[ut2004] speed of loading maps

Douglas Wade Needham cinnion at ka8zrt.com
Sun Feb 27 19:20:34 EST 2005


Quoting augustus at linuxhardware.org (augustus at linuxhardware.org):
> Definately not memory related.  One system has 1GB and the other has 2GB. 
> Kernel 2.6.10.
> 
> This one gets me because only UT-based games act up.  For example, Enemy 
> Territory and Doom 3 run fine.  ArmyOps only has slow load times.  UT2004 
> has slow load times and it glitches out some during the game.
> 
> What I see happening during map loads is that UT or AA complete use up all 
> processor and saturate the system.  Once the game starts things go back to 
> normal for the most part.  I wonder what the UT engine could be doing that 
> other games/programs aren't.  Any ideas Ryan?

This is odd, as I have a system I multi-boot on which I play UT2004
(3323), which exhibits exactly the same behaviour, only under XP (I
also have NetBSD and FC3 installed).  Worse, it is not only during map
loading and startup, but thanks to a second monitor running SpeedFan
or other tools, I can tell it happens during game play or just sitting
there on the menus!!  I too know it is not RAM, as the system has 1GB
of RAM, a AMD Athlon XP 2500 (Barton) and an ATI 9500Pro with recent
drivers.  The one thing I have managed to find out in all my
investigation is that almost 0 paging/swapping, 100% CPU utilization,
and very low disk I/O.

Other items to note:
- I had similiar problems back in December, and reinstalled XP, but
  the problem did not disappear with the re-install.  It did, however,
  go away on its own, and reappeared about a week ago.
- There have been no appreciable changes on this machine, other than
  the fact that those morons in Redmond keep making it necessary to
  apply patches.
- This machine gets very regular full scans for viruses and other
  malware.
- This behaviour is independent upon the game type, server SW version,
  and map/game sizes.

If others have seen similiar behaviour, where the problem magically
appeared and disappeared, I have to wonder about the problem either
being in the code common to both platforms, two separate but
similiarly behaved problems, or a lesser likelyhood of something such
as the master servers blasting us with data causing ucc to get lousy
performance.

A question comes to mind at this point.  Has anyone tried to trace the
syscalls ongoing during the sluggish period??  It would be really nice
to know what it seems to be doing during this period.  I remember
investigating a problem with an app while I was responsible for UN*X
systems at CompuServe.  It turns out that the program in question was
doing 30 or so gettimeofday() calls a second, which was traced back to
a point in the code where they were trying to implement a delay.  It
could be that with a trace, we could see something to help us track
down the problem in a similar way.

BTW.  I have been trying to get the ATI driver to run my display under FC3,
so that I can ditch XP and play on UN*X, but even with this latest
driver, it still does not work, and instead I end up with it acting as
if the system were powered down.  I just have not yet had a chance to
collect the diagnostics and send it off to ATI.  Anybody have any luck
with FC3 and the 9500Pro?  If so, please reply to that in a separate
thread or directly.

- Spacey

-- 
Douglas Wade Needham - KA8ZRT        UN*X Consultant & UW/BSD kernel programmer
Email:  cinnion @ ka8zrt . com       http://cinnion.ka8zrt.com
Disclaimer: My opinions are my own.  Since I don't want them, why
            should my employer, or anybody else for that matter! 



More information about the ut2004 mailing list