mattgadient.com

VMWare Fusion 3 vs Parallels 5 – Windows GAMING on the Mac

New versions of these popular virtual machines (also known as emulators) have recently come about. In our old comparison we took a look at 4 games, and we do the same again here. Last time, VMWare Fusion was the winner. This time… it’s Parallels. You’ll see why as you read ahead.

VMWare Fusion 3 and Parallels 5 were tested, using a Windows 7 Professional 32-bit virtual machine. Windows 7 was chosen because both Fusion and Parallels now support it, and with XP being continually phased out (and Vista being bloated), it’s the operating system that most people are likely to choose.

So without further ado, let’s take a look at the 4 games chosen this time around. I’ve packed the screenshots at the front with a tiny explanation, and if you scroll down a bit further (about 1/2 way through the article, you’ll see the actual writeup.

Mass Effect (click to see full size):

mass-effect-parallels-5-img1 mass-effect-fusion-3

Both actually had the error message you see on the right. After playing with the Windows Compatibility mode stuff, I managed to get to the error message on the left.

Halo CE (click to see full size)

halo-2-parallels halo-fusion-3

The left side (Parallels) was nice and fast – the 59 FPS you see was the lowest it got, and that was just for the screenshot. Not so on the right side (Fusion) – it looks like snow in the right image. It’s just missing textures. At a whopping 2 fps.

Knights of the Old Republic 2 (click to see full size):

kotor-2-parallels-5 kotor2-fusion-3

Both were very playable, but Parallels offered a perfectly smooth experience. Fusion played quite well, but crashed in the “Advanced Video” option menu.

Cities XL (click to see full size)

cities-xl-parallels-5

Just Parallels is shown, because Fusion only lasted about 5 seconds – not enough time for a screenshot.


Now for a little detail… We’ll start with a disappointment, and then cheer everyone up a little with a few successes.

Mass Effect
Parallels – no rating
Fusion – no rating

In the last comparison, neither Parallels nor Fusion supported PS3.0, and therefore, Mass Effect didn’t stand any hope. With both of the latest versions now supporting DX9.0c and PS3.0, the hope was that Mass Effect would, at the very least, run.

In Parallels 5, there was no such luck. The game would crash before starting up.

VMWare Fusion 3 shared the same fate. Crash on start.

Pretty disappointing. At first I thought it might have been the Steam backup i was restoring from – however I tried creating an XP virtual machine in each VM, and was met with a GPF – something to do with GMatrix2D. A little research showed that many people have had this error even on PC’s – specifically relating to video card drivers. It’s likely that nobody short of ATI or nVidia will ever get this game to run. Update: That’s not quite true – after doing a little searching around I decided to try installing the WineD3D package in both Fusion/Parallels VM’s (the Wine D3D package incidentally also comes with the free VirtualBox VM and has some support for DX9, as well as some experimental support for DX10). In Fusion the game still crashed at startup, but in Parallels the game started up and I was able to create a character, although it crashed the VM as soon as the game’s opening sequence started. The menu reacted slowly, and presumably had it worked the game wouldn’t have been exactly fast. It’s possible that VirtualBox would fare better than Fusion/Parallels, but it choked during the XP install so I didn’t get to try it.

Cities XL
Parallels – 4 stars (the game already has speed issues, emulation makes them even more apparent)
Fusion – 0 stars (it almost ran)

Recently, a SimCity4-like game was released called Cities XL. The game requirements aren’t overly specific, although it states that it’s “Enhanced for DirectX 9.0c”, and lists a GeForce 6600GT or ATI X1600 as minimum video card requirements.

In Parallels, the game ran, and worked quite well. In a city with about a population of 800,000, navigating, building, etc went along without a hitch. It was however noticeably slower than through the PC, particularly on the loading screens (which doesn’t really matter), as well as when zooming (which does matter). Moving around the map from the overhead view had a slight choppiness to it as well.

Much of this can probably be attributed to the game itself – Cities XL has been attacked by many for not being multi-threaded / multi-core. Many PC players have issues with it behaving slowly even on a multi-core machine. Running the game through an emulator certainly doesn’t help an already difficult situation.

Fusion was much slower than Parallels getting in. Once in your city, you have about 5 seconds of extreme slowness before it either bluescreens, or the video becomes checkerboarded.

Halo CE (custom edition)
Parallels – 5 stars (so good you’d never know it’s emulated)
Fusion – 1 star (technically it runs….)

Halo’s a fairly old title, and it only requires DirectX 9.0, so it would probably run on the older versions of Fusion/Parallels. In any case, it ran extremely well on the new version of Parallels at 1440×900 resolution, with absolutely no issues. By default it’s set to run at a fixed rate of 30FPS but once it became clear that it wasn’t having any issues, it was time to see how high we could go. Fusion was a whole other story.

Parallels framerate were generally between 200-300 FPS. Even in windowed mode it didn’t drop very much. Every so often with the right scenery and shooting it would drop below slightly 100 FPS – in fact the lowest it ever got to was about 60 FPS during the screenshot.

VMWare unfortunately didn’t fare well at all. It averaged 1 FPS, maxing out at 7 FPS. Very frequently the textures turned white as well (see the screenshot at the beginning of this article) both in Windowed and Full Screen modes. If all the video options were turned down, the white texture issue didn’t occur, but it remained at the low FPS mark.

KOTOR 2 (knights of the old republic)
Parallels – 5 stars (you’d never know it’s running under an emulator)
Fusion – 4 stars (solid, very playable, but a couple minor issues)

Despite being almost 5 years old, KOTOR 2 requires DirectX 9.0c, and thus, may not have run on the older versions of Fusion/Parallels. The resolution chosen was 1024×768. It does go higher, but doesn’t support any wide-screen resolutions, and if I ran anything higher in windowed mode, I ended up with scroll-bars.

The game play is perfectly smooth on Parallels, and while KOTOR2 doesn’t have a frame-rate meter built in, you’d never know it was running under a virtual machine. I can say it certainly didn’t have any problems – it was screaming fast & smooth. Video options for “anti-aliasing”, “frame buffer effects” and “soft shadows” were greyed out however (other options could be turned on).

On VMWare Fusion, the game is smooth, but not perfect. It’s similar to when in some games VSync being off causes a bit of a jaggy look during movement (although I tested with VSync both on and off). The game was very playable, although it did crash anytime I tried to go into the advanced video options. The screenshot was taken at 800×600, but it remained smooth (but again, not perfect) at 1024×768.

So the verdict? Parallels is the clear winner this time around. It’s just much faster in the games that were tried, and the games that worked were flawless graphically. Perhaps with a few version revisions we’ll see both VM’s pick it up a bit, but for now the choice is pretty clear.

A few things/tidbits worth mentioning:

-Parallels Coherence generally causes major issues with games (don’t use)
-VMWare doesn’t handle switching between full screen and windowed mode very well. Parallels tends to do alright if you start fullscreen and then change to windowed, but not the other way around
-VMWare won’t mount BIN/CUE disk images

There is a fundamental difference in the way they handle resolutions.
-In Windowed mode, VMWare excels because there’s a setting that can locks the cursor to the screen (one of the things the “optimize mouse” setting does) – it helps in games where having the mouse at the edge rotates the view.
-In Windowed mode, Parallels falls behind in some games because the mouse will move off the screen (better Mac integration, but not great with games).
-In FullScreen mode, VMWare excels because if the game resolution isn’t the same as the screen resolution, it will scale the game to fit the screen
-In FullScreen mode, Parallels falls behind because if the game resolution isn’t the same as the screen resolution, the game will be shown in a “box” in the middle of the screen.

Final thing to mention: Don’t believe all the reviews you read – from what I’ve read, Parallels stuffed their Amazon review page at launch – it’s disappointing when a company stoops to those levels. No, they didn’t stuff this one :p

And there you have it! It’s worth noting that similar to the previous comparison, this was done on a “hackintosh” – this time, an i7 920 machine with 12GB of RAM and an ATI 4850 video card, running under Snow Leopard. For comparison purposes, it would be very similar to a current high-end iMac or Mac Pro.

Each VM was given 2 processors, and 4 GB of RAM under Windows 7 Professional 32-bit.

  • Dave

    Hey Matt,

    Fantastic review. We’d like to post part of it on the Parallels blog then link to you for the rest. Interested?

  • Mateo

    It is NOT emulated!!!!!!! There is NO emulation here. Only virtualization.

  • Mateo:

    A touch of background: when explaining something like Parallels or VMWare Fusion to people (often not technically inclined), I’ve found it easier in many cases to use the term “emulator” – many have used NES/Genesis/etc emulators, and if you use the same terms, they tend to get what you’re saying right away.

    Therefore, I tend to use “emulator” fairly loosely – even if it’s not the proper term when referring to a VM, it’s better understood (similar to how some people won’t understand “limited slip” until you refer to it as “Posi-Traction”, or understand that 2WD is technically 1WD). For many people the technicalities don’t matter, and when they do, you can simply go into the detail.

    But just to be clear, even though I’ve used the term emulator a few times throughout the write-up, I’m not in any way trying to imply that Parallels/Fusion are emulators rather than virtual machines (note that I’ve used “VM” most of the time). Whether or not any emulation takes place or not, I’ll admit I don’t know – I would guess that the video end of things is handled by some form of emulator or translator (DX > OGL), but I really don’t know. It doesn’t really matter for this comparison – I’m not looking at how the VM’s go about handling the games, only how well they do it.

  • Peter

    Thank you for the great review!

  • “For comparison purposes, it would be very similar to a current high-end iMac or Mac Pro.”

    Actually no, it wouldn’t. Who are the video drivers you’re using designed by? VMware works with ATI and NVidia when we find bugs, but it’s only designed to work on Apple hardware. There’s no Mac you can configure with a 4850, the Mac Pro comes with the 4870. We have specific checks for this, so it may be playing a factor in your results. I almost have to invalidate them.
    I’m going to test some of this out on my Macbook Pro5,5 with the 9400M and see what what difference I can come up with.

    Regarding the term ‘Emulation’… using the term incorrectly only keeps other people using it incorrectly.

    Otherwise, well presented, I’d like to see more people doing the same. Now, I just have to figure out how to hackintosh my evga x58-based i7-920 😉

    • Michael:
      I appreciate the response. I checked the Apple store before putting up the write-up. The $1999 iMac uses the 4850 (the 4th one shown at http://store.apple.com/us/browse/home/shop_mac/family/imac ). So there are indeed Apple Macs that can be configured with the 4850.

      That said, the “for comparison purposes” was mentioned to give a ball-park area – I wouldn’t want somebody with a regular Macbook, Mac Mini, etc etc to read this and think that they’re going to be running these games on their machines and have a playable experience – they probably won’t.

      The driver used is included with Snow Leopard (was current as of 10.6.1). The only change made was adding the 0x94421002 string to the IOPCIMatch section of Info.plist within the standard ATI4800Controller.kext to get the MacOS to recognize the device ID of the card. If VMWare has specific checks for device ID’s (rather than relying on what Snow Leopard reports), this could certainly be an issue.

      I’d also like to see others doing their own comparisons. The only reason I’ve done them is because there are so few others that seem interested in comparing game performance. I certainly wouldn’t suggest that anybody make their decision based on my experience alone. Maybe my testing was flawed. Maybe nVidia cards perform differently than ATI cards. Maybe the most recent Apple update (10.6.2) has resulted in these tests being obsolete. Maybe I all out lied and skewed the results for a hidden agenda. I certainly haven’t tried to hide anything, but nobody reading this has any way of knowing this for certain. People should always take what they read with a grain of salt, gather as much information as possible from many sources, and if possible download both the Fusion and Parallels demos and try for themselves (heck, they’re what I used!)

      As far as the evga machine, I’d probably suggest the evga snow leopard guide at infinitemac.com (Web Archive) as a starter. The machine I used was based on an MSI X58 Pro-E, and the process could be slightly different.

  • Just to add here: if you get varying results on the Macbook Pro, feel free to pass them along here. I just made a call and I’ll be heading out shortly to give this all a try again on a genuine iMac and see if there are any differences.

    • Alright, just got back.

      Tested on the premise that VMWare Fusion may not have been given a fair test because it wasn’t running on a true Mac (which could cause issues if it didn’t recognize the video hardware amongst other things for example).

      The iMac that I used (7,1) was running the latest version of SL (10.6.2), had a Core2Duo 2.4Ghz processor, 1GB RAM, and an ATI HD2600.

      The Fusion VM was given 1 processor, and 512MB of RAM, again with Windows 7 Pro 32-bit. The RAM limitation of the machine unfortunately meant that I’d only able to re-test 2 of the 4 games originally tested (Cities XL requires 1.5GB, and Mass Effect requires at least 1GB).

      Running Halo CE was again extremely slow to the point of being unplayable. I can’t say it was any worse, but it certainly wasn’t any better either – still hovering around the 1 FPS mark. Again, there were areas with missing textures that show up as white. (screenshot here – nothing new)

      Knights of the Old Republic 2 was a little slower – still playable, but not as good as the first time around. Again, it crashed when attempting to go into the advanced video settings. I took a screenshot of the error message this time around (screenshot here).

      In any case, had the issues in the writeup been a result of VMWare not running on a true Mac, I would have expected at least one of the following when running the games on the iMac:
      -No graphical anomalies (missing textures) in Halo
      -No crash at the Advanced Video Settings in KOTOR2
      -Faster gameplay (despite the slower video card) in both games

      That simply wasn’t the case. I therefore have to conclude that the hackintosh used is reflective, at least to a large degree, as to what a regular Mac with similar specs would see. There really isn’t any evidence to dictate otherwise.

      That said, I will emphasize that the results in the write-up do apply to the specific games mentioned, and that things are obviously subject to change as both Parallels and VMWare release updates as time goes on. There’s certainly a lot of room for improvement in both VM’s as it pertains to gaming.

  • Bernardo

    “-In FullScreen mode, Parallels falls behind because if the game resolution isn’t the same as the screen resolution, the game will be shown in a “box” in the middle of the screen.”

    All you need to do is go to Configure -> Options -> Fullscreen -> and check Adjust Mac in Full Screen…

    This takes out the box and adjust the game to the full screen…