Fusion 3.0.1 heavily beaten by Parallels 5 in performance

[ehendrix]

ok,

So Macworld did a new performance test between the 3 different virtualization solutions and Parallels did not just beat Fusion in almost every aspect of the performance, it overall did it on such a large scale that it's not even funny anymore. One would almost think that Fusion must still have a bunch of debug code turned on compared to Parallels yet we know this is not the case.

See: http://www.macworld.com/article/145878/2010/01/virtulapptesting.html?lsrc=rss_main

What is actually sad is that VirtualBox seems to be sometimes faster, sometimes on par with Fusion, but VirtualBox is the free option (does not have all the options of Fusion though).

So my question to the VMWare developers, what are the plans to bridge this gap in performance? It really seems to me that Parallels is doing something different compared to Fusion resulting in these enormous performance differences. Will this be addressed so that performance will come more to an equal status? Also interested in why VMWare, such a big player in the virtualization, is beaten so heavily in performance by Parallels? Is there something done differently making the Fusion solution better yet slower? Not from what I can see but maybe the developers have some comments on it?

One of my main reasons to continue to use Fusion is on the way snapshots work, but looking at these performance differences one does have to question if it is worth it.

Thanks.

[GrumpyDave]

I just looked over that test and IMO there's a fatal flaw. I don't understand why in the world he ran the XP test with dual CPU support. Every recommendation I've ever read says you're better off running your XP VM in a uniprocessor configuration. Multiprocessing PCs simply weren't that common when XP was designed and although it does support it, it's just not optimized for it. I'd be willing to bet that the vast majority of XP VMs out there are running single core, at least for those people who listened to the advice and followed it. I'd also bet those benchmarks would be different (better) with a uniprocessor VM. Maybe not as good as the competition, but maybe so. Of course, VMWare still got waxed on the Win7 test, at least for this particular setup.

[ehendrix]

Your comment regarding Windows XP might be the case, but since 2 vCPU was used in both Parallels and Fusion for Windows XP I would think that this would be an issue with Parallels as well then yet it still blows Fusion out of the water.

Also to take note is for example the copying via shared folder and drag&drop performance. I would consider those to be mostly an factor of the efficiency of the virtualization layer and drivers written for this; and we can see how Parallels is much faster.

[Kenneth868]

I hope Fusion developers will really consider this fact.

I owned fusion since v1 and found this v3 really the worst fusion I've ever used.

I've tried it on iMac and Macbook Pro and all shows undesired performance.

Hope that they will stop using the hackintosh excuse for their own product troubles.

[rcardona2k]

Hope that they will stop using the hackintosh excuse for their own product troubles.

I've been avoiding this rant/whine thread but I challenge you to produce one message where a VMware employee user (vm icon on the left) makes this excuse! As a non-VMware volunteer I once asked if it was possible some issues reported here are specific to hackintoshes and not even as an excuse.

[Kenneth868]

Please accept my sincere apologies.

I was naive to assume a highly active, very dedicated and extremely helpful community member was somehow affiliated to the company.

Once again my apologies to a non-VMware volunteer.

[lugesm2]

I have been using Fusion since version 1 and have been really satisfied with it . . . . up until V3.

Yes, I know this song has been sung over and over here, and I am just adding my 2C out of frustration.

I have found v3 to be unusually sluggish and sometimes downright non-functional. Thus far, I have just put up with it in the hopes that the good folks at VMWare would fix the problem with a revised version.

However, after reading the above referenced test against Parallels 5 I have to look seriously at giving Parallels a try.

Sad.

[GlennyG]

I used to use Parallels, but got utterly fed up with them not fixing basic problems and demanding an annual upgrade fee for marginal improvements. One wonders if Parallels learned their business skills (=greed) from Adobe.

Having said that, Parallels definitely has the edge now and it would be hard to recommend Fusion over Parallels except for Fusion's cross-platform working -- take a Fusion VM and run it on a Windows machine; then copy it back.

It's as if VMWare's management don't care any longer. Given that we've all paid VMWare a pile of money JUST to upgrade to Snow Leopard, I would personally like some development time to fix the bugs and tune it. Or maybe VMWare's management have been attending Fagin's school of grabbing money from customers.

[noetus]

For what it's worth, I'm going to add my 2 cents here.

I think it's quite difficult to conclude anything definitive from the test linked to on this thread. A more meaningful test would be for a tester to run and live with Fusion for a couple of weeks with a wide variety of applications representing as wide a swathe of users as possible, and then try the same thing with Parallels for two weeks. Test bench scores are a very narrow sort of evaluation and don't necessarily tell you very much about real-world livability and performance.

I looked at switching to Parallels because I had heard its performance is better and I read around a lot online to try and get a sense of the difference. My conclusion was that Parallels may indeed be faster, but that it is also more unreliable and unstable, and tries to 'wow' its users with somewhat gimmicky features. I decided I prefer a virtualization environment that concentrates on all-round quality and performance, and Fusion strikes me as having the edge here still. I also concluded that Fusion has much better support. This conclusion is not based on experience of Parallels mind you, it is just a conclusion garnered from much reading around of the people's experience of it.

There's also a lot of devil in the details, small ways in which Parallels and Fusion do things differently, a consideration not reflected in a test like this. I have lived with Fusion for a while and gotten used to all its little quirks, and managed to figure out workarounds for most of the things that I have found I don't like about it. Switching to a whole new virtualization environment seemed like just one big headache as I figure out the new features and the new quirks and try to evolve a whole new everyday method of getting along with it.

[ehendrix]

Hey noetus,

One of the items the writer/tester of the article also mentions is that Parallels 5 just overall always felt faster to him compared to Fusion and VirtualBox. So it was not only throughout his testing over several months but also his own impression was that Parallels was faster.

Also, VMWare recommends to store as much as possible in a folder shared with the Mac to keep the VMDKs small and allowing one to execute backups using TM for those items. So "My Documents" is one of those folders then located on this shared. The copying of files etc. can not be ignored for this as it also shows how reading/writing through this shared folder is much slower in Fusion compared to Parallels.

I do agree with you that from overall comments in both Parallels and Fusion forums that Fusion seems to be overall more stable and that patches are more tested compared to Parallels.

My only issue with this is that Fusion is not even close on performance towards Parallels and to me they should be able to do this. Also, Parallels supports the Magic Mouse in the virtual environment in all its glory, yet Fusion can not do this.

To anyone thinking that I started a rant here, no, I just wanted to make absolutely sure that the people working on VMWare Fusion understand they have a lot of work to do to get Fusion up there in performance. And comments in these forums also often mention performance issues with Fusion.

[HPReg]

ehendrix,

So my question to the VMWare developers, what are the plans to bridge this gap in performance?

We as a company have neglected the performance of hosted products (not just Fusion, but also Workstation on Linux/Windows) for too long, in favor of a strong performance effort focused on our server product (VMware ESXi).

This wake-up article from macworld has definitely sparked some internal discussions at VMware. I expect good things to come out of these discussions...

[lugesm2]

ehendrix

This wake-up article from macworld has definitely sparked some internal discussions at VMware. I expect good things to come out of these discussions...

This comment makes me hopeful. As stated earlier, I have been very, very impressed with Fusion, versions 1 & 2.

Guess I'll hang in here a while longer .

[Kenneth868]

This wake-up article from macworld has definitely sparked some internal discussions at VMware. I expect good things to come out of these discussions...

The good news is finally VMware recognize and are aware of the product problem.

The bad news is, at least seems to me, they completely ignore this forum threads regarding the issues.

[lugesm2]

Kenneth,

My impression is that VMWare are aware of the comments on the slug-like performance of Fusion 3.01, especially with Windows 7. But, they don't have any solutions at this time and are tired of responding on this subject.

[Kenneth868]

Kenneth,

.... and are tired of responding on this subject.

</div>

I just hope this is not the case because once been tired this means a miserable client relationship.

My experience tells me that they used to be quite responsive and responsible.

[rcardona2k]

I would say "performance" is a lot like "security" it's a process not a feature. It's easy to be lazy and say Linux should just be fast, or Windows should just be fast out of the box. VMware as a company has gotten themselves in a position where this cannot be fixed overnight, just ask Vista SP0 users about this!

The situation is not hopeless, as users can participate by optimizing their installations, ask any game rigger. In fact, it's a lot like physical exercising and learning about new options and improvements can be additive. What I mean are there limits and boundaries to understand but are also many options to chose, tweak, uninstall, and things to avoid. Some user experiences are counter-intuitive in the virtual world like thinking "2 vCPUs will be faster than 1" without considering your core count, or adding more RAM to a virtual machine will be faster (starving OS X of necessary resources). I suggest reading through a document like as a basic start. Before anyone screams "excuses", these can apply to any virtual system like VirtualBox or Parallels.

Finally, anyone who thinks Win7 is a slug on OS X with an x86 processor, you never used Microsoft VirtualPC to run an emulated x86 environment! I still have a mac mini running OS X PowerPC with an instance of XP and that is the standard by which I compare today's wicked-fast x86 virtualization. 😛

[noetus]

Hey noetus,

One of the items the writer/tester of the article also mentions is that Parallels 5 just overall always felt faster to him compared to Fusion and VirtualBox. So it was not only throughout his testing over several months but also his own impression was that Parallels was faster.

I didn't realize this. Good to know.

Also, VMWare recommends to store as much as possible in a folder shared with the Mac to keep the VMDKs small and allowing one to execute backups using TM for those items. So "My Documents" is one of those folders then located on this shared. The copying of files etc. can not be ignored for this as it also shows how reading/writing through this shared folder is much slower in Fusion compared to Parallels.

I remember choosing Fusion 2.0 on the basis that it was faster than parallels (and I tried both to verify this). Shame the tables have been turned, then!

I do agree with you that from overall comments in both Parallels and Fusion forums that Fusion seems to be overall more stable and that patches are more tested compared to Parallels.

My only issue with this is that Fusion is not even close on performance towards Parallels and to me they should be able to do this. Also, Parallels supports the Magic Mouse in the virtual environment in all its glory, yet Fusion can not do this.

Also it seems that the performance woes are mainly consigned to Vista and Windows 7. I tried several Windows 7 installations with Fusion and was unable to get them to work satisfactorily. I decided to use Win2003 Server instead (tweaked to run as a workstation) and am very happy with that - and for what it's worth, it seems to me to run faster than the same VM in Fusion 2.0. I'm happy with this solution (for me) as there is nothing about Windows 7 in a virtual environment that Windows XP/2003 doesn't have, and the older OSes are lighter on resource usage, so seem better suited to a VM environment.

[Kenneth868]

Amazing ! Why should someone upgrade a product if it does not perform better ?

If performance is not a feature ( or reason ) then way not stick on pervious version of same product or others free product ?

But on one thing I must agree with you. For me, Fusion 3 is VIsta.

[rcardona2k]

Amazing ! Why should someone upgrade a product if it does not perform better ?

In general, they shouldn't. For example if you read WoodyZ's postings here, you see he might rather be on Fusion 1! One big reason people upgrade is for support, i.e. latest bug fixes and safety in numbers, and asking for collective help on "what everyone else has". I like the challenge of upgrading and making things work. Very rarely do I back down, personally I'm very happy with Fusion 3.0.1 on my 4 GB, 13" Unibody MBP running primarily Win7 x64, XP SP3 and Ubuntu 9.10. I upgraded to Snow Leopard day one and broke my AT&T 3G modem but I could live with that and things got fixed. I'm also very patient.

If performance is not a feature (or reason) then way not stick on pervious version of same product or others free product ?

That's the wonderful thing about choices! I used to have Parallels installed but I dumped it back at 3. Now I have VirualBox and Fusion but I use Fusion 3, over 99% of the time. Mostly I was curious about VirtualBox.

Performance means so many things to people, i.e. network performance, shared files performance, graphics performance, Unity performamce, OS boot time, application load time, run my X batch processing. It's very, very subjective. I have several VMs to do my specialized tasks (sw dev) and each one works very well for what I built it for. But I invest a fair of time optimizing and tweaking like I wrote above.

But on one thing I must agree with you. For me, Fusion 3 is VIsta.

I think that's one point of this thread plus to beat VMware with a stick about Parallels 5.

[Kenneth868]

Congratulation ! I think you are one of the few lucky guy's that don't bother

paying for an upgrade and cannot get a better product but live happily with Fusion 3.

Let me guest, after this post many will post they LOVE Fusion 3.