Which Mac is better for running Windows?

MacTech releases its annual head-to-head analysis of how VMware Fusion and Parallels Desktop perform with various types of tasks and running both Windows 7 and Windows 8. It also examines how various Mac models performed.
Written by David Morgenstern, Contributor

This time of the year Mac managers may look forward to the new product announcements at Macworld/iWorld Expo, the advertisements during the Super Bowl, and MacTech Magazine's annual extensive analysis of VMware Fusion and Parallels Desktop performance on a variety of system configurations.

When I spoke to Neil Ticktin, MacTech Publisher, at last week's Expo in San Francisco, he stressed that the report isn't a product review. Instead, he said it is a "benchmarking analysis," meaning that it measured specific performance characteristics as well as issues that came up in the testing. It didn't look at product features and each products user interface, and how it might work in a specific workflow with certain applications, services and data types. This is also often the grist of discussion at MacTech's Boot Camp conference sessions.

In both cases, as with other MacTech benchmarks, we tested performance of the types of things that everyday users typically do. In this case, it was not just testing the raw performance of the Windows OS, but also commonly used Windows applications. Like last time, based on reader feedback, we paid attention to 3D graphics and gaming. We also looked at how well the products performed supporting the new Retina screens.

The goal was to see how VMware Fusion and Parallels Desktop performed, under Windows 7 and Windows 8. Furthermore, we wanted to see some of the differences with different Mac models with different graphics and processor types.

And MacTech found differences.

For example, the report looked at how the Windows virtualization was handled on a variety of sizes of the Retina Display. Interface items such as icons looked much sharper on the Retina Display (we would hope so). I was interested by a problem with Windows 8 launching IE in a small-size virtual machine window. Parallels handled the issue while VMware Fusion couldn't launch from the Start Screen interface.

Windows 8 has requirements for screen size that can be an issue for the smaller screen Macs such as the MacBook Air 11". We noticed this when launching Internet Explorer on Windows 8 in a smaller virtual machine window. VMware Fusion cannot launch the version of IE shown on the Windows 8 Start Screen, but it can launch Internet Explorer 10 from the Windows 8 desktop (it's a different version of IE). Parallels, on the other hand, is able to launch both versions of Internet Explorer.

Both VMware Fusion and Parallels Desktop tout support for Apple's Retina displays. In reality, however, they do it quite differently. In VMware Fusion, you check a box to enable Retina support. When you do that, the resulting window is tiny (about one quarter the size). What we didn't realize until VMware explained it to us is that at that point, you need to manually change the size of "text and other items" in Windows (7 or 8). To do this, within Appearance and Display, you select the "custom sizing options" and choose the scale (about 200%, but since Windows has a bug at 200%, it's best to use 199%).

In Parallels, you simply choose whether you want scaled (for those items that aren't using scalable elements), Best for Retina (scaling items and giving you higher resolution) and More Space (which gives you the maximum screen real estate). This is similar to how Apple gives you choices for the Retina display in OS X's Display preference pane. Parallels takes the approach of automatically changing the sizing of items, as well as making other adjustments to the interface so that they are reasonable sizes automatically.

However, both VMs have "incredible clarity of text and objects" on a Retina Display, MacTech said. Even better for Windows users on Macs with Retina Displays, the Parallels experience is better than the usual native Windows experience.

Trust me when I tell you that, in person, the difference between Retina on Windows and is even more astounding than for OS X.

One worry with earlier iterations of these virtualization architectures was overhead. In this year's analysis, Parallels came out on top by a few percent. MacTech said that neither solution will bog down a Mac's performance. However, there was a battery power difference between them, which could be important for some machines and some users. It appears that Parallels moved stretching out battery life up the new feature list.

The result is noticeable. In our tests, Parallels got 40 percent more battery life on a virtual machine sitting idle than VMware Fusion on Windows 7. This was a couple of hours more of battery life on the MacBook Pro.

If you are using your virtual machine for typing in Word, or using Excel, this is a reasonable approximation. That said, if you are "pegging" the CPU on your machine, or using an optical drive or hard drive, the difference will be less. For most travelers, the light CPU impact of Word/Excel or Internet Explorer use is probably what they are using a virtual machine for.

Check Out: MacTech announces BootCamp II seminar schedule

MacTech ran extensive real-world tests on games and 3D performance. Now, I don't really consider this important in a business setting. If you are working on graphics or video, then you should be in OS X, not Windows. But I recognize that everyone needs their own pastime.

The report also made an interesting point about memory with virtual machines, that may be counterintuitive for our experience with everyday native OS computing.

When it comes to RAM, less is often better. You see, the virtualization products need to do a lot with RAM, so only increase it if you really need it and make sure you have enough actual RAM in your hardware. As a general rule, 1GB virtual machines are the way to go for best boot, suspend and resume times. If you are running virtualization, try to have 8GB or more actual RAM in your machine.

There's so much more in the report. As Ticktin said, each customer will have to determine which solution fill be best for their workflow and clients. Each will provide reasonable performance and good value.

Editorial standards