Parallels 9 vs Fusion 6 on nMP

[Merlin65]

In case anyone is interested in a quick speed comparison between Parallels and Fusion on the nMP. Here goes:-

I have mainly used VM Fusion for running Windows/SQL Server and an in-house data intensive application that we developed. Parts of the application processes large amounts of SQL data which pushes cpu, memory and disk IO. I have tested this on both Parallels and Fusion. Sadly, as a Fusion user I have found Parallels runs 25% faster.

Both virtual PCs are identical with the same data.

I wasn't expecting to see such a difference when running a real world application. The benchmarks I have seen on the web have measured Windows startup and shutdown times and they are both fairly similar.

Just thought it worth reporting.

 

[Bear]

Have you checked that both Fusion and Parallels had the same amount of ram, vram and cores allocated to the VMs?

 

[Merlin65]

Have you checked that both Fusion and Parallels had the same amount of ram, vram and cores allocated to the VMs?

Absolutely. The parallels VM was converted from the Fusion VM. They have identical hardware spec and are running the same Windows image and application data. I have even tried increasing the cores and memory on the Fusion VM to see if it can beat the lower spec Parallels VM and it can't.

 

[lemonade-maker]

What about virtualbox/vagrant?

 

[NewbieCanada]

I run Windows 7 under Parallels.

My Windows Experience Index is attached. Geekbench 3 scores are 3026 single and 14,888 multi-core

That comes between a Intel Core i7-3630QM 2400 MHz (4 cores) and a Intel Core i7-4700HQ 2400 MHz (4 cores) on the single core tests.

On the multi, to a Intel Core i7-3770K 3500 MHz (4 cores) or Intel Core i7-2700K 3500 MHz (4 cores)

Pretty respectable, all in all

 

[Merlin65]

For comparison I have run the same virtual machines in Parallels 9 and Fusion 6 on high spec 2012 retina Mac Book Pro and Parallels is 14% faster than Fusion. Interestingly when the precise same test is run on the nMP the gap increases to 25%.

At all times the tests have been run on exact VM clones running Windows 8.1 with 2 cores and 4GB of memory. Parallels appears to be able to make better use of the nMP than Fusion. Very disappointed as a Fusion user.

I am not using benchmark tools to rate the processor, etc. I am using a real world SQL application with real data and a lot of data processing.

 

[AidenShaw]

For comparison I have run the same virtual machines in Parallels 9 and Fusion 6 on high spec 2012 retina Mac Book Pro and Parallels is 14% faster than Fusion. Interestingly when the precise same test is run on the nMP the gap increases to 25%.

At all times the tests have been run on exact VM clones running Windows 8.1 with 2 cores and 4GB of memory. Parallels appears to be able to make better use of the nMP than Fusion. Very disappointed as a Fusion user.

I am not using benchmark tools to rate the processor, etc. I am using a real world SQL application with real data and a lot of data processing.

What kind of virtual disks are you using? Fat-provisioned disks could be much faster for a disk intensive workload.

 

[Merlin65]

What kind of virtual disks are you using? Fat-provisioned disks could be much faster for a disk intensive workload.

I am using thin disks running off the internal SSDs. I have tried to avoid fat disks to save the SSD. I will do a test on fat though to see how it comes out.

 

[bist]

For comparison I have run the same virtual machines in Parallels 9 and Fusion 6 on high spec 2012 retina Mac Book Pro and Parallels is 14% faster than Fusion. Interestingly when the precise same test is run on the nMP the gap increases to 25%.

At all times the tests have been run on exact VM clones running Windows 8.1 with 2 cores and 4GB of memory. Parallels appears to be able to make better use of the nMP than Fusion. Very disappointed as a Fusion user.

I am not using benchmark tools to rate the processor, etc. I am using a real world SQL application with real data and a lot of data processing.

Although it's not specific to the new MP, I can confirm that Parallels is faster than Fusion (On a 2010 Mac Pro).

But (There is a but), for an Hyper-v enabled Windows 2012, Fusion worked (nearly) out of the box. Parallels didn't work at all.
The guest is a W2012 server hosting some network services and three Hyper-V VMs.

So, depending on your needs, Fusion may still have an edge over Parallels.

 

[AidenShaw]

I am using thin disks running off the internal SSDs. I have tried to avoid fat disks to save the SSD. I will do a test on fat though to see how it comes out.

With fat disks, be sure to defragment both the .VMDK file on the host and the virtual disk from the guest to make the comparison more consisten. (Never defragment a thin disk unless you are getting ready to clone or compress it.)

 

[deconstruct60]

...I have even tried increasing the cores and memory on the Fusion VM to see if it can beat the lower spec Parallels VM and it can't.

There is a layer here that consists of the Parallels and Fusion apps. Cranking up the Fusion VM resource consumption if the Fusion kernel is under resourced won't help.

Similarly, if the virtual network is resourced differently then cranking up the vitalized RAM/CPU will have little impact. ' SQL Server and an app' is suggestive of a client server contribution.

 

[Demigod Mac]

Confirms what my experience has been. Parallels is faster than Fusion in most respects, but has all sorts of bugs and issues (some which have sustained across multiple upgrades). Ultimately I went with Fusion because stability, consistency and dependability is more important than performance to me.

 

[AidenShaw]

I am using thin disks running off the internal SSDs. I have tried to avoid fat disks to save the SSD. I will do a test on fat though to see how it comes out.

Also, I assume that the latest VMtools and the similar package for Parallels is installed.

 

[Merlin65]

Also, I assume that the latest VMtools and the similar package for Parallels is installed.

Yes both are the latest versions and the tools are also up to date.

To clarify other posts, the application being tested is a SQL client C# .NET application and SQL Server 2012 is installed on the same VM machine. Everything is self-contained within the VM.

 

[crjackson2134]

Confirms what my experience has been. Parallels is faster than Fusion in most respects, but has all sorts of bugs and issues (some which have sustained across multiple upgrades). Ultimately I went with Fusion because stability, consistency and dependability is more important than performance to me.

This was my experience too. Parallels felt a little more responsive but for me it was nothing to write home about. However, VMWare is much more stable (for me), and this outweighs the increase in speed.

 

[AidenShaw]

This was my experience too. Parallels felt a little more responsive but for me it was nothing to write home about. However, VMWare is much more stable (for me), and this outweighs the increase in speed.

+ one.

Parallels has always been a bit flaky for me - yet VMware has been very solid and I have no problems moving VMs between my laptop and Xeon workstation and the vSphere (ESXi) servers in the data centre.

IMO, Parallels is amateur hobby software and VMware is professional enterprise level software.

And if you knew the price for my vSphere license for eight 16-core servers you'd know that I'm putting my money where my mouth is.

 

[Merlin65]

+ one.

Parallels has always been a bit flaky for me - yet VMware has been very solid and I have no problems moving VMs between my laptop and Xeon workstation and the vSphere (ESXi) servers in the data centre.

IMO, Parallels is amateur hobby software and VMware is professional enterprise level software.

And if you knew the price for my vSphere license for eight 16-core servers you'd know that I'm putting my money where my mouth is.

I think I have read enough to stop looking at Parallels. I am very disappointed with Fusion performance on the nMP given that Parallels is a lot faster for the same tasks but I don't need something which is unstable.

I will hope that VMware take note of Parallels improved performance and do something in the next Fusion release.

 

[A8NSLIDELUXE]

Is there a difference in performance using bootcamp partition or a "really" VM with Parallels/Fusion?

 

[Merlin65]

Is there a difference in performance using bootcamp partition or a "really" VM with Parallels/Fusion?

I would imagine a bootcamp partition will always be faster than any virtual machine. I doubt there is any setup when the bootcamp partition would be slower.

 

[A8NSLIDELUXE]

I would imagine a bootcamp partition will always be faster than any virtual machine. I doubt there is any setup when the bootcamp partition would be slower.

Oh, sorry, I meant using the bootcamp partition in Fusion/Parallels, not native bootcamp of course.

 

[ianj1972]

Some good info here guys, thx. I don't game under my VMs so rendering performance is less important but I do do a lot of disk intensive processing of audio and video files due to some great windows tools which haven't been ported to mac yet and familiar workflows I haven't adapted yet.
My Windows8.1 VM permanently resides in RAM and is active and I haven't seen any instability under Parallels 9 at all. Ive been using it since day 1 (as well as P8 and P7 before those). When you guys complain of stability, how does it manifest itself, applications in the VM crash or the VM itself goes down?
Genuinely interested because I haven't observed any problems at all but considered jumping onto VMWare to take a look at the Yosemite public beta which Parallels doesn't run currently.