Plus, as far as I could see, there were no graphical benchmarks (2D or 3D) so SPICE is irrelevant.
... Never the less, I do agree that virtio should have been added to the mix.
As the OP's post about disabling HT, I couldn't disagree more.
With 4 VM's, each with 2 till 8 vCPU's and 2-4GB RAM (Total vCPU's: 16, total memory: 10GB), I've seen noticeable performance increase on my Xeon 5530 workstation (8 cores, 16 threads, 12GB).
While I didn't bother to benchmark the setup, subjectively, w/ HT disabled, the machine would literally crawl to halt, while on the other hand, w/ HT enabled, I could continue working more-or-less as usual on the host.
Intel's Core7 / Xeon x5/x6xx HT implementation is -far- better than the P4 implementation.