Workstation 5.5 and Server 1.0 only officially support Vanderpool in long mode. Even on an EM64T system, Vanderpool instructions will only be used (by default) when the guest is executing in long mode. In particular, this means that 32-bit guests will not use Vanderpool instructions, even on an EM64T system. There is a monitor control flag to modify that behavior, but its use is not supported or recommended.

Thus, in a 32-bit guest, Petr's program should report 'LIMIT=FFBFFFFF', indicating that VT is not in use, regardless of the capabilities of the host. Since VT is required for 64-bit guests on EM64T hosts, these guests should report 'LIMIT=FFFFFFFF', indicating that VT is in use. Of course, 64-bit guests cannot be executed on existing Core Duo processors, since they do not support EM64T.

You can use Petr's program to check for VT usage, or you can check the vmware log file. In the log file for a 64-bit guest on a VT-capable EM64T system, you should be able to find the message, "vmm64: VT enabled".

I can confirm your findings with VMWare Workstation.

32-bit VT works, is not tuned, and won't be officially supported unless it can offer the same performance that users of 32-bit VMs expect. Which probably won't be for another generation or two of VT-like instructions.

At this point, 32-bit VT is about as useful as support for a 387 math coprocessor on a Pentium - in both cases, the overhead of the support wipes out the gains. 64-bit VT is necessary because Intel CPUs need that to run 64-bit guests (and it is tuned such that performance is similar to 64-bit non-VT); 32-bit VT just isn't necessary, unless you have a reason why it should be?

Why do you want 32-bit VT support? In what case is 32-bit VT desirable?

Benchmarks on one of the competing programs I mentioned shows that there's a significant speedup when VT is enabled.

In fact, enabling VT makes the competing product faster than VMware in *every single bench* and the response times in that virtualization product when VT is enabled feels like native performance (or very close to it).

If you've read any blogs and some review sites lately, you'll notice people have been raving about the Mac version of that competing companies product that is impressing people due to it's speed on the (VT-enabled) Intel macs.

I would urge the VMware devs to get a copy of the VT enabled product I'm referring to since the company has a trial available and test with VT on and off and also compare it to VMware with VT on. I'm not making up numbers.

I don't feel like repeatedly arguing this point of "VT support makes it faster" when the Vmware devs could just as easily bench the two products and see how VT is indeed faster. Just because your 32bit VT support is broken doesn't mean VT is going to make things slower.


I wouldn't call running 3 times slower than with VT disabled (as Sander_S above confirms) "works". It's broken plain and simple.

To the person who requested the name of the products I'm referring to--I don't think I could say it publically on this board and I don't want to be accused of shilling for the competing companies. If you want to know, send me a private message and I'll reply there.