Tyan 2720 motherboard for Xeons (fwd)
Robert G. Brown
rgb at phy.duke.edu
Mon Jun 24 10:29:45 PDT 2002
On Mon, 24 Jun 2002, Ivan Oleynik wrote:
> Does anyone agree with Steve's opinion?
I wouldn't have six months ago, based on their market reputation and my
experiences with single CPU boards, but I do now. At least in the
specific case of the dual AMD, the 2460 was and is semi-broken with all
sorts of oddities (riser in slot 2, certain cards not supported,
nonstandard power requirements, middling degree of amorphous long-term
hardware instability in the form of crashes with no obvious causes in
some nodes, the screwed up sensors support). The 2466 "seems" better,
but I'm still seeing a range of wierdnesses in brand new nodes,
hardware, bios, wherever. Between one thing and another, I'm seeing a
failure rate of nearly 50% out of the box, so to speak. After screwing
around with them for a while (sometimes involving some parts
replacements) I can generally get them going and then they seem stable,
but I've never had this much trouble with any system configuration
before in my fairly broad 18 year Unix experience. At a guess the BIOS
is still semi-broken and the hardware engineering is way too "marginal"
for robust operation.
I suppose Tyan isn't too likely to replace all of our 2460's with the
2466's that repaired this "beta" motherboard. They're of course
ultimately replacing all the 2466 motherboards that are DOA under
warranty, and one day we might even get a BIOS flash that really works
and that stabilizes our few remaining marginal nodes.
Would I buy Tyan again? I honestly don't know. There are features of
the 2466 motherboard that I really like. It has an onboard 3c920 with
PXE, for example, instead of the piece-of-junk RTL's one finds on many
motherboards. I like having a 64/66 PCI bus, of course. I REALLY like
the serial console option.
Still, the board could be FAR better designed for its obvious rackmount
server role. Stupid little details: the CPUs are oriented laterally so
that heatsink fans are perpendicular to airflow and obstruct it. Memory
DIMMS oriented laterally so that they are perpendicular to airflow.
Power supply connectors at the BACK of the motherboard (right for a
tower, wrong for a rackmount system). IDE and floppy connectors on the
"wrong side" of the motherboard for at least the 2U (AIC) cases we are
using, so that one has to stretch a wide, twisted ribbon across to form
yet another obstacle to airflow in the case -- we have adopted round IDE
cables to get "around" this one. Finally, they actually use jumpers to
control things like enabling/disabling the onboard NIC, and don't do
other BIOS things that would certainly be nice, such as providing a
serial console by DEFAULT (so a system could be setup without ever
needing a video card at all, with corresponding ignoring keyboard errors
and with some special keystroke sequence provided to replace
Ctrl-Alt-Del for serial console soft reboots).
Not impressive engineering, in other words. So yes I think I agree with
Steve -- the 246X dual AMD Tigers, at least, were clearly brought to
market hastily, sloppily engineered for its primary market purpose, and
in a somewhat "unfinished" state beyond even that. Not Tyan's brightest
move, given the number of irritated customers they've developed as a
> ---------- Forwarded message ----------
> Date: Mon, 24 Jun 2002 10:03:18 -0400
> From: Steve Gaudet <SGaudet at turbotekcomputer.com>
> To: 'Ivan Oleynik' <oleynik at chuma.cas.usf.edu>
> Subject: RE: Tyan 2720 motherboard for Xeons
> > I have heard from one of beowulf engineers that Tyan 2720
> > motherboard for
> > dual Xeon CPUs has some serious problems including Memory, ECC and
> > Bios. And Super Micro P4DME-M is offered as an alternative.
> > At the same time I have also been told by other people that
> > Tyan 2720UNG
> > is very fast motherboad for clustering solutions.
> > I would appreciate very much to get some information and guidance
> > regarding the best motherboard for dual Xeons (2.2 Ghz).
> Off the record, Tyan has always been junk. They have a tendency to put
> something new out in the market place and have the customers debug it.
> Recent case in point is the firmware issues Hard Data spoke of on the Tiger.
> However, after 6 months in the field its usually ok.
> SuperMicro on the other hand is very good.
> Steve Gaudet
> Linux Solutions Engineer
> | Turbotek Computer Corp. tel:603-666-3062 ext. 21 |
> | 161 Abby Rd fax:603-666-4519 |
> | Manchester, NH 03103 e-mail:sgaudet at turbotekcomputer.com |
> | toll free:800-573-5393 web: http://www.turbotekcomputer.com |
> Beowulf mailing list, Beowulf at beowulf.org
> To change your subscription (digest mode or unsubscribe) visit http://www.beowulf.org/mailman/listinfo/beowulf
Robert G. Brown http://www.phy.duke.edu/~rgb/
Duke University Dept. of Physics, Box 90305
Durham, N.C. 27708-0305
Phone: 1-919-660-2567 Fax: 919-660-2525 email:rgb at phy.duke.edu
More information about the Beowulf