> [ On Wednesday, August 27, 2003 at 23:04:03 (-0700), Bryan Vyhmeister
wrote: ]
> > Subject: Slow Xeon Performance
> >
> > Has anyone else observed similar issues on a Xeon system? I am
really
> > not sure what to think. It seems very strange.
> 
> Unfortunately I've don't think I've ever done much work on any Intel
> Xeon based system, but from what you've said I can ask some more
> questions that  might help identify the problem:
> 
> How much cache is there on the Xeon processor?  Is it a P-III Xeon
(I'm
> assuming it is though I've not looked at your mother's specs)?

The Xeon is a P4 Xeon. It is a 2.4 GHz CPU with 512 KB of cache and a
533 MHz FSB.

> How much, if any, cache is on the Adaptec 2010S?

48 MB of ECC SDRAM

> How much cache is on the particular drives you are using?

The drives each have 8 MB of buffer cache.

> Did you enable write caching on the disks (dkctl setcache rw)?

No. I tried enabling this and I just get errors about "inappropriate
ioctl for device."

The command I tried using is:

dkctl /dev/rld0d setcache rw

I also tried other variations of ld0, ld0d, rld0, etc.

> Were the drives on the P4 system also set up as RAID-1?  With
RAIDframe?

The drives on the P4 system right now are in a RAID-5 using three
Western Digital 120 GB drives with 8 MB of buffer cache connected to an
Adaptec 2400S IDE RAID card. Before it had a single 120 GB drive and it
seems only marginally faster. Even with the single drive it was far
faster than the Xeon.

> What kind, quantity, and speed of RAM was on the P4 system?

The P4 system has 1 GB of RAM in two 512 MB PC133 SDRAM chips.

> I'm assuming all your sources and objects are on the local disks...

That is correct

> I'm guessing your problem is partly to do with disk throughput, which
may
> be slower for sequential tasks such as a straight compile on the Xeon
> system despite the use of such drives and controller.

I suppose that might be the case. Since the drives are quite a bit
faster in the Xeon system than they are in the P4 system I would think
that the Xeon would at least be about the same speed as the P4 even if
it is running much slower than actual peak performance. The fact that
the P4 is 1.5 GHz and the Xeon is 2.4 GHz really makes me wonder as
well.

> You may also not be making the best use of that amount of RAM with a
> GENERIC kernel, assuming that's what you've been booting on both
> machines.

That very well could be the case. I have not compiled a custom kernel
yet because I just installed NetBSD today on this system. That still
should not account for the major slowdown.

> Are you using a memory filesystem for /tmp?  On both machines?

I am only using an mfs /tmp on the P4 machine, not the Xeon. I am
enabling that on the Xeon machine right now.

> Have you added "COPTS+=-pipe" to /etc/mk.conf?  On both machines?

No. What is the advantage in this?

Thank you for your help.

Bryan