Re: high load with win7 & usb tablet

[Michael Tokarev <mjt@tls.msk.ru>]

10.09.2010 15:47, Jes Sorensen wrote:
> On 09/03/10 14:21, Michael Tokarev wrote:
>> Hello.
>>
>> I noticed that window7, unlike windowsXP as I played with
>> previously, poses quite high load on host when idle.  On
>> my machine, host cpu usage is about 25% when one single-cpu
>> win7 guest is running with "-usbdevice tablet" option.
>> Here's a typical kvm_stats when such a guest is idling:
> 
> Hi Michael,
> 
> I am curious, the XP numbers you are referring to, are they against the
> same version of KVM/QEMU or is it something you saw in the past? It

I ran winXP guest and win7 guest one after another on the
same machine using the same version of kvm (it was 0.12.5).

You run a freshly installed win7 guest and at idle it consumes
about 25% of host CPU.  You shut it down and run a freshly
installed winXP guest - about 3..4% host CPU usage at idle.

(This is AthlonII CPU, 3.2GHz.  Cpufreq is enabled, and that
activity is not enough for ondemand governer to increase the
frequency from the lowest awailable 800MHz, so that's 25% of
800MHz, not 25% of 3.2GHz).

> would be interesting to track down whether this idle time consumption is
> only with win7 or if it happens with all versions of windows, as opposed
> to being a general problem in current KVM/QEMU.

It does not happen on winXP and happens on win7.
I haven't tried vista - don't have it around.

Since my initial report, I learned a bit about USB
(trying to find a way to change pooling frequency).
I don't actually understand how it works, but some
sources claims that USB device descriptor, among
other things, also may include preferred pooling
interval.  It has also been mentioned that win7
has different USB driver for HID devices, and some
sources report that it now checks for mouse moves
1000 times per second, instead of 120 or 250 as
previous versions of windows.  I've no idea how
accurate these sources and claims are, however... :)

Thanks!

/mjt