Speed/performance comparison-Fusion vs Parallels-how does yours compare?

[wheat]

I have two Macs running Parallels and Fusion. Although I prefer the stability of Fusion I am taking a greater performance hit on Fusion than Parallels on almost all aspects such as boot up, same programs running compared, etc.

Am I an exception or is this common for others?

[admin]

Can you elaborate on the benefits of running a 64-bit OS for performance?

The 64-bit cpu architecture is cleaner and easier to virtualize (more registers, more memory, we're less constrained).

Are typical desktop tasks actually more efficient under a 64-bit OS for Windows?

Maybe. Realize that driver support (and in some cases even application support) might not be quite up to your expectations yet though. It's unfortunately not a no-brainer to make the switch.

[admin]

Oh - one performance tip I've found...

Running Skype on the guest OS causes much higher CPU

usage even when idle.

Interesting tip. Can't you just run Skype on the host? Or do you depend on features that they have for Windows and not for Mac yet?

[limebreeze]

Yes - this is what I've learned to do. Exit Skype on the guest and run it under the host. Works great...

I find it interesting that both Fusion and Parallels have similar CPU increases when running Skype.

j.

[AbortRetryFail]

Hi, Bgertzfield,

I successfully ran VM Converter 3.0.1 to convert a Parallels Build 3122 VM running a 3.1 GB NTFS + WinXP Home Edition with SP2 that had been compressed several times with Parallels Compressor. This is running on a 2.33 GHz MBP C2D 15.4" with 2 GB of RAM and Tiger 10.4.9, fully patched.

Fusion B4 is configured without the debugger, has 1024 MB of RAM allocated, with the soundcard disabled. I run it in windowed mode, not in Unity.

Fusion seems noticeably slower. Even pressing Windows Start (both on-screen with the mouse as well as pressing the Command key) can take up to a second to pop the Start menu. I can visibly see the Start menu painting in Fusion. Everything seems more sluggish in Fusion B4 than in Parallels Build 3122, including keyboard response, application startup, and processing (e.g. big indexing runs of database applications). Accessing files on the Mac filesystem via the VMWare network shares also seems to take longer, too.

I've attached some screenshots comparing the two VMs (running at different times but with just Outlook 2003 active). Parallels consumes more virtual memory and CPU than Fusion, but less wired memory. Being lazy, I didn't do full Mac OS X reboots between these tests, so there may be some caching effects implicated somewhere. I switched back and forth between Parallels and Fusion several times, so they should have been equally penalized if that's involved somehow.

If there's a configuration tweak somewhere that I've missed, I'd be delighted to hear of it.

Regards,

StevenC

[bgertzfield]

Hi AbortRetryFail,

I would recommend making a fresh virtual machine with Windows Easy Install to directly compare the speed. There may be several settings that were not set in the conversion process, and I'm not sure if you ended up with a virtual IDE or a virtual SCSI disk in your converted machine.

Did you install VMware Tools in the virtual machine?

[AbortRetryFail]

Hi, BGertzfield,

Thanks for the suggestion. I'd thought that something along those lines might be implicated.

I've installed VMware Tools in the converted VM. Going into Virtual Machines > Removable Devices > Hard Disk shows that the Bus Type = SCSI. I have only one CPU allocated to VMware.

I have over two dozen applications installed in the original Parallels VM source instance, many of them requiring third-party vendor activation, including WinXP and MS Office, which now require a call to MSFT every time I do this. Reinstalling them all and reconfiguring them in VMware would take days.

Is there a way of running a verification script to compare expected-vs-actual settings? I'd hate to do this reinstall, only to find that performance is still so poor.

Regards,

StevenC

[bgertzfield]

If that's the case, it sounds like you're running low on memory in the host, and things are starting to swap.

Try editing the file:

~/Library/Preferences/VMware Fusion/config

and adding the line:

hard-disk.useUnbuffered = "TRUE"

This will disable the host disk cache we enabled by default in Beta 4.

[AbortRetryFail]

Thanks for the suggestion. I couldn't find a "config" file in that location, but I did find a "preferences" file. I added the line to that one.

If you think the slow performance is due to page swapping, I'll try a few clean benchmark comparisons of the two with full Mac OS X reboots between them, and full reboots of the respective VMs, too. That'll ensure a fairer comparison, I hope.

Out of curiosity, does this mean that Beta 4 uses lazy writes to files on the virtual filesystem as well as across the network shares to the Mac OS X native filesystem? If a VM running on Beta 4 crashes unexpectedly, does this mean that any buffered (pending) writes are lost, even if the applications (e.g. Excel) think that they've been committed?

Regards,

StevenC

[martindolphin]

I seem to get broadly similar performance from the two products once they are loaded and running, but find the biggest difference is that Parallels resumes a suspended VM a lot faster than Fusion. Fusion is good though - I like the way it draws windows better and Unity looks better than Coherence. Also find that Parallels attitude to support sucks!

[rhind]

I've been using parallels for a few months and have finally started trying VMware with beta 4. I have set up similar VMs in both:

XP SP 2

1GB RAM for VM (3GB total in host imac 2.33 c2d)

Nod32 anti-virus

Visual Studio 2005

I use coherence/unity. To test performance for me, I simply build my VS 2005 projects in both products (not running at the same time) multiple times in a row (in case caching etc has an effect).

Parallels consistently takes 20-30% longer to compile than Fusion. I have 2 solutions, one which is C C# project that takes about 1min 20 to build in Fusion and another is a mixture of C#, C+/CLI and C+ projects which takes roughly 7min 20 in Fusion.

Fusion also uses less idle CPU (5-6% as opposed to Parallels 3's 10-12%). Parallels 2.5 only used 6-7% with sound disabled, so not sure what has increased this in V3, but its a bit of a disappointment.

The speed increase in fusion (if it stays the same in the final release) will be enough to switch me to fusion full time.

Current impressions: Fusion faster, Parallels much better host OS integration and coherence usability (easy access to start menu etc, remembers UI mode).

Currently Fusion is very annoying in unity in that every time I close my last XP window, it pops up the launcher window. I like to keep an uncluttered desktop so this is annoying. More annoying is the fact that I can't get to the proper start menu easily and also that applications menu is disabled when all windows are closed (which is my typical case when I come back to my Mac as I leave the VM open).

But I realise that VM is still a beta and this is the first beta with unity so I have high hopes that the usability will increase up to the actual release.

I would really like to see OpenGL acceleration as some of the apps I develop are scientific imaging apps and use OpenGL (we don't push it like games, and the software rendering can cope, but Parallels OpenGL acceleration does a very good job for us, although 3.0 is garbling some rendered text that 2.5 didn't have a problem with).

Cheers

Russell

[bgertzfield]

Thanks for the suggestion. I couldn't find a

"config" file in that location, but I did find a

"preferences" file. I added the line to that one.

You'll need to create the file "config" if it doesn't already exist.

The preferences file is only for UI-related preferences.

[Andreas_Masur]

Thanks for the suggestion. I couldn't find a

"config" file in that location, but I did find a

"preferences" file. I added the line to that one.

Ben,

Quick question...didn't this parameter used to be in the 'vmx' file? Thanks.

Ciao, Andreas

[msvista]

Fusion Beta 4 performance is much much better than Parallels...and Parallels I keep getting blue screen even and then. Atleast Fusion seemed to be working fine and performance is good. May be we should see the unity support for vista in coming months

http://venublog.com/

[manfredell]

Since I converted my VM to scsi disk, Fusion is running acceptably.

But I certainly cannot subscribe that it runs faster than Parallels.

But which really is slow is the time it takes to resume from a suspended state. It takes minutes when P takes seconds....

[hilmartor]

Is it possible to swap ctrl/cmd key on the keyboard? It is better for copy/paste.