Shootout: Parallels outperforms VMware Fusion in many tests
A head-to-head test of Parallels Desktop for Mac and VMware Fusion - the two leading competitors in the Mac Virtualization segment - crowns Parallels as the clear winner, but the testers found both products performed well and VMware can even be a better choice under certain circumstances.
MacTech Magazine ran both virtual machines through thousands of real-world tests, timed with a stopwatch, to find out which performs the fastest. The results are in a five-page story at its website, which AppleInsider highly recommends that prospective buyers of either solution read in full.
Acknowledging Apple's own Boot Camp, which can boot natively into Windows, MacTech decided to look only at virtual machines since they do not require restarting to switch between operating system environments. The publication was also interested in determining which operating system - Windows XP or Vista - offers the best performance. The testers kept an eye on the differences between larger memory configurations, multiple processors, and 64-bit versions of the OS.
To facilitate these tests, they used four different models of Macs to cover the range of machines actual customers would use with more than 2,500 total tests. Again, each was timed with a stopwatch for real-world results to avoid difficulties in running diagnostics on virtual machines that don't always properly keep track of the time it takes for tasks to execute.
Specific tests included: performance (launch and CPU), file and network IO, footprint, application launch and performance, and tests of 3D and HD graphics and games.
Parallels Desktop was the clear winner, running 14 to 20 percent faster than VMware Fusion in the majority of overall averages. There was only one exception when VMware was faster: 32-bit Windows XP on two virtual processors, which ran about 10 percent faster than Parallels.
Overall Test Results
Parallels
14% faster running 32-bit XP and Vista on a single virtual processor20% faster running 32-bit Vista on two virtual processors15% faster running 64-bit Vista on two virtual processors
VMware Fusion
10% fasster running 32-bit XP on two virtual processors
Launch and CPU Tests
In using Windows' compression tools and MP3 to WAV transcoding, Parallels was faster in every version of the test, ranging from 7% faster when launching virtual XP and resuming from suspend after a full restart, to 31% faster after a full Windows boot on Vista. Â*The contest was closest when both Parallels and VMware Fusion were launched without restarting, allowing the benefits of caching. Â*In that test, Parallels was 17 percent faster in Vista and 22 percent faster in XP, but the range was only zero to one second.
Application Launch Test (Word and Outlook)
Here, there was no clear advantage for either virtualization engine. Â*When launching Microsoft Word after a full reboot, Fusion was faster by 8 percent on XP while Parallels trumped with Vista at a 33 percent faster clip. Â*Without a restart, Fusion notched a sweep, while the two again split with launching Outlook (Fusion narrowly won with XP, Parallels with Vista).
Application Performance Tests
MacTech did a global search and replace of about 95,000 items in a very large Word document, and in Excel, they ran a macro to generate a large amount of random numbers, filling cells with them. Â*PowerPoint rendered slides with transitions as quickly as possible.
In the Word test, Parallels was faster under both XP (4 percent) and Vista (9 percent), with similar results on Excel (14 percent faster on XP, 13 percent on Vista). Â*However, Fusion did score a victory with XP on the PowerPoint test, albeit a small one, at two percent. Â*On Vista, it was a tie.
Internet Explorer Speed Test
\t
MacTech created a large, complex page in HTML served from a local LAN server using common HTML elements and no JavaScript.Â* VMware Fusion took significantly longer across the board, so they removed the results from the overall analysis to avoid it skewing the results. In general, Parallels was more than 80% faster in XP and Vista both with and without SSL.
Footprint Tests
Three footprint tests were also run. The first booted Windows, let it finish, and then measured CPU while both Mac OS X and Windows were sitting idle.Â*The second test was similiar but evaluated launches of Word and Outlook, and the third gauged CPU usage while playing a DVD. In all three tests, Parallels and Fusion fared well.
Virtual Machine Memory Usage
As part of its Virtual Machine Memory Usage tests, MacTech noticed that Parallels only took memory from Mac OS X when Windows needed it. The publication ran an "endless loop" macro in Excel that generated random numbers.Â* When battery life approached the end, Parallels paused the virtual machine and the Mac went into hibernate, saving the state of the virtual machine.Â* VMware Fusion suspends the virtual machine before sleeping. The end result was that Parallel's battery usage lasted at least 8 percent longer in both XP and Vista.
File and Network IO Tests
In a highly refined I/O test -- due to problems discovered throughout the testing proccess -- 3.7GB of data (two 1.85GB files) was duplicated and copied to and from a Mac hard drive, LAN Server, and USB Flash drive under both XP and Vista. Parallels beat out Fusion hands down in all but 2 of the 14 tests.
3D and HD Graphics Tests
XP: Smoothly played both 720p and 1080p videos in both environments.
Vista: Couldn't play 720p at all in Parallels. VMware Fusion stuttered on every machine except the Mac Pro. Given 720p results, MacTech didn't bother trying 1080p.
Games
MacTech found that frames per second was a poor indicator of quality, as VMware Fusion, despite running Civilization IV: Colonization at a lower frame rate, had better-looking graphics. Parallels ran Portal better, but the testers were impressed with overall game performance on both platforms.
Final Conclusions
The publication determined the performance differences between XP and Vista are small enough to where users can choose whichever OS they want; that is, the "Vista penalty" is gone. However, the testers did prefer XP's stability and "less annoying" behavior.
When it comes to multiple virtual processors and running in 64-bit, it was concluded that 64-bit wasn't helpful for anything but computationally intense applications like CAD, so most users won't need them. The ability to address more memory in 64-bit mode was similarly deemed irrelevant since many Windows applications and drivers aren't yet 64-bit compatible, and there are plenty of reports of problems in virtual machines trying to run them to recommend against it until technology improves.
Finally, MacTech concluded that XP with 1GB of RAM and one virtual processor will be more than adequate for the majority of users. In fact, more RAM meant longer virtual machine launch times, suspends, and resumes. The testers recommend using only as much memory as needed.
Parallels Desktop was the clear favorite for its small footprint (it only takes more RAM from OS X when needed) and performance advantages in every environment except one. MacTech, though, believes both are excellent products that can run Windows XP and Vista well, so it will be up to the user to decide based on whichever factors are most important in each individual case.
The full 5-page report includes a plethora of detailed test results and graphs. It's recommended for all prospective or current Mac virtualization users.
Purchasing Parallels and VMWare Fusion
Parallels 4 retails for $79.99 but can be purchased and shipped from Amazon for just $49.99 after an instant $10 savings and an additional $20 rebate. For those readers looking to purchase Parallels 4 and download it immediately, Parallels offers this option through its own website for $79.99. However, readers can get $10 off the digital download through this link. (Note that Parallels automatically adds a $6.99 "Extended Download" option to the shopping cart of all of its online orders, which can be removed easily by clicking the "X" to the far right.)
Fusion also retails for $79.99 but Amazon is offering 24% off VMWare Fusion 2.0 ($60.49), though readers interested in saving another $8 may elect to purchase 1.0 for $52.29 and upgrade to 2.0 at VMWare's website for free.
MacTech Magazine ran both virtual machines through thousands of real-world tests, timed with a stopwatch, to find out which performs the fastest. The results are in a five-page story at its website, which AppleInsider highly recommends that prospective buyers of either solution read in full.
Acknowledging Apple's own Boot Camp, which can boot natively into Windows, MacTech decided to look only at virtual machines since they do not require restarting to switch between operating system environments. The publication was also interested in determining which operating system - Windows XP or Vista - offers the best performance. The testers kept an eye on the differences between larger memory configurations, multiple processors, and 64-bit versions of the OS.
To facilitate these tests, they used four different models of Macs to cover the range of machines actual customers would use with more than 2,500 total tests. Again, each was timed with a stopwatch for real-world results to avoid difficulties in running diagnostics on virtual machines that don't always properly keep track of the time it takes for tasks to execute.
Specific tests included: performance (launch and CPU), file and network IO, footprint, application launch and performance, and tests of 3D and HD graphics and games.
Parallels Desktop was the clear winner, running 14 to 20 percent faster than VMware Fusion in the majority of overall averages. There was only one exception when VMware was faster: 32-bit Windows XP on two virtual processors, which ran about 10 percent faster than Parallels.
Overall Test Results
Parallels
14% faster running 32-bit XP and Vista on a single virtual processor20% faster running 32-bit Vista on two virtual processors15% faster running 64-bit Vista on two virtual processors
VMware Fusion
10% fasster running 32-bit XP on two virtual processors
Launch and CPU Tests
In using Windows' compression tools and MP3 to WAV transcoding, Parallels was faster in every version of the test, ranging from 7% faster when launching virtual XP and resuming from suspend after a full restart, to 31% faster after a full Windows boot on Vista. Â*The contest was closest when both Parallels and VMware Fusion were launched without restarting, allowing the benefits of caching. Â*In that test, Parallels was 17 percent faster in Vista and 22 percent faster in XP, but the range was only zero to one second.
Application Launch Test (Word and Outlook)
Here, there was no clear advantage for either virtualization engine. Â*When launching Microsoft Word after a full reboot, Fusion was faster by 8 percent on XP while Parallels trumped with Vista at a 33 percent faster clip. Â*Without a restart, Fusion notched a sweep, while the two again split with launching Outlook (Fusion narrowly won with XP, Parallels with Vista).
Application Performance Tests
MacTech did a global search and replace of about 95,000 items in a very large Word document, and in Excel, they ran a macro to generate a large amount of random numbers, filling cells with them. Â*PowerPoint rendered slides with transitions as quickly as possible.
In the Word test, Parallels was faster under both XP (4 percent) and Vista (9 percent), with similar results on Excel (14 percent faster on XP, 13 percent on Vista). Â*However, Fusion did score a victory with XP on the PowerPoint test, albeit a small one, at two percent. Â*On Vista, it was a tie.
Internet Explorer Speed Test
\t
MacTech created a large, complex page in HTML served from a local LAN server using common HTML elements and no JavaScript.Â* VMware Fusion took significantly longer across the board, so they removed the results from the overall analysis to avoid it skewing the results. In general, Parallels was more than 80% faster in XP and Vista both with and without SSL.
Footprint Tests
Three footprint tests were also run. The first booted Windows, let it finish, and then measured CPU while both Mac OS X and Windows were sitting idle.Â*The second test was similiar but evaluated launches of Word and Outlook, and the third gauged CPU usage while playing a DVD. In all three tests, Parallels and Fusion fared well.
Virtual Machine Memory Usage
As part of its Virtual Machine Memory Usage tests, MacTech noticed that Parallels only took memory from Mac OS X when Windows needed it. The publication ran an "endless loop" macro in Excel that generated random numbers.Â* When battery life approached the end, Parallels paused the virtual machine and the Mac went into hibernate, saving the state of the virtual machine.Â* VMware Fusion suspends the virtual machine before sleeping. The end result was that Parallel's battery usage lasted at least 8 percent longer in both XP and Vista.
File and Network IO Tests
In a highly refined I/O test -- due to problems discovered throughout the testing proccess -- 3.7GB of data (two 1.85GB files) was duplicated and copied to and from a Mac hard drive, LAN Server, and USB Flash drive under both XP and Vista. Parallels beat out Fusion hands down in all but 2 of the 14 tests.
3D and HD Graphics Tests
XP: Smoothly played both 720p and 1080p videos in both environments.
Vista: Couldn't play 720p at all in Parallels. VMware Fusion stuttered on every machine except the Mac Pro. Given 720p results, MacTech didn't bother trying 1080p.
Games
MacTech found that frames per second was a poor indicator of quality, as VMware Fusion, despite running Civilization IV: Colonization at a lower frame rate, had better-looking graphics. Parallels ran Portal better, but the testers were impressed with overall game performance on both platforms.
Final Conclusions
The publication determined the performance differences between XP and Vista are small enough to where users can choose whichever OS they want; that is, the "Vista penalty" is gone. However, the testers did prefer XP's stability and "less annoying" behavior.
When it comes to multiple virtual processors and running in 64-bit, it was concluded that 64-bit wasn't helpful for anything but computationally intense applications like CAD, so most users won't need them. The ability to address more memory in 64-bit mode was similarly deemed irrelevant since many Windows applications and drivers aren't yet 64-bit compatible, and there are plenty of reports of problems in virtual machines trying to run them to recommend against it until technology improves.
Finally, MacTech concluded that XP with 1GB of RAM and one virtual processor will be more than adequate for the majority of users. In fact, more RAM meant longer virtual machine launch times, suspends, and resumes. The testers recommend using only as much memory as needed.
Parallels Desktop was the clear favorite for its small footprint (it only takes more RAM from OS X when needed) and performance advantages in every environment except one. MacTech, though, believes both are excellent products that can run Windows XP and Vista well, so it will be up to the user to decide based on whichever factors are most important in each individual case.
The full 5-page report includes a plethora of detailed test results and graphs. It's recommended for all prospective or current Mac virtualization users.
Purchasing Parallels and VMWare Fusion
Parallels 4 retails for $79.99 but can be purchased and shipped from Amazon for just $49.99 after an instant $10 savings and an additional $20 rebate. For those readers looking to purchase Parallels 4 and download it immediately, Parallels offers this option through its own website for $79.99. However, readers can get $10 off the digital download through this link. (Note that Parallels automatically adds a $6.99 "Extended Download" option to the shopping cart of all of its online orders, which can be removed easily by clicking the "X" to the far right.)
Fusion also retails for $79.99 but Amazon is offering 24% off VMWare Fusion 2.0 ($60.49), though readers interested in saving another $8 may elect to purchase 1.0 for $52.29 and upgrade to 2.0 at VMWare's website for free.
Comments
It's free and arguably better than VMWare and certainly better than Parallels for running just about any OS you can throw at it.
Yeah, why do these comparisons never mention VirtualBox? Come on AI, you should have taken MacTech to task over this. All I know is that VirtualBox exists and runs on OS X, it'd be nice if someone would include it in these head-to-heads to shed light on whether it's any good or not.
This makes me feel good, since I own parallels, and I'm constantly hearing how Fusion is somehow better.
The abysmal stability of Parallels was the main reason why I switched to VMWare. I purchased Parallels first and was so unhappy with it because of this that I switched to VMWare and never had it crash on me. They may have caught up in the stability area now but the fact that they were willing to release obvious "beta" software to the masses and sell it irked me. It also sounds like VMWare will crush Parallels on one of the new Mac Pro's because of VMWare being able to take advantage of some hardware features of Nehalem.
This makes me feel good, since I own parallels, and I'm constantly hearing how Fusion is somehow better.
I use both pieces of software on two machines. I wish they were both VMWare, which I use on my personal computer. The performance differences are minimal, and I think VMWare is just better made and feels better and more intuitive to use.
All the updates provided by Parallels caused a lot of trouble. It was some kind of a nightmare to find the exact release as Parallels provided so many updates and two times I had to re-install Windows. Now, I only use vmware. It doesn't seem slower compared to Parallels but I never used a stopwatch.
I'm not sure which one is the best in terms of performance but in terms of stability vmware outperfoms.
I use both pieces of software on two machines. I wish they were both VMWare, which I use on my personal computer. The performance differences are minimal, and I think VMWare is just better made and feels better and more intuitive to use.
I guess I got in after the stability issues. The current software works great for me. Speed or not, it really makes no difference to me anyway because all I need it for is the ability to test websites in IE. I just can't get excited about running Windows :P
The abysmal stability of Parallels was the main reason why I switched to VMWare. I purchased Parallels first and was so unhappy with it because of this that I switched to VMWare and never had it crash on me. They may have caught up in the stability area now but the fact that they were willing to release obvious "beta" software to the masses and sell it irked me. It also sounds like VMWare will crush Parallels on one of the new Mac Pro's because of VMWare being able to take advantage of some hardware features of Nehalem.
The history of virtualization and emulation on the Mac is interesting. As best I can remember it, Parallels was the first to come out with a virtualization product for OS X, but VM Ware's long and solid history on other OSes brought in a solid fanbase when Fusion was finally released.
Parallels was the first I use. I liked it, especially when in comparison to MS' VirtualPC emulator for OS X, but that is a given for many reasons. My problem with Parallels started when their paid for update cycles became too frequent and the stability started to suffer with these more advanced versions that seemed only focusing on out doing VM Ware's Fusion. I eventually had too many issues with Parallels, tried out Fusion and haven't needed to check out Parallels again since switching.
Another point though is that I'm guessing their tests were done as actual virtual machine disk images. However, I'm sure many consumers would just run their Boot Camp partition in a virtual machine. Supposedly, Boot Camp VM is slower since it's an actual partition that can't be as effectively cached to memory. As well, I've been told that virtual machine image uses OS X's SCSI driver which is better optimized than the IDE driver that is used to read a Boot Camp VM. It'd be interesting to see someone do comparisons using Boot Camp VMs to see if that really changes things.
For most users, they should only dedicate a single processor to their VM since dedicating 2 processors makes it more difficult for OS X to schedule the program: ie. it'll need to find 2 cores that are free so that the both cores in the virtual machine can be kept in sync. Finding room for 2 cores simultaneously is obviously harder than scheduling 1 core. This is especially true for Macs with 2 cores which are most of them. You should generally dedicate fewer cores than your Mac has to allow room for other background programs.
The other thing I noticed is that OS X's scheduler doesn't seem to determine processor affinity very well. Like the process running the VM is constantly bouncing around between cores on my MacBook Pro. It isn't as much a problem on dual cores since they have shared L2 cache anyways, but on non-Nehalem Mac Pros, which use dual die quad core chips, if it's moving between cores on different dies there will be a lot of cache thrashing or memory reading since the caches aren't being used effectively.
I'm on the fence with this one, but I may go back to Parallels for a while.
edit: Title is "Surprised"
Then a client wanted me to do some work with Ubuntu 8.04 LTS. Parallels, even V4, sucked at it. Constant crashes, things not working right.
I tried VMWare and everything was perfect and runs fine. And I got VMWare for $5.00 on a special deal a while back.
My wife still uses Parallels, but I now use VMWare across the board.
Gregor
As far as performance and features go, they seem periodically to leapfrog one another with new versions. Which is faster may depend on the version life cycle point at which you test them, but in any case the differences aren't huge. Fusion, our more recent installation, runs very briskly on a 24" iMac. Frankly the virtual machine speed isn't a limiting factor at this level of performance. Of course I used to use VirtualPC, so by comparison both Parallels and Fusion are speed blurs!
I do agree that Fusion has a more polished feel to it. It's very unobtrusive in everyday use.
VirtualBox looks interesting for personal purposes, but in a production environment I need fully documented and supported commercial software.
I used parallels for a really long time. But it was unstable and the one factor never mentioned in the review is its CPU usage was constantly 25% or greater. Even when I used the latest version [installed it last night after Windows died due to registry corruption caused by a piece of software I installed] it still pegs CPU usage higher than VMWare. Uninstalled Parallels, installed Fusion, and sure it runs a bit slower but it is more stable and my laptop isn't overheating from constant CPU usage of 25% or better.
I really do like Parallels more overall, but could not put up with the instability and the need to reformat and reinstall Windows every 4 months because of Parallels ongoing instability. It might be stable now, but VMWare treats me well.. and I have been burned by too many Parallels upgrades being highly unstable.
- poor stability
- non-existent support
- ridiculous upgrade policies (you'd have had to purchase Parallels several times over by now to stay current, while my VMWare license, purchased for less than the price Parallels wanted to upgrade my buggy-as-heck first release, keeps going and going)
No, thanks.
It's free and arguably better than VMWare and certainly better than Parallels for running just about any OS you can throw at it.
Trying it now. Thanks for the tip.
I can't scale back a volume's size though, I wish I started out smaller. I have only installed three apps and the security updates, and the image is already about 8GB in size, and this is Windows XP. I thought XP was a lot smaller than this.
The abysmal stability of Parallels was the main reason why I switched to VMWare. I purchased Parallels first and was so unhappy with it because of this that I switched to VMWare and never had it crash on me. They may have caught up in the stability area now but the fact that they were willing to release obvious "beta" software to the masses and sell it irked me. It also sounds like VMWare will crush Parallels on one of the new Mac Pro's because of VMWare being able to take advantage of some hardware features of Nehalem.
This was also why I switched, & we run an ESX virtual server cluster so I like being able to just pull down an ESX VM to my machine & run it if I want.