Parallels, VMware trade shots in virtualization shootout
A new study from MacTech reveals that both Parallels Desktop and VMware Fusion have their strong points, but that software Mac users should choose depends heavily on Windows and the user's goals -- and that a Mac may outpace a separate PC in a native environment.
The technology magazine notes in its benchmark tests that the older Windows XP operating system runs considerably faster than Vista regardless of the environment, but also that the relative speed of each OS inside a virtual machine can vary widely.
Using a MacBook, MacBook Pro, and a Mac Pro tower as its comparison systems, MacTech finds that Parallels Desktop is generally faster than VMware Fusion in common networking and office tasks when using Windows XP. When switching to Vista, however, Fusion handles the performance hit more gracefully and drops by an average of 32 percent across the three systems versus a steeper 85 percent for Parallels.
Surprisingly, either can be slightly faster than running Windows through the native Boot Camp mode for some of these particular tasks, the publication says. Parallels earns additional recommendations for those who depend on a tight link between Mac OS X and Windows, though VMware's solution may be better for Windows software that depends on multiple cores.
Both have a good selection of virtual appliances and are easy enough to use that selecting a solution can be just a matter of opinion, according to the comparison.
Nonetheless, users who don't need the tie-in between the two operating systems are still better-served by running Windows in Boot Camp, MacTech tells readers. And when compared to a reference Fujitsu notebook running a 1.86GHz Core Duo, even the base MacBook was typically faster, making it more feasible to run Windows directly from the Mac itself.
"It's faster than a PC anyway," MacTech says.
Complete test results, comparison graphs, and further analysis are available at the MacTech site.
The technology magazine notes in its benchmark tests that the older Windows XP operating system runs considerably faster than Vista regardless of the environment, but also that the relative speed of each OS inside a virtual machine can vary widely.
Using a MacBook, MacBook Pro, and a Mac Pro tower as its comparison systems, MacTech finds that Parallels Desktop is generally faster than VMware Fusion in common networking and office tasks when using Windows XP. When switching to Vista, however, Fusion handles the performance hit more gracefully and drops by an average of 32 percent across the three systems versus a steeper 85 percent for Parallels.
Surprisingly, either can be slightly faster than running Windows through the native Boot Camp mode for some of these particular tasks, the publication says. Parallels earns additional recommendations for those who depend on a tight link between Mac OS X and Windows, though VMware's solution may be better for Windows software that depends on multiple cores.
Both have a good selection of virtual appliances and are easy enough to use that selecting a solution can be just a matter of opinion, according to the comparison.
Nonetheless, users who don't need the tie-in between the two operating systems are still better-served by running Windows in Boot Camp, MacTech tells readers. And when compared to a reference Fujitsu notebook running a 1.86GHz Core Duo, even the base MacBook was typically faster, making it more feasible to run Windows directly from the Mac itself.
"It's faster than a PC anyway," MacTech says.
Complete test results, comparison graphs, and further analysis are available at the MacTech site.
Comments
I'm using Parallels on the Mac on and off, at the moment I'm beta testing Windows software (highly specialized environmental monitoring database) and actually I'm glad I don't have to test it on my work XP boxen. At work I'm running virtual Linuxes under VMWare which is also a rather painless experience. Can't really comment on VMWare under Mac OS.
All that I still would really like from either of these products is Mac OS virtualization. I don't say I can't live without that but it would be a big nice-to-have!
- Vista under Boot Camp averaged 24% slower than XP
- Vista under VMware Fusion averaged 32% slower than XP
- Vista under Parallels averaged 85% slower than XP
Anyways, I'm not touching that resource hog of an operating system with a six foot pole. One more reason to phaze out Windows from any working environment.
check em out and let me know what you think
http://www.youtube.com/watch?v=2EBlYb-PY_E
http://www.youtube.com/watch?v=s7qxs-YHaGc
The only reason for me to use a Windows instance is because some sadly designed website requires IE.
Seriously, this is one of the few things that really annoy me. It *is* a PC you idiot. of *course* a 2 or 2.2 ghz core 2 is faster than a 1.83ghz core 2 running the same os, same ram, and probably even similar chipsets. That's a no-brainer. It's not 'cause it's a mac, it's 'cause it's the *same hardware*, but faster!
Seriously, the mac vs. pc thing has got to go. Windows vs. OSX? Sure. But hardware distinction, tenuous to begin with, went out the window completely when Apple shifted to intel chips!
First, I'm running Vista Ultimate under Boot Camp and it runs fast. Secondly, Windows isn't going to be phased out anytime in the forseeable future. Companies simply have to much invested in it from developed applications (Visual Studio), Win servers, Win Clients and Thin Clients. I like both Vista and Mac OS X. They both have their niche.
Vista has a future. But it will involve "XP-ing" it to some degree in the next few years, and to work out major annoyances. Microsoft now knows what works (XP!) and what mostly doesn't (Vista's "security"!). At least going into the end of this decade we won't have to wait and see what the "next big thing" out of Redmond is.
It's pretty much for most people, going to be some sort of XP4/ VistaSP3 hybrid-thingy. On one side of the coin people getting the last breath out of XP and on the other side Vista improving to be a reasonable option.
At this stage most people are better off with XP2Pro IMO.
I am waiting to upgrade (back) to Vista for DirectX10 on my PC.
When it stops dragging the whole gaming speed down.
I may get a PS3 because I am sick of all this hardware upgrading
stuff. But overclocking is teh fun and I can tinker with stuff
when I like, rather than a "black-box" console.
In the meantime, I look forward to some good
Unreal Tournament 3 tonight on my WinXP2Pro, DirectX9.0c
overclocked PC.
"It's faster than a PC anyway," MacTech says.
Seriously, the mac vs. pc thing has got to go. Windows vs. OSX? Sure. But hardware distinction, tenuous to begin with, went out the window completely when Apple shifted to intel chips!
Agreed - the architecture is so similar now that saying a Mac is faster than a PC is a contradiction, since a Mac now IS a PC (but with OSX). Still, it's good to have something "official" now and then showing that Macs perform very well as PC's (like we didn't know it already).
I made a couple of videos for one of parallels promotional contests on the topic of "why choose?"
check em out and let me know what you think
http://www.youtube.com/watch?v=2EBlYb-PY_E
http://www.youtube.com/watch?v=s7qxs-YHaGc
Love the stop motion on the second video :-)
VMware's solution may be better for Windows software that depends on multiple cores.
VMware absolutely is better for software that depends on multiple cores, because only VMware supports multiple cores in its virtual machines.
VMware also runs X64 software infinitely faster than Parallels, because Parallels only supports the old 32-bit X86 instruction set.
Use of the X64 instruction set often yields better performance (up to 25% or more). I would therefore have liked to see benchmarks of 64-bit Vista and 64-bit XP, along with dual core benchmarks. This would give us an idea of how much "horsepower" one can utilize in a virtual machine, as well as the incremental efficiency of adding a second core.
Lastly, why wasn't Boot Camp benchmarked using multiple cores?
VMware absolutely is better for software that depends on multiple cores, because only VMware supports multiple cores in its virtual machines.
VMware also runs X64 software infinitely faster than Parallels, because Parallels only supports the old 32-bit X86 instruction set.
Use of the X64 instruction set often yields better performance (up to 25% or more). I would therefore have liked to see benchmarks of 64-bit Vista and 64-bit XP, along with dual core benchmarks. This would give us an idea of how much "horsepower" one can utilize in a virtual machine, as well as the incremental efficiency of adding a second core.
Lastly, why wasn't Boot Camp benchmarked using multiple cores?
I think the article may not have been targeted as much to "pro" users? But you are right, 64bit multicore can be an important advantage when choosing VMWare.
I am trying now to install Unreal Tournament 3 Server in Parallels on Mac Pro 2.66ghz 4-core 1gb RAM (I know, it is demo Mac Pro that is why so little RAM). In this case I don't mind Parallels just hogging 1 core and 512mb allocated RAM. In this case I like the fact that it only uses 1 core and leaves the other 3 freeeeee for other goodies like OS X Leopard Server NetBoot and NetInstall. Our Mac Pro RAM upgrades have been approved, coming soon hopefully...
Agreed - the architecture is so similar now that saying a Mac is faster than a PC is a contradiction, since a Mac now IS a PC (but with OSX). Still, it's good to have something "official" now and then showing that Macs perform very well as PC's (like we didn't know it already).
I believe there are many reports circulating around now that spec-for-spec comparisons with MacBookPro show that they outperform PCs for Vista, maybe XP too. I think the article was trying to say that for XP virtualized in Mac the performance is very close to Bootcamp or native PC install speeds.
The latest version dramatically improves performance over 1.0
I recently downloaded both demos and found VMware faster than Parallels. . I also found Parallels support to be lacking. I eventually bought VMware over Parallels
The summaries for this test seems so confusing. It's weird how they worded most of the test.
The best solution is to download both demos and find out which one works best for you. Each has their strengths and weaknesses and one will probably run some obscure program you need better than the other.
But I've had a string of issues with Parallels when doing non standard stuff, like with peripherals and with Linux. So for that stuff I use VMWare.
I'd expect Parallels to capture the home user market and VMWare will capture the professional/enterprise market. Keep in mind that VMWare has been around for ages and is version 1.x only in the Mac world. Still, Parallels was first in the Mac world so getting people to switch is an uphill battle for VMWare.
I agree that the competition between the two will be good for consumers. I hope they keep at it.
I think the article may not have been targeted as much to "pro" users?
True, pro users might be the most dis-served by the review, and non-pro users might be the most misled. For years now, every Apple computer sold contains multiple cores. It looks to me like the reviewer stopped when they got the result they wanted: Parallels (supposedly) faster than VMware Fusion. It either supports their inherent wish to see Parallels look best or it produces a result that will generate more ad revenue.
... I like the fact that [Parallels] only uses 1 core and leaves the other 3 freeeeee for other goodies
VMware defaults to using 1 core and is configurable to using 2. That still leaves 2 (6) cores free in a 4 (8) core Mac Pro, while giving the guest OS more oomph when it's needed.
I've seen other Parallels vs. Fusion reviews that found the load on the host system of an idle guest is quite high for Parallels and almost nil for Fusion.
I've used VMWare on Windows, for running Linux, and that has usually been a good experience, but truthfully, I'd rather dual boot to get the best performance.
As for why people would what to run Vista in a VM (other than testing purposes), I'd stick to a real box or XP.