Ken32701
Contributor
Contributor

Which is faster?

Jump to solution

Hello...

I used VMWare Converter to convert my Windows XP Pro machine into a virtual machine. I notice it a bit sluggish at times.

I know that Fusion can also run Boot Camp-Windows as virtual machines at run time.

Does anyone know if it would be faster to run my Windows Environment from Boot Camp via Fusion or continue running VMWare Fusion as I have it now on my Mac?

I'm trying to maximize my performance. I'm running VMWare Fusion on my MacBook Pro that has 4GB of Ram.

Any input would be appreciated!

Thx!

Ken

0 Kudos
1 Solution

Accepted Solutions
rcardona2k
Immortal
Immortal

Well I ran some tests tonight. I will say from the start it's unfair to benchmark a beta which is what 1.1b1 is. I can't publish hard numbers per the VMware EULA and because this is beta.

What I compared is the VMware Boot Camp Partition with Windows XP SP2 configured as 256 MB, no connected peripherals (CD/DVD disconnected), 1 network adapter in NAT mdoe, no 3D, vs.

a newly built WinXP SP2 Fusion VM, 256MB, no connected peripherals (CD/DVD disconnected), 1 NIC in NAT mode, no 3D.

The virtual disk for the Fusion VM is a pre-allocated IDE disk, defragged on the host and defragged internally.

What PassMark Performance Test v6.1 showed is the Fusion VM scored much higher in CPU and Disk. The CPU mark surprised me. I built my Fusion VM with 2x processors and I looked at the BC config and it was set to 1 CPU. I figured that was it, so I reconfigured the Boot Camp VM for 2 processors. I ran the CPU tests again and there was still a 2x difference. I deleted the enter Boot Camp VM bundle and let Fusion re-build it and it defaults to 1 CPU, so I changed it back to 2 CPUs. The VM came up and wanted to install the Tools so I let it repair the tools, reboot again. The PassMark CPU mark rating for the Boot Camp VM did not improve. I looked in device manager and two processors are in play under an ACPI Multiprocessor HAL, so I can't explain this. I verified through Activity Monitor that the Boot Camp VM is not using both cores during CPU test as it should be. If someone can confirm this or explain, I'm all ears.

On disk, this should have been a straight Boot Camp partition, which by definition is a flat host partition vs. a pre-allocated host file. Again the Fusion VM substantially outperformed the Boot Camp partition VM. I attribute this to the driver VMware uses to access the host partition. Maybe it doesn't cache as well as OS X does, I'm not sure. i assume if IDE does this well SCSI would only get better or at least remain the same.

If you are willing to leave the Boot Camp partition behind, based on these results (in a beta*) a clean new Fusion VM is the better option.

View solution in original post

0 Kudos
10 Replies
rcardona2k
Immortal
Immortal

IMO, it depends if "maximum performance" for you means actually rebooting into the Boot Camp partition. If you're willing to leave OS X, use a Boot Camp partition. If you're NOT willing to leave OS X, build and optimize a Fusion VM on your OS X filesystem. You can experiment with options like scsi drivers and pre-allocated storage and defragmentation to improve IO performance. For CPU performance you should run XP with 2-processors.

OTOH, nothing will beat Windows running natively directly on the host. If your VM is not in a Boot Camp partition you lose the option to reboot into Boot Camp proper.

So...again it depends on what "maximum performance" means for you.

Ken32701
Contributor
Contributor

Hi...thx for responding. This helps...

To clear things up...I do not want to leave OS X. I'm trying to determine if my VM will run faster on a Boot Camp partition thru Fusion vs just running my VM in Fusion on my OS X partition. Does it make any difference in performance if Fusion runs a VM from the Boot Camp partition vs off the OS X partition?

Thx

Ken

0 Kudos
rcardona2k
Immortal
Immortal

It's been a while since I benchmarked XP in a Fusion VM on OS X v. XP in the Boot Camp partition. Last time I did it, IO on my pre-allocated, defragged scsi Fusion VM was faster. CPU is the same. Someone with more recent data can chime in. I can also re-install Passmark and re-run the tests tonight and let you know.

0 Kudos
bflad
Enthusiast
Enthusiast

From my personal experience, there doesn't seem to be too much difference between the two. Even with that said, your milage may vary.

Hope this helps,

Brian

0 Kudos
Ken32701
Contributor
Contributor

Hi rcardona2k,

I take it that your pre-allocated, defraggred scsi fusion vm was a separate partition. If you don't mind running Passmark to test performance between running a VM on an OS X partition vs a Bootcamp partition that would be great. It would be nice to know the results.

Thx!

Ken

0 Kudos
BP9906
Expert
Expert

I wouldnt mind Richard running those tests again too Smiley Happy

I'd be curious to see how the following compares:

1.) Defrag'd Bootcamp partition

2.) Pre-allocated SCSI defrag'd

3.) Dynamic SCSI defrag'd

I usually lean on the Dynamic SCSI just cause I dont like wasted space. Smiley Happy

0 Kudos
rcardona2k
Immortal
Immortal

Well I ran some tests tonight. I will say from the start it's unfair to benchmark a beta which is what 1.1b1 is. I can't publish hard numbers per the VMware EULA and because this is beta.

What I compared is the VMware Boot Camp Partition with Windows XP SP2 configured as 256 MB, no connected peripherals (CD/DVD disconnected), 1 network adapter in NAT mdoe, no 3D, vs.

a newly built WinXP SP2 Fusion VM, 256MB, no connected peripherals (CD/DVD disconnected), 1 NIC in NAT mode, no 3D.

The virtual disk for the Fusion VM is a pre-allocated IDE disk, defragged on the host and defragged internally.

What PassMark Performance Test v6.1 showed is the Fusion VM scored much higher in CPU and Disk. The CPU mark surprised me. I built my Fusion VM with 2x processors and I looked at the BC config and it was set to 1 CPU. I figured that was it, so I reconfigured the Boot Camp VM for 2 processors. I ran the CPU tests again and there was still a 2x difference. I deleted the enter Boot Camp VM bundle and let Fusion re-build it and it defaults to 1 CPU, so I changed it back to 2 CPUs. The VM came up and wanted to install the Tools so I let it repair the tools, reboot again. The PassMark CPU mark rating for the Boot Camp VM did not improve. I looked in device manager and two processors are in play under an ACPI Multiprocessor HAL, so I can't explain this. I verified through Activity Monitor that the Boot Camp VM is not using both cores during CPU test as it should be. If someone can confirm this or explain, I'm all ears.

On disk, this should have been a straight Boot Camp partition, which by definition is a flat host partition vs. a pre-allocated host file. Again the Fusion VM substantially outperformed the Boot Camp partition VM. I attribute this to the driver VMware uses to access the host partition. Maybe it doesn't cache as well as OS X does, I'm not sure. i assume if IDE does this well SCSI would only get better or at least remain the same.

If you are willing to leave the Boot Camp partition behind, based on these results (in a beta*) a clean new Fusion VM is the better option.

View solution in original post

0 Kudos
BP9906
Expert
Expert

Agreed that it must be their driver to the host partition and caching. I have noticed that using a host file (VMDK) is faster so I've stuck w/ that, then created a VMDK that maps to a host partition that's Fat32 (my data drive). Having it this way still seems to slow down my Vista VM that uses a VMDK, so I have to concur with you're suggestion. Of course I'm basing my experience on Fusion 1.0 since I havent loaded up the Vista VM since 1.1, but according to release notes that must've not changed.

0 Kudos
Ken32701
Contributor
Contributor

Rcardona2k,

Thx for the terrific feedback! This helped me tremendously. I'm going to stay clear of Boot Camp and work just with Fusion VM instances as you recommended. You saved me a lot of time. Thx for doing these tests!

Kind Regards,

Ken

0 Kudos
drwff
Contributor
Contributor

Does this mean you would recommend getting rid of an existing Bootcamp partition if I want to just run Windows through VMFusion and don't need to reboot Windows natively ever?

0 Kudos