18

Virtualization Benchmark Showdown – Parallels 10 vs. Fusion 7 vs. VirtualBox

Posted by Jim Tanous on September 17, 2014
Parallels 10 Fusion 7 VirtualBox Benchmarks

Passmark PerformanceTest

Similar to PCMark, Passmark PerformanceTest 8.0 is another benchmark that aims to take an overall look at system performance. It measures CPU, 2D, 3D, memory, and disk performance to calculate an overall Passmark rating, which forms the basis of the Passmark benchmark database. Results are reported both as an overall score and as individual scores within each testing category. A higher score equals better performance.

Parallels 10 vs Fusion 7 vs VirtualBox Benchmarks Passmark PerformanceTest

From a computational and graphics perspective, Parallels 10 wins again, although Fusion 7 takes a small victory in terms of disk performance. Parallels’ large margin of victory in the 3D graphics test requires further explanation, however. Some benchmarks, such as PCMark, will refuse to post a score if a system fails any of the included tests. PerformanceTest, on the other hand, will still spit out a result in the event of a failure, but heavily penalize the system as compensation.

That’s what happened here; the 3D Graphics test includes portions that test DirectX 10 performance. Parallels offers limited, but functional, DirectX 10 support. Fusion and VirtualBox do not. Parallels would have still likely won the category absent this reality, but not by as wide of a margin.

Also interesting is VirtualBox’s abysmal Disk score, although you’ll see later on that the disparity is somewhat confirmed in real-word file transfer tests.

[one_half padding=”0 5px 20px 0″]
1. Introduction
2. Hardware, Software, and Methodology
3. Geekbench
4. 3DMark06
5. Cinebench R15
6. PCMark 8
7. Passmark PerformanceTest 8.0
[/one_half]

[one_half_last padding=”0 0px 20px 5px”]
8. Handbrake Video Encoding
9. iTunes Audio Encoding
10. File Transfer Speed
11. Virtual Machine Management
12. Battery Life
13. Conclusions
[/one_half_last]

18 thoughts on “Virtualization Benchmark Showdown – Parallels 10 vs. Fusion 7 vs. VirtualBox”

Georgina Lewington says:
excellent review. it is great to see standards from someone other than commonalities. I do find it annoying though, that all similar opinions center around operating ms windows VMs. While strong focus is put on managing a smooth ms windows incorporation on parallel’s part, it is in the end a hypervisor and would be awesome to learn more about efficiency with OSX and A linux systemunix VMs.

distance learning courses uk

Reply
Diggle says:
Would have been nice to see these tests against a pc with similar specs running windows.
Reply
TekRevue says:
Well, that’s not the point of this comparison. The focus here is, given identical hardware, which virtualization platform performs best? The actual speed of native Windows performance doesn’t matter in this context.

Where native Windows performance does matter is, once a user has chosen a particular virtualization platform, is it fast enough to meet their needs versus a separate PC or Boot Camp? That’s why we provide native (Boot Camp) numbers in our individual app reviews, like the one for Parallels 11 and one that should be available shortly for Fusion 8.

Reply
Chris says:
Great job – I would have loved to see how CrossOver (CodeWeavers) compares to these – even though it’s a slightly different product, I always wanted to know if it is really faster than Parallels?
Reply
TekRevue says:
Great idea, Chris. New versions of Parallels and Fusion should be coming this fall. We’ll try to add CrossOver to our next round of benchmarks for those products.
Reply
fusspot1 says:
If you’re running windows on the MAC then there is no question but to use parallels.

I’ve just switched to parallels after many months of frustration using VMWARE Fusion. Parallels is faster in every respect.

Using VMWARE Fusion, accessing a shared mac drive was extremely slow, and attempting to compile large projects with Visual Studio was terribly slow. In unity mode, moving windows across the desktop there was visible painting lag and access through to an FTDI USB device was also slow. This was using an iMac27 8 core i7 with 32GB ram and 3TB fusion drive running Yosemite so there are no excuses for the abysmal performance. I tried fiddling with the network adaptor settings etc with no real improvement gains.

Switching to Parallels, everything runs instantly, including accessing the shared drive, compiling with visual studio and accessing the USB controller.

I noticed that Parallels ships with its own set of drivers that probably explains the difference in performance.

Reply
Fran says:
I used Fusion 6 and didn’t experience your problems, although I used it only with Multicharts and on a Full screen set-up.
Actlually I was using Fusion 7 I, wothout any known reason my internet connection was dropping, leaving me with blind open orders: I was attributing the loss of connection to my brokers servers, but it wasn’t, so I switched to Parallels 10 and my problems were solved.
Maybe the connection problems in my Fusion machine were solvable, but, since it’s working fine with Parallels, I didn’t take the time to find the cause of the problem.
I agree with you that Parallels graphics engine is much faster than Fusion’s. Anyway, I cannot see why you had such a huge lack of performance using VmWare Fusion. Did you check the number of processors in WmWare. The default value in VmWare is 1 processor, while the default in Parallels is 4.
To get a fair comparison there is the need to set this to the same number. Also, as I said, If you really want to take the max performance, there is no need to be afraid of setting the nr of processors to the 8 threads of a i7 CPU.
Reply
TekRevue says:
Hi fusspot1,

Glad that Parallels is working well for you. I just wanted to mention that you may have had a configuration issue with VMware Fusion. You mention that “Parallels ships with its own set of drivers.” VMware Fusion, too, has drivers and custom software that you need to install in your guest operating system (with your guest operating system booted, go to Virtual Machine > Install VMware Tools from the OS X Menu Bar). Without these tools/drivers, your guest operating system performance, particularly with Windows, will be significantly impaired.

I’ve used/tested every version of both Parallels and Fusion since they were released, and while Parallels has indeed been faster in many versions or for certain tasks, the performance gap shouldn’t be as bad as you’re seeing. If you’re happy with Parallels, there’s no real reason to switch back to Fusion as both are relatively comparable on features at this point, but if you want to try the next version of Fusion (which should be released this fall) via a free trial or something, just check to ensure that you’ve installed the Fusion tools/drivers before starting to use your guest operating systems.

Happy virtualizing, and thanks for stopping by and taking the time to comment!

Reply
Fran says:
My problem wasn’t performance, but internet connection gaps. Since Parallels solved the issue I didn’t bother to fight the solution with VMWare.
Komrad says:
What if you want to run Linux on a Mac? Then which?
Reply
fusspot1 says:
Hi Konrad, I haven’t tried Linux with parallels so can’t comment.

I have a few Linux vms that I tried with Fusion that looked to be ok but I didn’t do anything computationally challenging with them.

fusspot1 says:
I’ve just read this article:

http://log.paulbetts.org/parallels-the-best-android-emulator-ive-ever-used/

I’ve imported my Fedora and Ubuntu VMs from Fusion into Parallels and all is well.
I’ve downloaded the pre-built parallels Android VM and that runs up without issue.

Robin Degen says:
You do realize that you can never “just run the install executable” from the disk right? That’s not a negative thing, that’s technically impossible. You as a reviewer should know this. It’s a Virtual Machine, not an emulator.
Reply
TekRevue says:
What are you referring to?
Reply
Robin Degen says:
Sorry, nevermind me i’m not completely awake and ended up relying on the wrong topic. Can you delete the message?
disqus_f23yWYhDc9 says:
Stupid article.
Fact: VMWARE is faster than Parallels.
Reply
Bao Nguyen says:
Awesome article, thanks! If you could please include bare-metal results as well for even better comparison!
Reply
Tim says:
I have just setup a windows bootcamp partition on my mac and am using it both to boot into and from OSX via parallels. This made me wonder if it might be more performant (or less) to have a windows VM using boot camp partition than a normal virtual disk. Have you considered this?
Reply
Sean says:
Excellent review! We all know there is a strong correlation between performance and energy usage. So I wonder if vmware fusion is performing worse compared to Parallel desktop in most test cases is because it’s trying to balance the power usage and that is why it beats parallel desktop handily in battery life test. If only there is a way to configure them to not care about power usage at all and then compare the number again.
Reply
Monomeeth says:
Good review. However, one thing I would like to see reviewed is how well the VMs manage hardware devices. I know this isn’t an easy thing to test per se, but it could be done.

For example, I have found that some USB devices don’t play nicely in some VM environments. They seem to connect ok, but then the connection is lost regularly and you need to disconnect/reconnect (sometimes multiple times) to get them working okay. That is, until the next time they disconnect.

Probably no need to test for things that would usually be connected to the Mac anyway (e.g. printers, iPhones/iPads, etc) – but you could identify a host of windows-only devices to test, such as consumer USB connectable weather stations, etc or even testing compatibility with professional devices (such as medical equipment).

Obviously you couldn’t test every model, but perhaps you could test a couple of popular devices in each category. Of course, in some cases, the cost may be prohibitive (such as in the case of medical equipment), but then you could approach the hardware manufacturer to determine if they have done any tests?

Anyway, just thought this was worth mentioning since the reason why some people need VMs is to be able to still use hardware they’ve previously invested in.

Reply
Bart says:
Would like to see the native benchmarks for a lot of this. Interesting comparisons though. Also would be nice to see some more typical Macs (not many of us have the new Pros.)
Reply
RJ Cx Love says:
Useful article
Reply
James says:
It would be great to see the last previous version compared to the latest version here so we can answer the unanswered question -> is it worth upgrading.
Reply
Haravikk says:
Great review, really making me want to give Parallels another try (I got licences for both a while back, but VMWare Fusion was a clear winner at the time). My one concern is the implication of the battery life results, as it implies that while Fusion may be slower in some tests, it’s also more efficient at running virtualisation. While battery life is a clear advantage of that for laptops etc., it also implies that Parallels (and VirtualBox) are going to have a great impact on the host’s performance while running virtualisation.

It would be interesting to see some kind of benchmark that runs a virtualised task, and a host system task simultaneously, to measure the impact that virtualisation is having on the host. Worth considering for a future review when the next Parallels and VMWare Fusion versions arrive?

Reply
Mridul Mahesh says:
An excellent, unbiased review! You have covered most of the important details and provided adequate visuals and description. Thanks a lot Jim! Because of you I could make up my mind to decide which virtualisation software to buy
Reply
Fran says:
Thanks for the review. My only comment is about your conclussions. If you want to extract the most out of your system with math intensive calculations, such as when using Mathematica, Mathlab or intensive stock market simulation then I think Fusion 7 is marginally better ( maybe 10% speedier). If you are using it for heavy 3D graphs and/ or gaming then Parallels 10 is a better choice.

Also, a pitty you didn’t use the whole cores on this test and, also, display the host power figures ( using bootcamp), as to see the amount of penalty when using the virtual machine.
.

Reply
TekRevue says:
Great observations, Fran. To answer some of your points, we didn’t test VMs configured with all of the cores because virtualization with all of a host’s resources can degrade performance of both the host and guest operating systems, resulting in worse than expected performance in many instances.

Somewhat related to that, we didn’t compare native Boot Camp performance because Windows running on Boot Camp would have access to all cores, and only some of the tests, like Handbrake and Cinebench, allow you to manually limit the number of cores/threads used. As a result, Boot Camp would appear unfairly faster in most results, not because it’s inherently faster (which it is), but because it has access to more physical hardware than the VMs.

If there’s demand, however, we’d be glad to run additional tests with VM vs Bootcamp, especially in areas such as gaming. Fran, and everyone else, just let us know what kind of scenarios you’d like tested and we’ll do our best to comply.

Reply
Fran says:
Thanks for the answer. Maybe performing tests on a vm with all cores would get interesting facts.

I tried using win 8.1 on Vmware Fusion 7 running with 8 cores on my Macbook pro mid2012 and I don’t see almost no degradation. Geekbench 3 ( 32 bits) MAC native: 11950, win 8.1 8 cores: 11584.
I’m also using Multicharts and actually running multiple and extensive trading simulations along the day, and I must say the VM is completely stable, although temperarture goes up to 80 degrees celsius( during the intensive simulation task).
What I mean is: the way a virtual machine behaves at its hardware limits says a lot about the goodness of the virtual engine.

TekRevue says:
Hi Fran,

My apologies. I just realized as I was replying to another comment here that I neglected to get back to you regarding VM tests with all cores, and VM vs native Bootcamp tests. We’re not currently set up to perform additional tests at this time (we only get trial licenses to use the software), but we will be ready and waiting to test the next versions of Parallels Desktop and VMware Fusion (which, based on historical release schedules, should be released this fall), and we’ll make sure to add the configurations you recommend to our next round of tests.

As always feel free to contact us if you have any specific testing requests (certain software, a particular Mac model, etc.) and we’ll do our best to include those tests in our review and benchmarks.

fusspot1 says:
I doubt it!
Reply
Fran says:
Please, What do you doubt?
fusspot1 says:
Hi Fran,

I haven’t factored in that this post is 9 months old – I am comparing Parallels 10 with Fusion 6 so things may have been different then – (see my latest post 5 hours ago)

My findings are that the two VM’s cannot be seriously compared. Fusion is so slow as to be a none starter. Parallels is providing instant response from the same windows 7 64 bit vm imported from Fusion as was running really slowly with Fusion.

I don’t believe for 1 minute that anything could run faster with Fusion that with Parallels. My concern is that its misleading for anyone looking to run VMs to even consider Fusion as a viable option for the Mac.

Fran says:
Hi, fusspot1,

The performance in both engines are pretty much the same, but, sligtly higher using Fusion 7.

I did benchmarks in Windows 8.1 using Geekbench3 ( 32 bits) in Fusion 7 and Parallels 10.

The number of processors was set to 8 and 4 GB ram.

Fusion 7: 11.019 http://browser.primatelabs.com/geekbench3/2760384

Parallels 10: 10.893 http://browser.primatelabs.com/geekbench3/2760566

Native: 11.244 http://browser.primatelabs.com/geekbench3/2760589

My test computer is a MacBook Pro (15-inch, Mid 2012), 2.7 core I7 16GB Ram.

fusspot1 says:
Thanks for your feedback Fran.

I’m using 4 cores with 4GB ram and 100GB disk for the windows VM.

Same VM run between Fusion and Parallels.

I’m running a late 2013 iMac 27 i7 8 core 32GB RAM.

Maybe there’s a problem between the Fusion 6 drivers and my machine but either way since Parallels 10 is providing comparatively blistering performance I’ll probably stick with it for the foreseeable future.

Fran says:
I remember the new versions were released just as Yosemite’s release, and I remember some incompatibility of the older versions of both engines with the new OSX version. That, I think may be the main cause of your lack of performance with Fusion 6.

https://communities.vmware.com/thread/480479?start=0&tstart=0

fusspot1 says:
Cheers Fran.

I appreciate your tip on assigning the max cores to the virtual machine as are available on the host machine – I was under the mis-apprehension that I should balance them between the host and VM. This should make my Parallels VM even more responsive.

I will very soon be doing lots of intensive computational maths on a 3D image processing application – I may give Fusion 7 a try at that point.

Many thanks

Pete

Fran says:
The difference is too tiny. Go to the links I provided above and see for yourself if it’s worth. I wouldn’t spend my time on it.
Chris says:
Very interesting observations and thanks a lot for sharing the figures as well – I will definitely give Fusion 7 another try now, since I mostly want to do math intensive calculations (FEM). I really should be doing this natively, but let’s see how it works out 🙂
Reply
DMANO GARLOVA says:
excellent review. it is nice to see benchmarks from someone other than parallels. I do find it irritating though, that all parallel reviews center around running windows VMs. While strong emphasis is put on running a seamless windows integration on parallel’s part, it is in the end a hypervisor and would be nice to hear more about performance with OSX and Linux VMs.
Reply
Warper says:
Excellent article! Thank you, Jim!
Reply

Leave a Reply

Your email address will not be published. Required fields are marked *


Disclaimer: Some pages on this site may include an affiliate link. This does not effect our editorial in any way.