How fast is Parallels VM?

[dhjdhj]

I'm just curious about this - I'm assuming that someone at Parallels has done some measurements to determine exactly how fast a VM session runs as compared to a native XP install (bootcamp or just plain PCs). Clearly for most (possibly all?) usual applications, the speed differential is not important but I'm intellectually curious.

D

 

[Parallels User]

I'm just curious about this - I'm assuming that someone at Parallels has done some measurements to determine exactly how fast a VM session runs as compared to a native XP install (bootcamp or just plain PCs). Clearly for most (possibly all?) usual applications, the speed differential is not important but I'm intellectually curious.

D

My benchmarks showed a 3-5 times increase in speed over VPC 7 running on my 1.33 Ghz Quicksilver.

CineBench Results Parallels Boot Camp
Rendering (Single CPU) ]993 294
Rendering (Multiple CPU) --- 535
Multi-Processor Speedup 1.82
C4D Shading 610 351
Open GL SW-L 484 1422
Open GL HW-L 356 2949
Open GL Speedup 0.79 8.41

(Sorry, I don't know how to do tables and the editor omits extra spaces.)

 

[mattham]

Well - I shall try to give yo a bit of a guide (if I can)

I use eon softwares Vue5 Esprit 3d rendering software. There is a default benchmark to run & it gives you a total time to render scene result at the end. So obviously lower times are better.

Vue is very processor intensive & uses a lot of memory

Computer - default mac mini core duo 1.66 512mb ram.

Software :
Parallels beta 2 (with virtualisation working) (ive only got 512mb ram so the virtual pc is only running on 200mb ram)
Render time 17:03
(note: running vue on win xp with 200mb ram is insane - i'm surprised it even worked.)

BootCamp - windows XP - same macine same software
Render time 8:41

osx 10.4.6 same machine - osx PPC version of vue (same release as windows version)
render time 28:43 (rosetta!!)

As a reference
p4 3ghz
2gb ram
windows xp
render time 15:09

Review: I'm very very impressed with the Native XP (boot camp) performance. and amazed that it even run in a VM with 200mb ram let alone return a half desent render time. PPC version running under rosetta - well that's to be expected.

Hope this helps

Matt :-D

 

[dhjdhj]

Hmm, that doesn't really tell me anything as I don't know how much slower VPC is than a native machine. Much more relevent is the speed of Parallels VM compared to native Windows.
D

My benchmarks showed a 3-5 times increase in speed over VPC 7 running on my 1.33 Ghz Quicksilver.

 

[dhjdhj]

Yes, this is very interesting.
D

Well - I shall try to give yo a bit of a guide (if I can)

I use eon softwares Vue5 Esprit 3d rendering software. There is a default benchmark to run & it gives you a total time to render scene result at the end. So obviously lower times are better.

Vue is very processor intensive & uses a lot of memory

Computer - default mac mini core duo 1.66 512mb ram.

Software :
Parallels beta 2 (with virtualisation working) (ive only got 512mb ram so the virtual pc is only running on 200mb ram)
Render time 17:03
(note: running vue on win xp with 200mb ram is insane - i'm surprised it even worked.)

BootCamp - windows XP - same macine same software
Render time 8:41

osx 10.4.6 same machine - osx PPC version of vue (same release as windows version)
render time 28:43 (rosetta!!)

As a reference
p4 3ghz
2gb ram
windows xp
render time 15:09

Review: I'm very very impressed with the Native XP (boot camp) performance. and amazed that it even run in a VM with 200mb ram let alone return a half desent render time. PPC version running under rosetta - well that's to be expected.

Hope this helps

Matt :-D

 

[aileron51]

Some Java benchmarking

I have done some testing with a very processor-intensive java process on both booting my MacBook Pro into Windows and using a Parallels VM. This process reads 20,000 records from a MySql database and then does some intensive calculations on the data. The process took almost exactly the same amount of time (19 minutes) in both environments. This process does not use graphics, so it seems to show native speed for parallels in processor-intensive activities.