Maybe I spoke too soon. We begin to see a little weirdness as we hit Sandra's all too common memory benchmark. Look at the huge differences on the 2.4GHz Xeons. With hyperthreading enabled, there appears to be a ~400MB/second penalty in performance. The testing on the 2GHz Xeons appear to be a wash.

Cinebench's raytracing has always been a favorite with us as it's an excellent testing of processor and memory performance. We've always noticed that Intel's shared-bus Xeon platform has scaled almost perfectly when going from the single-cpu to the multi-cpu testing. It appears that hyperthreading may have been playing more of a role in this fact that we originally expected. Disabling hyperthreading caused a dramatic decrease in performance.

Let's transition into some real-world testing; LAME, MPEG1 encoding and DivX encoding.

This is exactly what i expected to see from the LAME mp3 encoder. Of course LAME is not multithreaded and as such hyperthreading shouldn't have any impact whatsoever. Heck, having two physical processors isn't going to make a difference either. Not dramatic, but we decided to include it anyways as it's something we've grown accustomed to including in all of our motherboard reviews.

Video encoding is all the rage these days with geeks everywhere spending far too much time buzzing through VCDs, SVCDs and even the occasional DivX. In our MPEG1 encode performed in TMPGEnc, we see that again hyperthreading seemingly offers some performance advantages. The time factor here is relatively substantial considering this is a rather short encode.