On Mon, Jan 10, 2000 at 04:48:08AM +0100, Christian Carstensen wrote:
> i'm using a ahc 7895 onboard on a tyan thunder 100. i've seen my cdrom
> occasionally not coming up until power cycling. is this a known bug with
> that controller?
Mine is an AIC 7860 (i.e. 2940UA). I see my hard disk failing to come up
occasionally, but only after I turn my power on after the system had been
turned completely off (which is not often).
> scsi performance is very bad on my system for a long time now. operating
> on many small files, for example, my IBM DCAS-34330W is 100% busy with
> only 1.2 MB/s.
> i don't know why, but i suspect vinum in combination with some old scsi
> devices to cause that.
What I'm going to say here may be complete crap, but it has been on my mind
for some time now, so I may just as well speak up with my suspicions.
I have read on several occasions that people find -current performing very
poorly lately. I recall someone comparing -current's network performance to
that of 3.4 and finding almost a factor 2 difference.
I myself have found the following facts lately:
1) scsi performance is down to 4 Mb/sec MAX, writing simply zero's (this
used to be around 12 Mb/sec)
2) network performance (NIC = 3COM 3C905B-TX 100Mbit) is 5 Mb/sec MAX
instead of the 8 Mb/sec I'm used to.
3) IDE UDMA66 performance is down to 12 Mb/sec writing zero's (this used to
be near 16 Mb/sec).
4) buildworlds are at an all time slow. I haven't been able to compile a
current in less than an hour for many weeks now. This used to be more
like 40 minutes or so (make buildworld -j 9 on my smp system).
Add to this the strange lockups we've been having recently that somehow
seem i/o or load dependent.
I'm no expert on this at all, but it seems legitimate to conclude that
something is seriously wrong somewhere. It seems to me that this should
_definitely_ get sorted out to the bottom before release time.
Regards,
Dave Boers.
--
[EMAIL PROTECTED]
To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-current" in the body of the message
