thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) (09/23/90)
Well, we got our Mentor upgrades to the HP9000 series 400t. I still want
to call it a DNxyzz name -- guess I prefer Apollos.... :-)
Some initial comments on them --
According to our local office, we got the first 14 Mentor 400t nodes
that they shipped in the world. (This isn't an evaluation of the nodes,
but it's something to be proud of. Don't burst my bubble if they lied.)
They are blazing fast. Our configurations are 16MB, 2x210MB SCSI disk,
with mainly monochrome, but 2 19" color (lowest end graphics speed). Our
initial Dhrystone checks list it at 20833 Dhrystones. By comparison, our
DN10000 is 41666 (yes, exactly 2x!). our DN4500 lists at 15625, our DN3550
gets 7936, our VMS Vax 8800 gets 12500, Sun 3/50 gets 2976, and the
venerable Apple IIe has 39! We don't have many Suns (Scums) in-house, so
don't have numbers for many of those. Please don't flame on the numbers
if you disagree. They are _only_ our benchmark -- not official; they come
under the "lies, damn lies, and benchmarks" quote; and when it comes right
down to it, nobody runs benchmarks as their normal production software.
The $%^#$ color monitors do NOT SWIVEL!!! Apparently, for about $350,
you can buy a swivel (and tilt?) stand for it. The resolution on them is
fantastic compared to the 1024x800 monitors.
Most software seems to run perfectly fine on them. We have encountered
a floating point error (crashed the program) in one of our packages, but
it's too early to say whether that's our problem or theirs.
Setting them up is a breeze. In our case, Mentor pre-loaded the O/S
so it was a definite case of put-together-and-turn-on.
The O/S is a little flakey in the 10.3 pre-release that we have. This
is supposedly post-beta, but pre release (lists as 10.03.a. The tcpd "can
intermittently disappear under heavy loads, specifically when X windows are
running...." This _will_ be fixed before the actual release. I don't play
around in Unix much any more (I _really_ prefer Aegis), but for the most part,
everything looks ok in it.
As with the 10.2 rgyd mixing with 10.1 rgyds, there's a problem with 10.1
and 10.3 rgyds mixing. No mention of problems with 10.2 and 10.3, and an
implied "no problem", since Mentor states that the 10.3 rgyd will name itself
'rgyd.1.2' (same as 10.2 does).
The nodes themselves don't look like Apollos any more. They look kinda
like HP or Sun systems. <sigh>
The token ring controller is right there like they said it would be, along
with an ethernet controller on the motherboard. The SCSI connector is a
high-density pin arrangement, unlike the Western Digital controller or the
DN2500 connector. There's some redundancy in connectors, with an HP keyboard
connector as well as the Apollo one, and a speaker port that only works with
HPUX. There's also an ethernet tranceiver and thinnet port.
All in all, I like 'em. Here's hoping that the software problem clears up,
that 10.3 comes out in actual release without bugs, and that the 68040
upgrades we ordered comes in real soon.
John Thompson (jt)
Honeywell, SSEC
Plymouth, MN 55441
thompson@pan.ssec.honeywell.com
As ever, my opinions do not necessarily agree with Honeywell's or reality's.
(Honeywell's do not necessarily agree with mine or reality's, either)system@alchemy.chem.utoronto.ca (System Admin (Mike Peterson)) (09/23/90)
In article <9009222009.AA24527@pan.ssec.honeywell.com> thompson@PAN.SSEC.HONEYWELL.COM (John Thompson) writes: > >Well, we got our Mentor upgrades to the HP9000 series 400t. > >Some initial comments on them -- > [ ... many comments deleted ... ] > The O/S is a little flakey in the 10.3 pre-release that we have. This >is supposedly post-beta, but pre release (lists as 10.03.a. The tcpd "can >intermittently disappear under heavy loads, specifically when X windows are >running...." This _will_ be fixed before the actual release. I don't play >around in Unix much any more (I _really_ prefer Aegis), but for the most part, >everything looks ok in it. I certainly hope the perennial TCP problems will be fixed before the actual release, but I won't hold my breath on it - we have been having TCP "disappearing" problems for 2 years (since SR10.0), and they still have not stopped (our DN10000 still hangs once a week (MTBF), even with a special OS patch to kickstart the VME Ethernet controller - in fact that patch which is supposed to go into SR10.3.p makes our system behave worse in terms of some really long delays - up to 5 minutes between characters). Our M68K systems can now last about a month between hangs - we run X on all our systems as the default window manager and login environment. -- Mike Peterson, System Administrator, U/Toronto Department of Chemistry E-mail: system@alchemy.chem.utoronto.ca Tel: (416) 978-7094 Fax: (416) 978-8775