VMware Communities
Stefsun
Contributor
Contributor

Fusion pros and cons, and things to be adressed.

To sum up:

Pros:

1. Obviously using less resources, (OSX flies). Parallels can bring OSX to a crawl for seconds at a time.

2. It can suspend (from inside the VM) a Bootcamp VM, using 2% CPU. Parallels Pause function still uses 9-10%.

3. It can hibernate a Bootcamp VM, where as in Parallels every time you want to quit the program you need to shutdown Windows completely. ( I am talking about Bootcamp VMs)

4. Faster CPU, memory and IO, (at least in Bootcamp).

5. Does not corrupt Windows VM. (The last version of Parallels has a lot of problems with crashing Bootcamp for no apparent reason)

6. As a cocoa app, it keeps WindowServer memory to a minimum. (Parallels ups WindowServer memory to 170-300 MBs, cause they say it's a QT app and needs to present 60 open windows to windowserver, even with no VM loaded). I would really like an explanation from the VMware engineers that so kindly help us here, if that is true, that a QT app needs 300 MBs of WindowServer memory and what is the downside for the host system (OSX).

7. It starts and stops much faster than Parallels.

8. Much better USB2 support, with more special devices like WM phones and stuff.

9. Impressed with this forum, where I see Vmware engineers and programmers actively participating and helping, when in Parallels forums plain users try to help or fight each other, with Parallels people absent.

10. Extreme stability, didn't crash once. Parallels crashes at least once a day.

Cons:

1. In my case (Bootcamp-Core2Duo Macbook Pro, 2GB RAM, 10.4.9), 2 CPUs in the VM makes guest XP slower.

2. Slower 2D graphics. (40-50% of Parallels, depending on the action)

3. When in full screen, Command+H is not working. After hiding it through menu and bringing it back, it does not come completely in front unless you click on something on the Mac tray (clock let's say), or right click on it's dock icon. Very annoying. Also annoying is the warning that comes up every time I reopen Fusion and go full screen.

4. At work I use bridged networking. My ethernet is connected to the internal network, and my airport to my own personal ADSL fast line. In OSX I give the gateway to the Airport, and I use Ethernet to mount my servers on the desktop. My Windows VM is configured to use Ethernet so it can join and use the Domain and active directory. With Fusion that can not be the case. The moment I activate Airport with a gateway IP, Fusion decides that that is the primary interface and ignored ethernet. In Parallels I can choose in bridged mode if I want it connected to Airport or Ethernet.

5. At work and connected to the Domain, it never opens it the first time. (??). I double click, it produces an error message, and the second time it connects.

6. Drag and drop between guest and host is not working all the time, sometimes it produces an error message.

7. It doesn't use a seperate hardware profile in Bootcamp VM. Result is that in Device Manager is still a "Multiprocessor ACPI PC".

That's all for now guys, to sum up the goods and bads. I am sticking with you, especially after the rumours of the latest version of Parallels ruining Macbook's batteries, but please do something at least with the 2D speed.

Thanks

Message was edited by:

Stefsun

Reply
0 Kudos
52 Replies
Stefsun
Contributor
Contributor

I would also like to add I don't understand the non-use of VT for 32-bit guests. Judging from Parallels again, which without VT suffers. The speed difference in Fusion it's self, when using Windows x64 as a guest (with VT), versus XP32 (without VT), is very big, when in a real PC it does not exist.

Message was edited by:

Stefsun

Reply
0 Kudos
RDPetruska
Leadership
Leadership

>I don't understand the non-use of VT for 32-bit guests

Have you read the whitepaper at http://www.vmware.com/vmtn/resources/528 ? Basically, all VT does is make it easier[/i] to write a VMM. And VMware has 10+ years of experience using Binary Translation in their VMM - in almost every case it performs better.

Reply
0 Kudos
Stefsun
Contributor
Contributor

I don't disagree with you RDPetrsuska, I am not that technical. All I am saying is that even in Fusion, Windows x64 feels so much faster that XP32, when on a real 64-bit PC the difference is negligable.

When I installed Windows XP x64 in Fusion, I was blown away, it feels twice the speed of my 32-bit Windows XP SP2 VM.

Reply
0 Kudos
aliasme
Enthusiast
Enthusiast

Ack, don't say things like this or I might be tempted to figure out how to do my development on Windows 64. I assume it feels a lot faster, but can you run some quick benchmarks on CPU and disk and see if there's an actual difference.

Reply
0 Kudos
admin
Immortal
Immortal

Did you try installing a "Standard PC" HAL in your Fusion Windows XP VM?

Reply
0 Kudos
Stefsun
Contributor
Contributor

OK, are you sitting down? I have downloaded Passmark 64-bit, and I compared the two VMs, Windows XP SP2 32-bit and Windows XP-x64.

In the 2D graphics department, the 64-bit Windows VM is equal to Parallels performance, it gives a number of 490, when the 32-bit Fusion VM gives a number of 211, and Parallels 512.

In CPU, Windows XP 32 gives 638.5, where Windows x64 gives 770, all in Fusion.

In memory, XP 32 gives a memorymark of 441, where x64 gives 520.

In disk IO, XP 32 gives 152, x64 gives 1061, over 9 times more!!!!!!!!!!!!!!!!

In general, Windows x64 feels like Parallels in 2D speed, smokes everything in everything else.

What a find!

Reply
0 Kudos
aliasme
Enthusiast
Enthusiast

Whaaaaaaat? Thanks for running through these. I'm going to check it out as well.

Reply
0 Kudos
aliasme
Enthusiast
Enthusiast

Here's what I found. Both single CPU, 18GB pre-allocated disks. I'm too busy right now to write up a summary. But in short: 2D performance is night and day. Windows XP 64-bit feels as fast as Parallels and is snappy. Windows XP feels sluggish in the redraw. I didn't notice much in the disk and didn't get a chance to run the 64 test multiple times to see if the seq write was just a glitch.

\----


Windows XP, SP2

\----


Test Name: This Computer

CPU - Integer Math: 116.6

CPU - Floating Point Math: 305.9

CPU - Find Prime Numbers: 449.9

CPU - SSE/3DNow!: 2546.8

CPU - Compression: 2982.6

CPU - Encryption: 18.2

CPU - Image Rotation: 601.8

CPU - String Sorting: 2074.4

Graphics 2D - Lines: 52.1

Graphics 2D - Rectangles: 88.0

Graphics 2D - Shapes: 26.0

Graphics 2D - Fonts and Text: 223.3

Graphics 2D - GUI: 109.5

Memory - Allocate Small Block: 2526.2

Memory - Read Cached: 1963.2

Memory - Read Uncached: 1885.8

Memory - Write: 1661.0

Memory - Large RAM: 64.4

Disk - Sequential Read: 33.5

Disk - Sequential Write: 20.6

Disk - Random Seek + RW: 2.4

CD - Read: 22.1

CPU Mark: 775.1

2D Graphics Mark: 355.5

Memory Mark: 598.9

Disk Mark: 204.1

CD Mark: 2710.9

PassMark Rating: 651.7

System information: This Computer

CPU Manufacturer: GenuineIntel

Number of CPU: 1

CPU Type: Intel(R) Xeon(R) CPU 5150 @ 2.66GHz

CPU Speed: 2644.2 MHz

Cache size: Unknown

O/S: Windows XP (WIN32)

Total RAM: 511.5 MB.

Available RAM: 296.0 MB.

Video settings: 1280x1024x32

Video driver:

DESCRIPTION: VMware SVGA II

MANUFACTURER: VMware SVGA II

BIOS: VMware SVGA II

DATE:

Drive Letter: C

Total Disk Space: 18.0 GBytes

Cluster Size: 4.0 KBytes

File system: NTFS

\----


Windows XP 64

\----


Test Name: This Computer

CPU - Integer Math: 307.7

CPU - Floating Point Math: 414.6

CPU - Find Prime Numbers: 366.5

CPU - SSE: 2545.7

CPU - Compression: 3603.2

CPU - Encryption: 18.9

CPU - Image Rotation: 1287.9

CPU - String Sorting: 2193.1

Graphics 2D - Lines: 116.7

Graphics 2D - Rectangles: 157.8

Graphics 2D - Shapes: 44.1

Graphics 2D - Fonts and Text: 205.5

Graphics 2D - GUI: 274.0

Memory - Allocate Small Block: 2561.8

Memory - Read Cached: 2309.9

Memory - Read Uncached: 2136.6

Memory - Write: 1730.8

Memory - Large RAM: 63.3

Disk - Sequential Read: 32.4

Disk - Sequential Write: 8.3

Disk - Random Seek + RW: 2.1

CD - Read: 21.6

CPU Mark: 930.6

2D Graphics Mark: 559.5

Memory Mark: 650.8

Disk Mark: 154.5

CD Mark: 2649.0

PassMark Rating: 716.1

System information: This Computer

CPU Manufacturer: GenuineIntel

Number of CPU: 1

CPU Type: Intel(R) Xeon(R) CPU 5150 @ 2.66GHz

CPU Speed: 2661.1 MHz

Cache size: Unknown

O/S: Windows XP Professional (WIN64)

Total RAM: 511.4 MB.

Available RAM: 312.8 MB.

Video settings: 1280x1024x32

Video driver:

DESCRIPTION: VMware SVGA II

MANUFACTURER: VMware SVGA II

BIOS: VMware SVGA II

DATE:

Drive Letter: C

Total Disk Space: 18.0 GBytes

Cluster Size: 4.0 KBytes

File system: NTFS

Reply
0 Kudos
Andreas_Masur
Expert
Expert

That's all for now guys, to sum up the goods and

bads. I am sticking with you, especially after the

rumours of the latest version of Parallels ruining

Macbook's batteries, but please do something at least

with the 2D speed.

Eh? How do they manage to ruin the battery?

Ciao, Andreas

Reply
0 Kudos
Pat_Lee
Virtuoso
Virtuoso

Stefsun,

Thanks for your detailed write up of your findings with VMware Fusion.

I especially appreciated the feedback about the team participation in the forums, as I am very fortunate to work with this group of dedicated professionals wanting to ship the best possible product designed from the ground up for the Mac.

Best,

Pat Lee

Senior Product Manager - Mac Products

VMware

Reply
0 Kudos
Stefsun
Contributor
Contributor

To Pat: You are welcome, continue doing a great job. By using a 64-bit Windows client, the speed is phenomenal, no reason to look at Parallels anymore, and I am sure things will improve for 32-bit clients in the versions to come. Like I said, the contrast in these forums compared with the other product is huge.

To Andreas: If you visit www.macwindows.com you can see for yourself, but it seems the last P version 3188 effectively disables Macbooks batteries, sometimes with no fix, not even a PMU reset.

To Aliasme: Thanks man, your findings are almost like mine, amazing how different the two systems feel, isn't it? What is your actual Mac?

Thanks

Message was edited by:

Stefsun

Reply
0 Kudos
DaveP
Commander
Commander

If you want to try VT with 32-bit guests add this to the VMX file:

monitor_control.VT32 = "TRUE"[/code]

Undocumented and unsupported!

Reply
0 Kudos
Stefsun
Contributor
Contributor

And if you try it you end up with a system 50% slower.

Reply
0 Kudos
Andreas_Masur
Expert
Expert

To Andreas: If you visit www.macwindows.com you can

see for yourself, but it seems the last P version

3188 effectively disables Macbooks batteries,

sometimes with no fix, not even a PMU reset.

Wow...glad Fusion is out and I do not need the other one...\*shudder*. Thanks for the information.

Ciao, Andreas

Reply
0 Kudos
DaveP
Commander
Commander

Didn't say it would be faster!! Smiley Wink Did you rerun the benchmarks you had done before?

Reply
0 Kudos
Stefsun
Contributor
Contributor

Didn't have to, it was so slow I had to wait for almost 30 secs just to open the recycle bin.

Reply
0 Kudos
aliasme
Enthusiast
Enthusiast

"Interesting numbers. Can you add more details, like what type of machine you used, what program, settings if non-default, whether you used the hidden vmx config options (and which), how much free space you had on the host (to make sure fragmentation wasn't an issue)?"

Sure.

2 dual core (4CPU) Mac Pro. 2GB RAM.

The opengl multithreading and aiomgr buffer settings are both turned on.

Free space on host is 80GB. The virtual drives exist on a Mac software RAID 0 (striped) pair of 73GB Raptors.

Program is what Stefsun suggested, Passmark 32 and 64-bit versions.

Anything else?

Reply
0 Kudos
Pat_Lee
Virtuoso
Virtuoso

Stefsun,

In reviewing your comments this morning, I didn't quite understand your #5 Con about the domain. Can you provide steps to reproduce this problem or explain it further?

Thanks,

Pat

Reply
0 Kudos
slicedbread
Contributor
Contributor

Just thought I'd like to add some benches.

My specs: Quad core Mac Pro 2.66, 6GB RAM (1.5G in Parallels, 2G in Fusion), VM's running off same HDD, ATI 1900XT.

Both done from same HDD, VMs NOT active at the same time (did one VM, shut down then booted the other VM). Done with Win XP 32bit, SP2 w/ all latest patches etc. Passmark 32bit v6.1

Parallels score (1.5GB ram assigned)

CPU Mark: 716.0

2D Graphics Mark: 751.7

Memory Mark: 602.4

Disk Mark: 1056.3

PassMark Rating: 618.3

Fusion score (1 core, 3D enabled, 2GB ram assigned)

CPU Mark: 784.2

2D Graphics Mark: 308.1

Memory Mark: 604.8

Disk Mark: 258.2

PassMark Rating: 415.4

NB. Definitely not trying to put VMware in a bad light, just highlighting my (and others) experiences of running Fusion Beta 3. Seems there is some quantifiable difference in 2D speed between the 2 programs. I really want to help VMware come out with a better product, your level of support is much better than the other guys. Just hoping there is a similar level of performance in Fusion once it ships.

Reply
0 Kudos