Qiangning Hong wrote:

Hardware: Thinkpad R52, pentium-m 1.73G,
Host: suspend2-sources-2.6.13-r4, vmware-workstation-4.5.2.8848-r7
Guest OS: Windows 2000 professional

The windows guest os is extemely slow. After double-click the date at
bottom-right corner and show the second hand of the clock, I find that
it cost more than 10 seconds for the guest os to increase *one* second.
That is, the timer in the guest os is 10 times slower than the host.
How can I resolve this?


The problem is probably because VMWare uses the processor TSC counter for timing, the rate of which varies with the processor frequency.

The following works for me on a pentium-m 2.13ghz system:

1. Add "processor.max_cstate=1" to your kernel boot options.
2. If you are running cpufreqd, stop it with "/etc/init.d/cpufreqd stop"
3. Do "echo performance >/sys/devices/system/cpu/cpu0/cpufreq/scaling_governor"
4. Run VMWare.

When you are done with VMWare, you can restart cpufreqd to get the frequency scaling back.

Actually, #2 isn't strictly necessary, but you have to add a rule to /etc/cpufreqd.conf to use the performance governor when VMWare is being run. For example, mine contains the following:

...
[Profile]
name=Performance High
minfreq=100%
maxfreq=100%
policy=performance
[/Profile]
...
##
# VMWare
[Rule]
name=VMWare
programs=vmware-vmx
battery_interval=0-100
profile=Performance High
[/Rule]

The above rule is the very first rule in my configuration. With this, I do not have to do steps 2 or 3 above.

HTH,

-Richard

--
[email protected] mailing list

Reply via email to