virtualbox.org

Discussions about 3.0.6 here please.

People who use the repositories and have the Beta installed, check http://forums.virtualbox.org/viewtopic.php?f=1&t=22210

I just upgraded to 3.06. Linux Kubuntu 9.04 host, AMD 64 dual-core, Windows XP guest.

My CPU is now sitting at 100% with the XP guest at idle. This did not happen with 3.04.

Are you running your fuest virtual machine with 1 or 2 cpus?

As I wrote in another thread (http://forums.virtualbox.org/viewtopic.php?f=1&t=22211), I've had similar problems with dual-cpu guest Windows XP with all 3.0.x releases. And I have an Intel Core2 Duo E6300, so AMD-restrictions mentioned in the user manual should not apply. My symptoms are that guest OS is terribly slow with host CPU utilisations staying at 100 % for most of the time.

I posted a message in the beta discussion topic just before this discussion topic was created, however my comment applies to the release, not just the beta. Rather than double post, here is the link.

bitti wrote:Are you running your fuest virtual machine with 1 or 2 cpus?

As I wrote in another thread (http://forums.virtualbox.org/viewtopic.php?f=1&t=22211), I've had similar problems with dual-cpu guest Windows XP with all 3.0.x releases. And I have an Intel Core2 Duo E6300, so AMD-restrictions mentioned in the user manual should not apply. My symptoms are that guest OS is terribly slow with host CPU utilisations staying at 100 % for most of the time.

The XP guest only is using one CPR, for the reasons you describe. This 100% CPU usage by the XP guest is new to 3.06. It did not happen with 3.04.

@rober1s

I've just checked my CPU, and it's fine. VBoxSvc isn't totally idle, but it doesn't get above 5%, and that is only every few seconds. It's mostly at or near zero when one VM is running (I tend not to run more than one).

mpack wrote:@rober1s

I've just checked my CPU, and it's fine. VBoxSvc isn't totally idle, but it doesn't get above 5%, and that is only every few seconds. It's mostly at or near zero when one VM is running (I tend not to run more than one).

Yes, the behavior you describe is how 3.04 behaved for me. The switch to 3.06 changed it.

BTW, I reran /etc/init.d/vboxdrv setup

and

installed the 3.06 guest additions. No change, still 100% CPU.

I had to go back to 3.04, I couldn't get any work done with 3.06.

Ok, I re-uninstalled 3.0.4 and re-re-installed 3.06, and the performance problem has gone away. Sorry for the false alarm.

I have not yet played with 3.0.6 but I am curious why this was done:

Main: restrict guest RAM size to 1.5 GB (32 bits Windows hosts only)

I believe there are some versions of Windows 32-bit (e.g., Windows 2003) that support the expanded page tables and PAE, etc. and can actually use more than 4GB of memory (if that was the reasoning for limiting it). Of course, no single process can have more than 4GB of address space but I can envision allocating 3GB to a VM without issue (because that does not leave the rest of the system with less than 1GB as one might first think).

Thanks

For some reason NAT doesn't seem to do anything under 3.0.6 (XP64 host, any guest). There are no obvious connection errors, I can ping any IP address I like successfully, but no other traffic -- network mounted drives gone, browsers show blank pages, etc. Weird. Was certainly working under 3.0.4.

Uzume wrote:I have not yet played with 3.0.6 but I am curious why this was done:

Main: restrict guest RAM size to 1.5 GB (32 bits Windows hosts only)

I'm wondering if there's a mistake in the release notes. I have an existing 32-bit Windows XP guest with 3.5 GB of RAM. It works normally in 3.0.6 and XP sees the whole 3.5 GB.

Uzume wrote:I have not yet played with 3.0.6 but I am curious why this was done:

Main: restrict guest RAM size to 1.5 GB (32 bits Windows hosts only)

I may be totally wrong on this, and I've done no research in order to refresh my memory, but...

I have a vague recollection from when I first started developing Win32 apps that every app is given a 4GB virtual process space (virtual here having nothing to do with VBox), of which 2GB was assigned to the application, and the other 2GB contains the OS. In 1996 2GB was an idiotically large allocation for the app, so of course nobody cared.

Now if that's true, and if your VM is living inside the application space of a VBoxSVC instance, and assuming VBoxSVC needs some memory itself, then...

bitti wrote:

Uzume wrote:I have not yet played with 3.0.6 but I am curious why this was done:

Main: restrict guest RAM size to 1.5 GB (32 bits Windows hosts only)

I'm wondering if there's a mistake in the release notes. I have an existing 32-bit Windows XP guest with 3.5 GB of RAM. It works normally in 3.0.6 and XP sees the whole 3.5 GB.

this was explained by frank from sun in one of the 306beta1 discussions, its not a mistake.