Re: Poor Xorg performance in Etch
On Sun, Aug 19, 2007 at 10:42:22AM -0400, Marty wrote:
> Douglas A. Tutty wrote:
> >On Sun, Aug 19, 2007 at 09:04:13AM -0400, Marty wrote:
> > >
> >>>New observations after further testing: First, I think I was mistaken,
> >>and >this problem does not occur in my other Etch system. Secondly, top
> >>shows >only one CPU pegged at 100%, while the overall CPU utilization is
> >>50%, >suggesting that the other CPU is idle. Taken together these seem
> >>to >suggest a possible SMP or motherboard issue (MSI K9MM-V).
> >>
> >>I posted too soon. Apparently the image setting had not taken hold, and
> >>when I did disable it the problem went away, both in Iceape and
> >>Konqueror. This seems odd because the page is mostly text, with only a
> >>few small images. Now I am guessing there is some kind of image library
> >>issue.
> >
> >Just to clarify what you're seeing in top: 0.0%wa ?
>
> I'm not sure I understand myself. I stopped it while scrolling and copied
> the first few lines below:
>
> Tasks: 126 total, 2 running, 124 sleeping, 0 stopped, 0 zombie
> Cpu(s): 49.9%us, 4.8%sy, 0.0%ni, 45.1%id, 0.2%wa, 0.0%hi, 0.0%si,
> 0.0%st
You have the (1) toggle on so that all CPUs are aggregated into Cpu(s).
Turn off the toggle to get a separate line for each CPU. According to
this your system is idle about half the time. I suspect that one
processor (the one running Xorg) is going flat out and the other is
idle. I don't know if Xorg is multi-threaded and able to run on
multiple CPUs or single threaded where SMP doesn't help. Hopefully, the
application can use the other CPU to spread the load.
> Mem: 2073092k total, 1917620k used, 155472k free, 18832k buffers
> Swap: 6434024k total, 0k used, 6434024k free, 1631596k cached
>
> PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
> 3353 root 15 -10 98128 24m 6460 R 100 1.2 10:36.02 Xorg
>
If you have Irix mode set, then Xorg is using 100% of the CPU on which
it is running. If you have Solaris mode set, then Xorg is using 100% of
all CPUs, for the time period since the last refresh.
Doug.
Reply to: