Vista - Ed Bott Blog on Windows XP Mode

Maybe it was technically necessary in order to move forward, but this need for
hardware virtualization for Windows XP mode (and VPC 7) is going to cause a lot
of trouble.

See Ed Bott's blog at

<http://blogs.zdnet.com/Bott/?p=946>

We are not just talking about old hardware here.

Part of the problem is that hardware virtualization is not a well-known feature.
Indeed I bought a 64-bit dual core machine 3 years ago (before I got into
virtualization) and I never gave it any thought. So of course my machine does
not have it (Athlon64 X2 on socket 939), though I know there were chipsets
available at that time that did support it.

--
David Wilkinson
Visual C++ MVP

David Wilkinson wrote:And back then, the pre-announcements of hardware virtualization made by
AMD and Intel looked like the feature would become available as a
CPU stepping or microcode update for existing CPU designs, ready to
be plugged into existing socket 939 and Xeon based computers.

But both AMD and Intel have decided to break backward compatibility with
motherboards/computers whenever they make even the smallest CPU
improvements. So users cannot simply buy a processor upgrade and pop it
into their existing computer, they have to replace a lot of non-outdated
hardware (sometimes the whole machine), just to get about 5 extra CPU
instructions and a few extra control registers.

It was much nicer during the transitions from 286 to 386, from 386 to
486, and from 486 to Pentium. In each case the chipmakers offered
"upgrade" or "overdrive" processors which had the latest hardware inside
but would fit in the old sockets on the old motherboards. (386SX was a
386 that would fit in a 286 motherboard design, there were
daughterboards for putting a 486DX in a 386 motherboard,
486DX2-OverDrive and 486DX4-OverDrive would fit in most 486DX boards and
there were a few Pentium 1 generation chips that would fit in a 486
motherboard).


--
Jakob Bøhm, M.Sc.Eng. * jb@xxxxxx * direct tel:+45-45-90-25-33
Netop Solutions A/S * Bregnerodvej 127 * DK-3460 Birkerod * DENMARK
http://www.netop.com * tel:+45-45-90-25-25 * fax:+45-45-90-25-26
Information in this mail is hasty, not binding and may not be right.
Information in this posting may not be the official position of Netop
Solutions A/S, only the personal opinions of the author.

Jakob Bohm wrote:Jakob:

Oh yes. If I could replace just the CPU to get hardware virtualization, I would
do it in a heartbeat.

--
David Wilkinson
Visual C++ MVP

> So of course my machine does not have it (Athlon64 X2 on socket 939),As I understand, it's CPU and BIOS related, and not a matter of chipset.

Have read that even if a CPU have Intel Virtualization Technology or AMD-V
on the paper, it might have been disabled by various reasons.

Believe Microsoft will run into some (Vista like) problems with this
solution.

jorgensen wrote:I don't think so. Athlon64 X2 for socket 939 does not support virtualization,
but for socket AM2 it does.

--
David Wilkinson
Visual C++ MVP

David Wilkinson wrote:Yes, and the only reason for that is that AMD Corp decided not to
include the feature on the chip they sell in a package with 939
pins, but do include it in a similar chip they sell in a package
with balls on it and slightly different electrical specifications.

The BIOS issue is just a matter of how it initializes the CPU etc.
and could be fixed with a routine BIOS update if a socket 939 CPU
with virtualization was on the market.

CPU-assisted virtualization does not require any special chipset
features, although it may happen to be in the same CPU versions
that support use with a TPM chip (that feature currently DOES
require chipset support, if used).

--
Jakob Bøhm, M.Sc.Eng. * jb@xxxxxx * direct tel:+45-45-90-25-33
Netop Solutions A/S * Bregnerodvej 127 * DK-3460 Birkerod * DENMARK
http://www.netop.com * tel:+45-45-90-25-25 * fax:+45-45-90-25-26
Information in this mail is hasty, not binding and may not be right.
Information in this posting may not be the official position of Netop
Solutions A/S, only the personal opinions of the author.

Jakob Bohm wrote:So technically, hardware virtualization could work on socket 939? It's just that
AMD have chosen not to include it on Athlon64 X2 CPU's for socket 939?

Very annoying. Any chance they could reverse this decision? They would get a CPU
sale out of me if they did.

--
David Wilkinson
Visual C++ MVP

> So technically, hardware virtualization could work on socket 939? It'sI have an Dual Opteron 939 CPU and it's not supported either, so no change
here.

Another problem with my somewhat outdated (however, fully optimized and
overclocked) system is the AGP 6800 graphic card. When running the Win 7
performance test, I get a pretty high score around 5 - except the graphic
system which comes out with a score of 1 and telling it's unacceptable.