Windows on Mac Performance, Parallels vs. VMWare vs. Boot Camp
Over the holiday, I played around with three Windows on Mac solutions mostly to find the best environment to run a multi-player game. But it lead me to benchmarking Parallels, VMWare and Apple’s Boot Camp. I decided a Windows on Mac performance showdown was in order. The results were surprising.
Parallels has been running on my Macbook Pro for a while. But I have been a little unhappy with its performance. I have had some frustrating pauses and general sluggishness. I had read that VMWare is a nice solution. Boot Camp always seemed like the best way to get performance and true compatibility for running Windows XP on a Macbook Pro. With Boot Camp, Mac OS X doesn’t even run. You boot directly to Windows.
Method
I found a copy of Boot Camp 1.4b, downloaded the latest VMWare beta v1.1, and tested with Parallels v2 and v3. I installed a fresh copy of Windows XP using each. Next, I stripped out all the junk that gets installed along with Windows, shut down unneeded Windows services, and installed some benchmarking software. I found Performance Test 6.1 by PassMark. It seems pretty decent and has a free trial. I configured each instance of Windows XP as similar as possible; same screen resolution, color depth, and memory. I didn’t have an option to reduce memory for Boot Camp so it had the full 3GB of RAM that is installed on my machine. I ran the benchmark three times on each instance and used the best result from each in the graph below.
Results
As you can see in the graph, Parallels is the winner. This was a big surprise to me. Given my experience with Parallels, I expected it to be the loser. I also expected Boot Camp to win because it doesn’t need to have Mac OS X running underneath it. I can’t explain why Boot Camp wouldn’t win in every category but it didn’t. Does anyone know?
Here are the results … higher numbers are better … click on it for a better view.
As you can see, Parallels v3 is a clear winner in almost every category. I triple checked the disk performance in Parallels 3 and found the result is correct. Parallels seems to do a very good job of caching for this particular benchmark.
Unfortunately, performance of 3D graphics is dismal in both Parallels 3 and Boot Camp. I can tell I won’t have a good experience with Call of Duty 4.
My understanding of how virtualization actually works is kind of shaky, so please feel free to mock and deride what follows…
Anyway, my understanding is that the OS virtualization takes place in the CPU. Thus, MacOS and Parallels+Windows are actually running side by side as equals in the chip. Windows isn’t actually inside OSX, it’s next to it.
If this is true, then Windows probably requires some sort of bridge logic in order to make the Mac hardware look like PC hardware. This hardware emulation or bridging is one of the features provided by Parallels and bootcamp.
If the above is true, then Bootcamp might have an advantage over for resource intensive apps, but if you are doing light stuff then the advantage would go to the system with better emulation software. Parallels might simply have done a better job at that.
Or it might be powere by unicorns. YMMV.
Wow, I would have thought your initial hunch that Bootcamp would perform better than Parallels correct. Thanks for running the test. Good info to know.
The biggest disappointment for me was 3D graphics performance. It means the Macbook Pro won’t run games well in any configuration.
I forgot to mention a key point. When viewed from Mac OS X, VMWare uses 0-15% cpu at idle. Parallels uses 15-30% at idle. That has an impact on responsiveness of your machine if you use it all day. It is another reason why I expected Parallels to lose.
One of the big reasons I moved to VMWare Fusion was memory footprint, I like to keep windows running along side my compiler, or Photoshop, etc., without the disk swapping. Do you have any comparisons on how invasive Parallels is vs. VMWare here? I moved to VMWare with the assumption that it would take less memory and less CPU on idle – so if that still proves true, VMWare is the winner in my book.
Please don’t get me wrong, but I can’t believe the information provided here is correct.
I have Parallels 3 and Bootcamp running on my machine. I can run some 3D-Applications with Parallels, but not all Direct3D-Apps. While I get about 15 up to 30 FPS with Parallels, I get 60 FPS with Bootcamp (Sam & Max – Icestation Santa).
I can even play Bioshock and the Crysis-Demo with Bootcamp. Not with full detail, but medium is good enough for me…
I haven’t done any CPU-Benchmarks yet, but how a Virtual-Machine could be faster than a CPU in native mode is beyond me and I simply can’t believe this…
The only thing I can imagine, is that the configuration got messed up somehow.
My machine: MBP Core2Duo @ 2 x 2.2GHz, 2GB RAM, Geforce 8600M GT 128MB RAM
I had trouble believing it too. So I reinstalled Boot Camp twice including partitioning, formatting and a fresh install of Windows. I don’t think it was a configuration problem. I am relying on Passmark Performance Test 6.1 for the numbers.
[...] http://alankeister.wordpress.com/2007/11/28/windows-on-mac-performance-parallels-vs-vmware-vs-boot-c... [...]
I really appreciate your test. I find that Safari in Parallels works faster than Safari in Mac 10.6.1 or any other Ma OS version for that matter. Except for the fact that the fonts are so ugly in Windows I would use Safari for XP in Parallels in Coherence mode.