory offers. We're still eagerly awaiting the first stock sub-2 hour clocking.
