This article originally appeared in TidBITS on 2011-09-19 at 10:35 a.m.
The permanent URL for this article is:
Include images: Off

Mac Virtualization Update: VMware, Parallels, and VirtualBox

by Joe Kissell

When the first edition of my book “Take Control of Running Windows on a Mac [1]” came out, just over 5 years ago, running Windows at anything approaching full speed on a Mac was still a novelty. At that time, the leading options were Apple’s Boot Camp [2] dual-boot system and Parallels Desktop [3], a virtualization program that let Windows run within Mac OS X — but without the heavy performance penalty of PowerPC-based software like Virtual PC. Then VMware got into the game with their own virtualization program, Fusion, and things started to get really interesting. Every few months or so since then, either Parallels or VMware has rolled out an upgrade that made their product the apparently superior choice for a while. The two companies have battled on features, performance, and price, and that competition has raised the level of quality of virtualization on the Mac. It’s been a good thing. (The emergence of a solid yet free competitor, Oracle’s VirtualBox [4], hasn’t hurt either! I’ll get back to VirtualBox later in this article.) Of course, Apple has consistently improved Boot Camp as well — and Windows itself has become much, much better too.

What hadn’t happened until this month was major releases of both Parallels and Fusion appearing at almost exactly the same time. I’ve used both Parallels Desktop 7 and VMware Fusion 4, and I thought this would be an appropriate time to offer a brief “state of the union” on virtualization software for the Mac. Before I begin, I want to make a few things clear:

The Similarities -- As I write this, I have yet to see a complete, objective feature comparison chart for Parallels 7 and Fusion 4. A checklist on the Parallels Web site hasn’t yet been updated to reflect the changes in Fusion 4 (and would, naturally, be expected to favor Parallels regardless). The Wikipedia comparison page [7], too, still reflects Fusion 3.1. In any case, checklists of this sort can be difficult to interpret when applications implement similar features differently; they can be spun all too easily to make it appear as though any application is superior; and they seem to give equal weight to all features, whereas most users care about only a few. Rather than attempt to create my own table comparing features exhaustively, I want to concentrate on the highlights.

Both Parallels 7 and Fusion 4 can do all of the following:

(Again, I’ll say a few words about how VirtualBox stacks up later on.)

Running Parallels 7 and Fusion 4 side by side on my Mac, I’m struck by how similar they are. Their terminology and user interfaces differ somewhat — but much less than in years past, and little enough that I can easily lose track of which application I’m running at any given time.

The Differences -- With so much the same, it seems that the two developers are increasingly struggling to come up with new ways to differentiate their products; there’s just not much left that one could ask from a virtualization program. However, I can still point to some spots in which one or the other appears to have the upper hand.

Parallels 7 lets you share video cameras (including built-in iSight or FaceTime cameras and external USB video cameras) between Mac OS X and Windows. In Fusion 4, by contrast, you can use a camera in either host or guest operating system but not both at the same time. For people who don’t already have Windows 7, Parallels offers a way to purchase and download it from within the application. I haven’t tried this myself, but I’ve heard the process is far from user-friendly. Parallels also offers an iOS app called Parallels Mobile [8] ($19.99, but currently on sale for $4.99) that lets you open, suspend, resume, and interact with virtual machines running on your Mac — as well as with Mac OS X itself. That’s nifty, but not extraordinary; you can use any VNC client (including some iOS clients that are free) to connect to any Fusion virtual machine, although configuration requires a few less-than-obvious steps. Parallels also lets you suspend a Boot Camp virtual machine, whereas in Fusion you can suspend only guest operating systems with their own virtual disks.

Fusion 4 offers a few nice capabilities not found in Parallels 7, too. Fusion is now a self-contained application that you can install or uninstall by drag-and-drop or even run from a USB flash drive, and when you quit it, no background processes remain active. (Parallels 7 insinuates itself more deeply into your system, and leaves some bits running after you quit.) Fusion 4 also supports virtual Bluetooth (letting a virtual machine directly access Bluetooth devices) and Remote Disc (so you can access a DVD or CD on another Mac or PC on your network), both capabilities Parallels 7 lacks. Both applications let you shrink virtual disks to save space; in Fusion 4 you can do so even if the virtual machine has snapshots, while in Parallels 7 you must delete all snapshots before shrinking the disk. And Fusion doesn’t restrict your choices to shutting down or suspending a virtual machine; you can also pause it, temporarily freeing up system resources, and then resume it instantly. Finally, Fusion can claim to be a completely 64-bit, Cocoa application, while Parallels has some 32-bit components and uses Cocoa along with a couple of other frameworks. I’m at a loss to know how being a 64-bit Cocoa app affects any ordinary user’s everyday experience, but it does at least seem to be a good foundation for the future.

Performance -- The very first version of Parallels I used — on a Mac that is now too old even to run Lion — ran Windows so fast I could hardly believe my eyes. It felt, to me, just like running Windows on modern PC hardware. I experienced no lags, no delays — nothing to complain about at all. So, when the next version of Parallels touted vastly improved performance, I didn’t see what was interesting about that. I hadn’t felt anything I needed to do was too slow, so I didn’t need or notice faster speed. Now that both Parallels and Fusion are many times faster than they were in the early days, and my hardware itself is also faster, I find almost all claims about speed to be meaningless in my particular case.

Now, understand that I’m not a gamer. (And if I were a gamer, I wouldn’t be the sort who’d get hung up about minuscule differences in frame rate.) I don’t edit video or render 3D animations or perform complex mathematical modeling in Windows. Maybe the users who do those sorts of things can genuinely feel the difference between one version of Parallels or Fusion and the next, but I can’t. What I have done is run office applications such as Outlook and Word, Web browsers such as Internet Explorer and Safari, and video apps such as the Amazon Unbox Video Player. For uses like those, even older versions of both Parallels and Fusion felt perfectly peppy to me — and even when running on much slower hardware than I have today. In short, for the ways in which I use Windows, I couldn’t care less about whether X is 10 percent faster than Y in some benchmark or other. It doesn’t affect me, and I’ll bet that the same is true for a large number of Mac users who run Windows.

I say all that to put my remarks about performance in perspective.

In 2010, MacTech did a series of benchmark tests [9] comparing the performance of then-current versions of Parallels (version 5) and VMware Fusion (version 3). In that comparison, Parallels soundly beat Fusion in all but a tiny handful of tests. It wasn’t close, and it wasn’t ambiguous: Parallels was definitively ahead in terms of objective speed tests. That fact apparently led a lot of people to conclude that Parallels 5 was therefore better than Fusion 3, and I thought that represented unsound reasoning. The fact that a Porsche can go faster than a pickup truck doesn’t mean it’s better; it’s only better if you’re planning on driving faster than the pickup’s top speed, and you value speed more than cargo capacity. I do not in any way dispute the benchmark results; I simply want to point out that for many users, they’re irrelevant.

Anyway, that was then; what about now? Both Parallels 7 and Fusion 4 claim much better performance than their predecessors, but what everyone wants to know is how they compare to each other. VMware sent me a reviewer’s kit that contained a bunch of benchmarks comparing Fusion 4 and Parallels 7. According to those test results — which VMware has not yet made public — Fusion 4 is very slightly ahead of Parallels 7 in most respects, and very slightly behind it in a few. On the whole, these tests suggest that the two applications are so close to each other in terms of overall performance that the differences are statistically insignificant. Naturally, I take test results with a grain of salt when they come from a developer comparing its own product to the competition’s. I have no reason to mistrust VMware, but it’s only natural to cast one’s own product in the best possible light. One of these days, MacTech (or another independent tester) will publish their own results, and we’ll be able to evaluate the comparative performance with greater objectivity. I can’t predict how those tests will turn out, but I can predict that they won’t change my perception of how fast I can type or check my email. For all practical purposes, these two programs are equally fast.

And that’s the real point. Benchmarks test specific, objective, measurable things. They may have no bearing whatsoever on the tasks you personally find important. A commenter on my announcement about Fusion 4’s release claimed, and I confirmed, that in certain situations, Parallels 7 can resume a suspended virtual machine more quickly than Fusion 4 can. That’s the sort of thing that may or may not show up in benchmark tests. And maybe it’s extremely important to you; it happens not to be important to me. As with all such statistics, I urge you not to get caught up in abstract numbers.

Instead, I suggest — as I have done for years — that you download the free trial versions of each app and try them out for yourself, running the kinds of software you need and doing the sorts of tasks that are important to you. If you can’t feel a difference in performance, then for you, there is no difference. And if you do feel a difference, that’s useful to know; but use your time with the trial versions to evaluate their other features, their interfaces, their documentation, and their technical support too. You may find that factors such as these tip the balance in one direction or the other.

Pricing and Licensing -- For a long time, Parallels and Fusion have had the same retail price, $79.99, but thanks to various widely available coupons, sales, bundles, and other promotions, the typical street price for either has been closer to half that. With their latest releases, upgrade pricing for both apps has raised some eyebrows.

Parallels charges $49.99 for an upgrade from versions 5 or 6 (it’s free for those who purchased Parallels after 1 August 2011) while offering registered Fusion users a competitive crossgrade price of only $29.99. I can’t help but notice an interesting trend here. Parallels has had paid upgrades each of the past 5 years; all were $49.99 except the upgrade to version 4.0, which was $39.99. If you had purchased Parallels 1.0 for full retail price and bought every upgrade thereafter, you would have paid a total of $559.91 to date (not counting the cost of Parallels Mobile, if you purchased that too). That makes Parallels an expensive product to keep up with, and the fact that Fusion owners get a lower price than those who have been giving Parallels their money every year is bound to feel a bit harsh to those long-time customers.

Meanwhile, VMware is selling Fusion 4 to everyone — upgrading or not — for the reduced price of $49.99 through the end of the year (it’s free for those who purchased Fusion 3 after 20 July 2011, the day Lion was released). Bafflingly, this promotion has provoked an outcry that VMware isn’t offering special “upgrade” pricing. But they are; it’s just that they’re offering that lower price to new purchasers too for a limited time. In its entire history, Fusion has had only two paid upgrades, the previous one being to version 3.0 (for $39.99). So a Fusion owner who bought the original release at full retail price and every subsequent upgrade would have paid only $169.97 to date, a mere 30 percent of what a Parallels owner would have paid. Moreover, Fusion’s new license agreement says that, except for commercial and educational settings, one license is good for all the Macs you own or control; the Parallels license requires you to buy a separate copy for each Mac. And to top it off, a coupon code has been making its way around the Internet; you can save 20 percent ($10) on your purchase of Fusion 4 by entering the code FUSION20 at checkout.

In summary, someone with Parallels 6 must pay $49.99 to upgrade to Parallels 7, but can get Fusion 4 for $10 less — $39.99 — while someone with Fusion 3 can upgrade to Fusion 4 for $39.99 or switch to Parallels 7 for $10 less — $29.99. But if Parallels continues its past upgrade pattern, switchers enticed with this low pricing will be invited to lay out another $50 next year for version 8 (and likewise in future years), which seems to me to blunt the offer’s appeal somewhat.

What about VirtualBox? -- I promised I’d have a few things to say about VirtualBox. VirtualBox started as an open-source project, and even though it’s now owned by Oracle, it’s still available in an open-source version (you compile the code yourself) or in precompiled, free-for-personal-use forms for various platforms. Much is made of its (lack of) cost, but keep in mind that the $50 you could pay for Fusion is a fraction of what you’ll have to pay for Windows itself. In other words, free is good, but running Windows on a Mac means shelling out some cash anyway; and to a certain extent, you get what you pay for.

Don’t get me wrong: VirtualBox is no slouch, and it’s gotten dramatically better since its early days. It supports a wide variety of host and guest operating systems and has a long feature list. Like Parallels and Fusion, it offers a mode in which the virtualization environment mostly disappears, putting Windows applications on the same visual footing as Mac apps; it also features shared folders, snapshots, 3D graphics acceleration, and support for most common hardware — including multiple displays. Its performance is solid, if not quite to the level of its commercial peers. And it even lets you assign up to 32 virtual CPU cores to a virtual machine.

So what’s not to like? Well, as of version 4.1.2, VirtualBox runs just fine under Lion but it doesn’t take advantage of Lion-specific features such as Mission Control, Launchpad, and full-screen mode; nor does it support running Lion or Lion Server as a guest operating system. It lacks automated setup of Windows virtual machines, won’t run Windows from your Boot Camp partition or use your Mac printers automatically, and has less integration overall with Mac OS X. And the user interface, as well as the documentation, betrays the software’s open-source roots: it just isn’t pretty to look at.

And yet, VirtualBox truly is adequate — and I don’t mean that as a dig. It gets the job done, and for those who need only occasional access to Windows on a Mac, giving up certain features for a savings of $50 might be an excellent compromise. You have nothing to lose by trying it, even for an extended period of time, and you can always migrate to a commercial alternative later if you want to.

Final Thoughts -- With each passing year, the number of tasks that can be accomplished only in Windows, and not natively in Mac OS X, decreases. This is due in part to the rise of Web applications, in part to Windows apps being ported to Mac OS X, and in part to new Mac-only or cross-platform applications that are superior to the older Windows-only options. Lots of people still legitimately need to run Windows on a Mac, but that number is certainly shrinking.

Boot Camp continues to work fine, but I gave up on it long ago. I consider it unnecessarily awkward and inflexible compared to virtualization software. Plus, the performance gap between Boot Camp and virtualization has shrunk significantly, and very few applications and external devices work under Boot Camp but not in a virtual machine. The only real question for the vast majority of people who still need to run Windows on their Mac is which virtualization program to use. As of Parallels 7 and Fusion 4, that’s more of a toss-up than ever. Although I’m sure both developers can still squeeze out a few more percentage points of performance, my sense is that we’re close to reaching the theoretical limit of how fast Windows can run on a Mac (and it’s plenty fast). In addition, improvements in Windows itself and in the various virtualization programs have made setting up and using Windows on a Mac so easy and seamless that users can often ignore the differences between operating systems, freely downloading and running nearly any software without regard for the platform it was written for.

Over the years, I’ve read hundreds of comments all over the Internet by users of one of these products or another who became disenchanted with their first choice for some reason and decided to switch to another one. Complaints have ranged from bugs and missing features to poor customer service to user interface aesthetics. Although I’ve never tried to calculate the totals, my impression is that users have historically been more likely to become unhappy with Parallels and switch to Fusion than the other way around, and that more people have switched from VirtualBox to a commercial alternative than the reverse. But as I said earlier, everyone’s experience is different. I wouldn’t try to convince you that any of these options is superior all around or the best choice for everyone. You’ve got to decide what’s best for you — and maybe that’s different now from what it was last year; maybe it’ll be different again in another year. If you’re happy with what you have already, you’ll be happier still after upgrading to the latest version of the same program. If you’re unhappy, this is the ideal time to check out the competition.

But be circumspect when you read claims about this type of software. I’m aware of some misinformation being spread deliberately, of some unscrupulous tactics being used to promote one of these products. That’s a pity; Parallels, Fusion, and VirtualBox are all fine programs that should be able to compete solely on their merits. Try them yourself and make up your own mind — remember, your experience is the only one that truly counts.