Performance gain through IVT-support under Parallels

[ohrnee]

Hello,

I'm planning to buy a new laptop and have to decide whether to choose one with or without Vanderpool / Intel Virtualization Technology (IVT). The Parallels website and postings in this forum state that there should be a significant performance gain (what is perfectly logical to me).

Interestingly a direct competitor of Parallels (...) claims that performance improvements should be minimal (what is questionable, of course).

Now I'd like to ask, if there is something like an objective comparison of Parallel's performance with and without IVT available somewhere?
I'm asking because I've found the perfect machine for me (unfortunately without IVT) and a comparable product with IVT would cost 500 euros more. As I'm planning to use virtualization excessively (preferredly with Parallels) this is the single crucial issue for me.

Don't get me wrong - I believe there are performance gains and I read the former postings in the forum stating these. But I'd like to see some figures telling me if it's rather about 1% or ... let's say 25% ... ?
I would also appreciate any personal user experiences (!), especially about smoothness.

Thanks in advance,
ohrnee

 

[SteveRichards]

Hello,


I would also appreciate any personal user experiences (!), especially about smoothness.

Thanks in advance,
ohrnee

I noticed a significant performance and stability increase using VT running OS2 as a guest.
I can't speak for other OS's though.
STeve

 

[ohrnee]

I noticed a significant performance and stability increase using VT running OS2 as a guest.

Thanks for your reply, Steve. Interesting that you mention stability as an issue, as all virtualization solutions (or the guest systems inside them, respectively) I used so far seemed perfectly stable.

 

[tacit_one]

...

Hardware virtualization does provide some functionality that could not be achieved in pure software virtualization at the competitive speed.
Even if some competitors like to say that in software it works much better.
Performance improvement will depend on your tasks, but you will surely notice it.

 

[ohrnee]

Hardware virtualization does provide some functionality that could not be achieved in pure software virtualization at the competitive speed.

I understand the underlying basics from my computer science studies and it's to clear to me, that the "binary translation"-approach leads to complex virtualization concepts. But I'm not sure, if (provocating example: ) IVT saves just that a little overhead and makes programming easier - as there are no public benchmark results proving the gains.

Even if some competitors like to say that in software it works much better. Performance improvement will depend on your tasks, but you will surely notice it.

Maybe I'm not enough of a power user, but this is why I ask: I've got a Core Duo desktop system with 2GB RAM, downloaded your trial version of Parallels Workstation and tried Windows and Linux guests with and without enabling IVT-usage. Because the systems perform so well I wasn't able to figure out a clear difference. The laptop system won't achieve that kind of performance, so this is very interesting for me.

That's why I liked to see some figures.

 

[tacit_one]

That's why I liked to see some figures.

The problem is that EULAs of current virtualization solutions do not allow anyone to *publish* benchmarks results. As an employee of Parallels i have no right to do that...
But you can always perform your own benchmarking.

 

[SteveRichards]

Thanks for your reply, Steve. Interesting that you mention stability as an issue, as all virtualization solutions (or the guest systems inside them, respectively) I used so far seemed perfectly stable.

Let me explain the stability...

I had two OS2/ECS systems that I virtualized onto a SuSE 10.2 Linux platform.
(2GB with P4 Hyperthreaded CPU @3.3GHZ)

One ran alone and was perfectly stable. When I would start the other it would run for a couple of hours then one or the other would die. I tried various combos of acceleration and memory config but they just both wouldn't run reliably with both active.

At various times I suspected the network emulation and memory config but the host never exhibited paging or memory related memory issues, and of course the host would never crash.

Enter a new mobo with qc2d (quad core 2 duo @2.4GHZ) and 4gb and VT.
Both virtualized instances now run together without issue, and boot time message fly by with a very noticable speed increase.



Regards,
Steve

 

[ohrnee]

The problem is that EULAs of current virtualization solutions do not allow anyone to *publish* benchmarks results.

You got me - of course I didn't read them. I don't have a background in questions of law but I would have never imagined, that such a claim is possible. Very strange.

As an employee of Parallels i have no right to do that...
But you can always perform your own benchmarking.

Thanks for your honest answer, I'll do that.

Edit: Now I realised this thread is located in the Mac-Section of the Parallels forum, where it shouldn't be. I apologize, if this was my fault.