CPU Testing with VMWare posts a low score

[millsc]

I am running a passmark cpu test on a ProLiant DL385 AMD Dual Core/Dual Processor Opteron Processor 280 2.4 Ghz, 16GB of RAM. It scores significantly lower than a similar workstation that has an Opteron 275 2.2 Ghz 4MB of RAM. Is there an issue with running Passmark on a VMware machine?

Thanks

[David (PassMark)]

From memory I think VMWare doesn't allow access to the real hardware of the machine but instead uses dummy emulated devices. So you lose the hardware acceleration offered by hardware like video cards. (emulation in software is always slower tham real hardware).

Eveyone has been talking about virtual machines in the last few months, but I really don't understand why they are attractive. For me it sounds like a good way to make your machine slow and loose compatibility with some devices. (For example, last time I checked VMWare didn't support USB2 devices).

Why would you want to use VMware?

---
David

[millsc]

We are looking at VMware as a Workstation replacement. Our workstations cost close to $6K, but looks like it's still more cost effective to go with separate workstations.

Where VMware really makes more sense is for servers. If you look at the stats most companies servers only utilize less than 10% of their CPU. VMware allows you to create a "farm" that consists of a number of actual servers that house a greater number of virtual servers. VMware has a product call VMotion that allows us to assign priority to servers, so high priority servers can be on an actual server that has a greater percentage of cpu and memory available. If the available cpu and memory goes down the server can be automatically migrated to another server without any downtime.

[VMdude]

Re: CPU Testing with VMWare posts a low score

I am running a passmark cpu test on a ProLiant DL385 AMD Dual Core/Dual Processor Opteron Processor 280 2.4 Ghz, 16GB of RAM. It scores significantly lower than a similar workstation that has an Opteron 275 2.2 Ghz 4MB of RAM. Is there an issue with running Passmark on a VMware machine?

I just finished running some tests on a system with VM Workstation installed. My results are pretty much exactly what I expected, the VM runs at about 95% performance, slightly slower due to Windows overhead. My overall results:

Host Machine: 458.7
Guest VM: 438.5

It is hard to say why you are getting the odd results, maybe your machine with the VM is running under load, or there may be some reporting differences between the CPU configuration. You are kind of comparing apples and oranges... I am very curious what type of scores you get when you run the tests on the same machine (Host and Guest VM) instead of comparing between two unrelated machine?

Also you should run the CPU test as a local connection, the network connection to VM desktop can produce some overhead as well.

[VMdude]

From memory I think VMWare doesn't allow access to the real hardware of the machine but instead uses dummy emulated devices. So you lose the hardware acceleration offered by hardware like video cards. (emulation in software is always slower tham real hardware).

Yes, some devices like Video cards are emulated, however, many of the key low level devices, such as the CPU, are accessed directly. VMs are primarily used in Server/Data Center environments, and aren't nearly as useful for the average desktop user, most people can live with a single OS. However, they are excellent for stuff like cross-platform development, and QA. All depends on what your needs are...

Anyways, was digging around and found this PDF with some details on benchmarking VMs that may be useful. Enjoy!

http://www.vmware.com/pdf/WS55_Bench...Guidelines.pdf

[David (PassMark)]

At one point we started to use VMWare for software compatibility testing. But in the end it didn't really work very well for us. The emulated devices don't emulate the real hardware very well.

So we would have software that would work on real hardware but not in VMWare and vice versa. But to be fair most of our problems were related to video cards and 3D code, DirectX, USB device drivers, and checking our benchmarks.

It was bad enough however that we couldn't trust VMWare for checking compatibility with various O/S versions. Now we just use multiple hard drives with various different O/S installed.

-----
David