Page 1 of 1
Client boot makes host unresponsive
Posted: 16. Nov 2013, 13:10
by SnowmanDK
I have a Windows 7 Ultimate host machine.
It has a AMD FX-8150 octacore cpu, 8gb DDR3 ram, 3TB SATA2 HDD's + 1 128GB SSD.
My vm is set up to use 3gb ram and 4 cores, running a Windows 7 Ultimate.
When I start the vm client normally then it stops during the windows logo. Animation is still running though.
When it stops there even my host becomes unresponsive. Joblist shows no programs doing anything.
If I start in failsafe mode in the client, it stops loading, but does NOT lock up.
If I use a snapshot then I can start it up.
So at the moment I have to start using a snapshot, do what I need to do and make a NEW snapshot to start from.
This also means I can NOT install the windows updates on my vm client.
Logfile attached.
Re: Client boot makes host unresponsive
Posted: 16. Nov 2013, 15:30
by mpack
Lets see... with 4GB memory available, you assigned 3GB to the VM (leaving 1GB for the host). With 4 cores available you assigned all four to the VM, leaving none for the host.
Neither of those things is recommended. I would change the VM to 2GB, and 2 cores at most (I would try 1 first, since unnecessary cores will make the VM slower, not faster).
Re: Client boot makes host unresponsive
Posted: 16. Nov 2013, 17:26
by SnowmanDK
mpack wrote:Lets see... with 4GB memory available, you assigned 3GB to the VM (leaving 1GB for the host). With 4 cores available you assigned all four to the VM, leaving none for the host.
Neither of those things is recommended. I would change the VM to 2GB, and 2 cores at most (I would try 1 first, since unnecessary cores will make the VM slower, not faster).
Ehm... Read my initial post again please

Re: Client boot makes host unresponsive
Posted: 17. Nov 2013, 13:51
by mpack
I read the log file you posted, that's what I base my comments on. If you'll be specific about your objection to my advice, I'll reconsider it.
Incidentally, there's no need to quote other participants message in full: that's a mailing list habit that isn't necessary in a phpBB forum. My message is right before yours if I need to refer to it. Quote small sections when it clarifies which parts you're responding to.
Re: Client boot makes host unresponsive
Posted: 17. Nov 2013, 15:42
by SnowmanDK
Sorry about quoting all. Replied using my tablet and it's browser sux tbh.
That aside, I guess there was some misunderstanding. If I caused that then I am sorry, as it was unintended.
My host pc is an AMD FX-8150 octacore cpu, 8gb DDR3 ram, 3TB SATA2 HDD's + 1 128GB SSD.
With the vm setup I use I should still have 4 cores and 5gb ram left for the host.
As mentioned then it runs fine when I restore from a snapshot, but makes both vm client and host unsresponsive if I try to boot normally in vm.
It started with VBox 3.0 and is still a problem in 3.2.
I tried to reinstall 2.8 but it's incompatible with the host setup changes I made after I got 3.0+.
As I am pretty new in using VBox then I am at a loss, when it comes to determining what goes wrong.
Re: Client boot makes host unresponsive
Posted: 17. Nov 2013, 19:51
by mpack
Hmm. The log file says that 4 physical cores are present. I take it you're sure about the CPU? It's a CPU I'm not familiar with, are these all "real" cores, not some AMD equivalent of hyperthreads?
Yes, the log file agrees that you have 8GB installed, however only 4.3GB of that is available, so if you allocate 3GB to the VM then were getting pretty close to the point where system congestion might be expected. Host and guest are sharing the physical hardware - if you allocate too much to one you risk starving the other. You should aim for an even split, or even better just give the VM no more than it needs.
Re: Client boot makes host unresponsive
Posted: 17. Nov 2013, 23:07
by SnowmanDK
I have taken a screenshot where you can see stats for both host and vm client.
Sorry about the language in it, I use danish

- CPU & Ram info
- fullscreen2.jpg (88.46 KiB) Viewed 1826 times
Re: Client boot makes host unresponsive
Posted: 3. Dec 2013, 19:38
by SnowmanDK
Updated to 4.3.4r91027
My problem persist. I can only boot from snapshot.
When I try to boot normally then both VirtualBox and host becomes unresponsive.
After leaving it at the animated Windows boot screen, VirtualBox shuts down after about 14 minutes (according to log)
The problems started after trying to install Internet Explorer 11 on the client.
I tried removing the update again, but it didn't help.
I have attached the newest log:
VBox.log
- (102.07 KiB) Downloaded 11 times
Update: After posting this I tried starting from a snapshot. Program crashed and Windows log shows this:
Code: Select all
Navn på program med fejl: VirtualBox.exe, version: 4.3.4.0, tidsstempel: 0x5298c42d
Navn på modul med fejl: VirtualBox.exe, version: 4.3.4.0, tidsstempel: 0x5298c42d
Undtagelseskode: 0xc0000005
Forskydning med fejl 0x000000000029e481
Proces-id 0x8a8
Programmets starttidspunkt 0x01cef04e9888a874
Programsti: F:\Programmer\VirtualBox\VirtualBox.exe
Modulsti: F:\Programmer\VirtualBox\VirtualBox.exe
Rapport-id: f8f2f65e-5c41-11e3-bdae-08002700ecc7
Re: Client boot makes host unresponsive
Posted: 4. Dec 2013, 14:03
by mpack
My observations are the same as before.
VBox.log wrote:
00:00:03.615716 Logical host processors: 8 present, 8 max, 8 online, online mask: 00000000000000ff
00:00:03.616167 Physical host cores: 4
If you only have 4 physical cores and assign all to the guest, then poor performance is not a surprise.
Re: Client boot makes host unresponsive
Posted: 4. Dec 2013, 16:55
by SnowmanDK
My problem is not that it's performance is poor, not when running from snapshot.
When running from a snapshot, both guest and host runs smooth.
The host even have leftover power for a game that uses close to 2gb of memory, and all is STILL smooth.
The problem ONLY occurs when I try to boot the guest normally.
It makes everything unresponsive, until the guest finally just closes down.
I must admit I suspect that VirtualBox misidentify my cpu.
This is AMD's own description of the cpu:
http://shop.amd.com/us/All/Detail/Proce ... OX#Details
I also found this I quote from Tomshardwareguide:
The 8150 does have 8 actual, physical cores. These are arranged in 2 core "modules" that each share a single L3 cache and a single floating point processor between the 2 cores. Each core has it's own L1 and L2 cache, and it's own integer processor.
This makes me believe that VirtualBox identifies using the wrong parameter.
Below is a grab using the latest CPU-Z. Please note the lower right (8 cores, 8 threads).

- CPU-Z.JPG (58.25 KiB) Viewed 1770 times
I don't know how else I can prove that it is an 8 core cpu (8 PHYSICAL cores). As said, I think VirtualBox identifies it wrong.
Re: Client boot makes host unresponsive
Posted: 5. Dec 2013, 13:08
by mpack
Then all you can do is raise a
BugTracker ticket for the issue, linking to this topic. BugTracker is a separate system, although you can use the same OSSO login you use here. On first login you'll be asked to choose a nickname for that site. It would make sense to use the same nickname as here, if it lets you.
Re: Client boot makes host unresponsive
Posted: 5. Dec 2013, 15:28
by michaln
SnowmanDK wrote:This makes me believe that VirtualBox identifies using the wrong parameter.
Below is a grab using the latest CPU-Z. Please note the lower right (8 cores, 8 threads).
VirtualBox uses the information provided by Windows. If you think Windows incorrectly classifies your processor as having 4 physical cores, you'll have to take that up with Microsoft.
I don't know how else I can prove that it is an 8 core cpu (8 PHYSICAL cores). As said, I think VirtualBox identifies it wrong.
Just because AMD calls it "eight-core" doesn't mean everyone will. See this article
http://www.anandtech.com/show/4955/ and especially the diagram. It's obvious that there aren't 8 independent cores but rather 4 sets of 2-cpu modules which share some hardware. There's not as much sharing as with Intel's hyperthreads, but there's a lot more sharing than with the older pre-Bulldozer AMD processors.
The Wikipedia article
http://en.wikipedia.org/wiki/AMD_Bulldozer and the diagams therein also clearly indicate that there's only a single L2 cache per module, and the L1 instruction cache is also shared, although each logical processor has its own L1 data cache. The upshot is that for scheduling purproses, it's much better to treat your processor as if it were hyper-threaded -- if you have 4 threads running simultaneously, it's much better to schedule them on per module rather than having 4 threads on two modules and two unused modules. See the note about Windows performance problems in the Wikipedia article.
Bottom line, we're not going to change it. If Microsoft changes its mind again, which seems extremely unlikely, VirtualBox will change what it's reporting as well.
Re: Client boot makes host unresponsive
Posted: 7. Dec 2013, 00:23
by SnowmanDK
michaln wrote:VirtualBox uses the information provided by Windows...
The entire CPU cores thing was a sidetrack in the first place in my opinion, as my thread was about client and host becoming unresponsive when trying to boot the client normally, but it works fine when starting through a snapshot.
No matter what type of survaillance I use to monitor ram/cpu usage, then there is plenty of power left over when I boot the client. It actually looks like it's just idle, at least until VirtualBox just shuts down.
I will try and do some more tracking using other tools for monitoring, and post here if I find anything new.