Have just updated to VMWare Server 2.00 GA from rc2 and for the first time have compared guest performance using two different host O/S on the same hardware.
The box has a quad core X5355 cpu @ 2.66GHz with 16GB RAM.
The guest config is Win2003 x86 (single cpu). The two host configs are Win2003 32bit and OpenSuse Linux 11.0 64bit with latest patches etc. Memory trimming switched off.
We used a benchmark tool in the guest called Sandra Lite2009.
Disk performance was comparable. (The virtual disks were kept on an NTFS partition used by both host O/S).
Memory-intensive performance was poor with the Linux host...most of benchmarks put it at about half the performance of the Windows host benchmarks.
The difficulty at this point is determining whether VMWare is to blame or if it's something that is not optimal in the Linux host config itself.
I'd like to be able to somehow compare performance of the two hosts, but across different O/S I'm not sure how you would do this.
Also I do remember seeing a thread in the Beta forum (which I can't find now) that criticised the VMWare behavior on Linux compared to Wiindows in respect of memory management...something to do with the disk-backed nature of the guest's memory footprint?
Appreciate any advice...this seems like too bigger performance penalty for taking the jump to 64 bit land. (I was more attracted to migrating onto a lean 64bit Linux kernel than going to Windows 2003 64bit.)