VPC/WMPlayer comparison?

I'm new to VPC ... downloaded it, installed on XPSP3 (2.66 GHz Core 2
Quad Dell, 2GB RAM), made a Vista Ultimate VM with a 25 GB VHD. It
works fine but it seems to be slow. This test (using JPSofts 4NT
command processor) gives interesting comparisons:

timer & for /L %i in (1,1,10) %comspec /c exit & timer

Run on the host (real) machine: 1.1 sec.
On the virtual Vista in VPC: 5.6 sec.

I also tried VMPlayer. I don't know if it simply loaded, or
converted, the MS VM, but it ran fine and the test above, in Vista in
VMPlayer, gave 1.6 sec.

I have repeated the tests many times with similar results.

Is this to be expexted? Is VPC just that slow? Are there any
suggestions for speeding it up (drastically)?

Thanks.

 

Install the Additions and it will speed up a bit.

 

They are/were installed.

 

Too bad, then you have done all you can. :-(
Except of course if you have not yet enabled hardware virtualization
for the guest, this is said to have profound effects. Can't vouch for
this myself since my PC:s all are not equipped with VT options...

Personally nowadays I most often use VMWare Workstation 6.51, which
works very well indeed....

BTW: When you used Player, how did you create the virtual machine
(since Player does not have facilities for creating a guest)?
Exactly what did you do to start it?

Maybe you could get even better performance by *creating* the guest in
VMWare (Workstation)???

 

I guided VMPlayer to the VPC virtual machine (in MyDocs) ... (had to
fiddle with the drop-down box showing allowable extensions) ... and
simply opened it. The first time VMPlayer took 3-4 minutes
(converting/interpreting?) but thereafter, it was remembered in
VMPlayer's UI and would start immediately.

I gather from your comments (and the lack of others) that the slowness
of VPC is well-known and accepted as a fact of life.

My only gripe about VMPlayer is the fact that merely installing it
gives you 7 new processes running all the time.

 

Yep,
VMWare puts a number of services into your system when it installs. It
is a much larger footprint than VPC2007. You could figure this out by
comparing the installer sizes:
VMWare Workstation 6.5.2 = 473 Mb installer package
VPC2007 SP1 = 32 Mb installer
More than an order of magnitude difference!

Right now I have my software development machine running in the
background in VMWare Workstation and TaskManager shows 8 processes
active.
In my Services display I see 5 VMWare services installed.

I guess it does take some software to have this running...

 

VPC VMs in general run just as fast as VMWare ones do for me. I run
VPC almost exclusively for Windows VMs.

Running timers in VMs doesn't really give you an accurate idea on
speed. Depending on how the VM software handles timers, they could be
jumping or skipping to keep up.

some of these you can disable with no ill effects. You don't need the
agent for a lot of things. Also, if you're not going to be using the
built in NAT and DHCP you can turn those off.
When I set up VMWare on a system, I always use the network based DHCP
so I disable the other 2 VMWare adapters also.

 

The timers were just a form of "hard evidence". **Everything** is
**obviously** a lot slower with VPC. It wouldn't surprise me if the
nearly 4:1 speed difference seen with the timer experiment was "across
the board".

You say "not so" for you. Have you any tips?

Perhaps you can tell me if this is possible, with VPC, VMPlayer, or
both. I have two NICs. NIC1 is dedicated to the internet (DHCP,
RoadRunner). NIC2 is dedicated to a home LAN but most of the time
there are no other machines on that LAN. Can I "export" the raw NIC2
to a virtual machine, say Vista, and configure Vista to (also) do DHCP
to RoadRunner (I'm allowed 2 RoadRunner IPs)? As it stands, I'm just
doing the NAT thing and have basic internet connectivity but I can't
get in from the outside.

Thanks!

 

I'm just saying that using timers won't necessarily give you an
accurate respresentation. Generally VMs will skip cycles to keep up.
Also, without syncing with the host, their clocks will fall behind.

No secrets. I give them plenty of RAM and run them on a disk that
separate from the host OS. Most all of my systems were purchased with
VMs in mind, so they have hardware VT also.

You can simply remove the TCP/IP components from NIC2 on the host
side, then the VM would get your 2nd IP from RoadRunner. I've never
tried it on VMWare Player, but I'd assume you could do something
similar by assigned the NIC to the VM directly.

 

After quite a search, I found out that my Core 2 Quad Q6700 supports
VT too. VPC doesn't see it. Must it be enabled in the BIOS set-up?
If so, is there any down-side to enabling VT ... any additional
requirements for using it?

 

Yes, you must enable it in the BIOS also.
Theoretically, it could be used for a malware attack, the "blue pill".
To use VT you need a BIOS and CPU that support it, that's really it.

 

I read somewhere that there was some connection between VT and DEP.

 

[/QUOTE]

Turning VT on made a significant difference. The results of my timer
tests were cut in half, and twice the speed seems a good subjective
estimate of the improvement. Thanks!

 

Is there any PC speed test software around that can be used to assess
performance of a physical and a virtual system in a "fair" way?
It would be interesting to benchmark the two against each other.

 

Benchmarks are always tricky with virtualization. It's hard to get
accurate results. I think that some VM benchmarks are in the works
though.

 

Here are my two recommended test methods:

1. Use a speed test where the important number is how long the test
takes to run (like the for command earlier in this thread or running
some real world task that takes time), and where that time is in the
order of 10 to 400 seconds. Time the test with a real non-computer hand
held stopwatch.

2. Use a test where the test start and test end can be observed over the
network and time the test from a computer that is not sharing any
hardware with the test machine. Because this does not involve human
patience and reaction time, test times can be measured with reasonable
accuracy in the range 2 to 100000 seconds.


--
Jakob Bøhm, M.Sc.Eng. * * 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.

 

The problem is designing such a test program...
For example if one just uses some simple loops then chances are high
that the optimizing compiler discovers this and just skips the loops.
There must be some math involved tha takes a longish time also on a
good PC.
Finding big prime numbers maybe? But I don't know of any such programs
(or better yet sourcecode so I can compile myself).

In principle I am loking for some test algorithms to include in a test
program or for a ready-compiled test program....

 

Loop while randomly adding +/-1 to a counter, then print the counter.
I don't think the compiler can fake that. If you use the same seed
for rand() every time it'll be utterly reproducible. This produces
better (more satisfying anyway) results in my comparison, giving 18.45
sec on the host and 20.65 sec on the guest. It could easily be made
more elaborate.

#include <windows.h>
#include <stdio.h>

INT wmain ( INT argc, WCHAR **argv )
{
srand(666);
INT x = 0;
for ( INT i=0; i<1000000000; i++ )
{
x += ( rand()%2 ) ? 1 : -1;
}
wprintf(L"%lu\n", x);
return 0;
}

I can think of reasons why my first test (for ... %comspec /c ...)
would give worse results. It ultimately invokes CreateProcess() a
number of times which may be slower on Vista (guest) because of added
security. And it causes some blinking of the console's title bar ...
so there's some graphics going on ... likely to be much slower because
of the emulated card and the lack of hardware acceleration.

 

In message <> Bo Berglund

Prime95...

 

In message <> "Steve Jain

Interestingly, it won't help malware much if you use a virtualization
solution yourself since the malware can't hide using VT while another VT
solution uses VT itself.