Additional ethernet (e.g. 100BASE-TX) NIC for Multia/UDB?

[MG]

The one PCI slot is vacant in my Multia/UDB (or equivalent LCA4 Alpha
system, e.g. AXPpci166 MT) Alpha system) and I was wondering, which
DEC/Compaq ethernet NIC would be ideal and is known to 'work'? I do
not need SRM-support, as long as it works once booted up into VMS
and/or Tru64 UNIX, that would be more than fine for me. The on-board
DEC "Tulip" works remarkably well (even in half-duplex) and I'm not
exactly 'relying' on this system, so it's not /too/ important.

I naturally won't assume a "DEGXA" would even remotely work, I don't
even think it did in a PWS 500au (from what I recall when I tried it).
But, what about a DE500-series NIC?

- MG

[MG]

Would anyone perhaps be interested in trading a DEC/Compaq BA500-BA
(50-24601-01/151780-B21; 32-bit PCI) for, say, a Broadcom "NetXtreme"
BCM5703 (64-bit PCI) successfully turned into a "DEGXA"? (Extensively
used under VMS and will, very likely, work just as well under Digital/
Tru64 UNIX.) Then I'll take the risk of finding out myself.

(I also have other PCI, PCI-X and the like cards here; HBAs, NICs,
video cards, memory for various types of systems, etc. ...)

- MG

[Hans Bachner]

Well, the problem with the DEGXA will be its physical size, I believe.
Isn't that a 64bit PCI card (though should work in a 32bit PCI slot)? I
guess the card just won't fit into the box. Beyond that, the DEXGA needs
OpenVMS V7.3-1 at least, and the Multia support kits that I know about
end with V7.2.

Having said that, I'd give the DE500 a try; the DE450 is an alternative,
but not 100 mbit capable. I'm curious enough to grab a Multia from the
shelf, blow the dust off and just try it - but I won't have the time for
the next few weeks :-(

Hans.

[Phillip Helbig---undress to reply]

In article <ke9g9...@usenet.bachner.priv.at>, Hans Bachner

<Ha...@Bachner.priv.at> writes:

> Having said that, I'd give the DE500 a try; the DE450 is an alternative,
> but not 100 mbit capable.

Can the DE450 do FD or just HD?

[MG]

On 29-jan-2013 21:51, Hans Bachner wrote:
> Well, the problem with the DEGXA will be its physical size, I believe.

That's correct (and I mentioned that in my post).

> I guess the card just won't fit into the box.

It should be able to fit, judging by the slot connector (seems to be
keyed for 3.3 and 5 V). But, like I said, if a PWS 500au can barely/
not handle one... then I somewhat doubt the Multia/UDB's ability.
(In fact, I may've tried it already. I could try it /again/, though,
just to be sure...)

I just thought, perhaps it's safer to go with something more in its
league and 'contemporary'.

> Beyond that, the DEXGA needs OpenVMS V7.3-1 at least, and the Multia
> support kits that I know about end with V7.2.

I had V7.3 running on it, but nowadays I'm running Tru64 UNIX V5.1B
(r2650, at the moment) on it. Unlike VMS, Tru64 UNIX doesn't seem
to have problems with the system and installs happily on it...

It's too bad that the original DEC employee 'midnight hack' files
aren't to be found.

> Having said that, I'd give the DE500 a try; the DE450 is an alternative,
> but not 100 mbit capable. I'm curious enough to grab a Multia from the
> shelf, blow the dust off and just try it - but I won't have the time for
> the next few weeks :-(

That would be excellent, I wonder what the outcome will be! Say, what
else do you have in your Multia/UDB? (Like in terms of storage, memory
and so forth.)

- MG

[Hans Bachner]

Phillip Helbig---undress to reply <hel...@astro.multiCLOTHESvax.de>
wrote:

Aren't all twisted-pair links FD capable?

Don't have specs for the DE450 at hand... but
<http://www.aei-
it.com/support/download/pfec_download/Windows/WinNT4sp4/OEMSETUP.INF>
offers a "10BaseT Full_Duplex" connection setting.

But you need a manageable switch where you can set the port to 10/FD. An
autonegotiating switch can sense the speed, but not duplex settings, and
will usually fall back to half duplex if negotiation doesn't work out. I
don't believe that autonegotiation properly works on this card.

Hans.

[John E. Malmberg]

On 1/29/2013 8:21 AM, MG wrote:
> The one PCI slot is vacant in my Multia/UDB (or equivalent LCA4 Alpha
> system, e.g. AXPpci166 MT) Alpha system) and I was wondering, which
> DEC/Compaq ethernet NIC would be ideal and is known to 'work'? I do
> not need SRM-support, as long as it works once booted up into VMS

The problem is that the Multia uses some sort of shared Interrupt
signaling and memory is that VMS only added support for that
configuration type in the 8.x stream to support some Itanium model, and
I do not know the details.

I do not know which 8.x release that was put in, and I would doubt that
any of it would work on the Multia even if you find a way to get VMS 8.x
to run on it and that code is present in the Alpha stream.

Regards,
-John
wb8...@qsl.network
Personal Opinion Only

[Hans Bachner]

Answering myself...

Hans Bachner <Ha...@Bachner.priv.at> wrote:

> Phillip Helbig---undress to reply <hel...@astro.multiCLOTHESvax.de>
> wrote:

<snip>

>> Can the DE450 do FD or just HD?
>
> Aren't all twisted-pair links FD capable?
>
> Don't have specs for the DE450 at hand... but
> <http://www.aei-
> it.com/support/download/pfec_download/Windows/WinNT4sp4/OEMSETUP.INF>
> offers a "10BaseT Full_Duplex" connection setting.

<snip>

Just found the DE435 manuals on
<http://deathrow.vistech.net/~cvisors/dec94mds/>
They explicitely mention full duplex capability for the DE434/DE435 so I
think it's fair to assume that the DE450 will do full duplex as well.

Only 10Base2/10Base5 are restricted to half duplex by design.

Hans.

[MG]

On 30-jan-2013 0:21, John E. Malmberg wrote:
> The problem is that the Multia uses some sort of shared Interrupt
> signaling and memory is that VMS only added support for that
> configuration type in the 8.x stream to support some Itanium model,
> and I do not know the details.

Yes, I'm most aware; actually, I think you may've brought that up on
the (or an) Encompasserve mailing list, at the time. (Because I did
not realize that, until you told me, about I64.)

> I do not know which 8.x release that was put in, and I would doubt
> that any of it would work on the Multia even if you find a way to get
> VMS 8.x to run on it and that code is present in the Alpha stream.

Tru64 UNIX, up to V5.1B, amazingly works fine. What I'm trying to
look into, is if I can get a faster NIC to work, even if while not
under VMS; in such a scenario, I could use the Multia/UDB as an X
terminal, more or less (whilst having some Alpha tools of interest
at my disposal).

- MG

[Steven Schweda]

I know nothing, but if you can't easily get a DE500-BA,
and if your SYS$SYSTEM:SYS$CONFIG.DAT mentions any of the
"Intel 8255x LOM (Fast Ethernet)" chips, then I'd look at my
local PC junk store for a card with one of those chips on it.
I think that I used DE500-BA cards in my oldest junk (AlpSta
200 4/233) systems, but, before I switched over to gigabit
cards, I had used Intel-chip cards in my newer junk (XP1000).
One with an empty flash memory chip socket says "COMPAQ
NETELLIGENT" "10/100 TX PCI Intel UTP Controller". One with
a soldered-in flash memory chip says "COMPAQ NC3121". Both
use an SB82558B chip. I also had a two-port (expandable to
four?) COMPAQ card of some kind, with two of those chips on
it. As I recall, Tru64 didn't complain, either.

If anything works in that slot.

[Hans Bachner]

MG <marc...@SPAMxs4all.nl> wrote:

> On 29-jan-2013 21:51, Hans Bachner wrote:

<snip>

>> Beyond that, the DEXGA needs OpenVMS V7.3-1 at least, and the Multia
>> support kits that I know about end with V7.2.
>
> I had V7.3 running on it,

You take me by surprise - do the V7.2 drivers (network, graphics) work
with 7.3 as well?

> but nowadays I'm running Tru64 UNIX V5.1B
> (r2650, at the moment) on it. Unlike VMS, Tru64 UNIX doesn't seem
> to have problems with the system and installs happily on it...
>
> It's too bad that the original DEC employee 'midnight hack' files
> aren't to be found.

Are you talking about the sources?
The binaries are available on the Freeware V5 CD.

>> [...] I'm curious enough to grab a Multia from the

>> shelf, blow the dust off and just try it - but I won't have the time
>> for the next few weeks :-(
>
> That would be excellent, I wonder what the outcome will be! Say, what
> else do you have in your Multia/UDB? (Like in terms of storage,
> memory and so forth.)

The one at home has a 233 MHz CPU, 152 MB of memory (added 128 to the
original 24), an internal 518 MB Toshiba SCSI disk drive with Windows NT
SP6 (several years ago, this was the PDC of my home Windows domain), and
an external disk enclosure with two more disks - one with additional
Windows file, and one with OpenVMS on it.

I don't remember the exact configuration of the other one (the one on
the shelf in the office), but it is a slower model (166 MHz) with a
single internal disk and OpenVMS on it. You can find it in operation
here:

<http://b.it.co.at/cluster/IT-Symp-2006_Cluster.png>

We built this cluster during the 2006 DECUS symposium in Dï¿œsseldorf,
Germany. Half a dozen folks brought hardware and we just connected it to
a three architecture, mixed physical/virtual, grossly mixed version,
highly unsupported cluster :-)

Hans.

[MG]

On 30-jan-2013 1:18, Steven Schweda wrote:
> "Intel 8255x LOM (Fast Ethernet)" chips, then I'd look at my
> local PC junk store for a card with one of those chips on it.

That's an excellent suggestion, thanks. I almost forgot about
some Intel NICs being a possibility as well.

> One with an empty flash memory chip socket says "COMPAQ
> NETELLIGENT" "10/100 TX PCI Intel UTP Controller". One with
> a soldered-in flash memory chip says "COMPAQ NC3121".

In my DS15, I have a so-called Intel 82559ER (in Tru64 UNIX,
identified as:

"ee0: COMPAQ Intel 82559ER (10/100 Mbps) Ethernet Interface"

I guess that will also be a possibility. But, I must say, the
dual-port one you mentioned looks quite interesting.

- MG

[MG]

On 30-jan-2013 1:44, Hans Bachner wrote:
> You take me by surprise - do the V7.2 drivers (network, graphics) work
> with 7.3 as well?

Yes, they do. You should probably contact Eberhard Heuser, he was kind
enough to share an image of his customized V7.3 installation CD.

I used that CD to install VMS on the tiny, stock, Toshiba 2½" Fast-
SCSI(?) 'notebook'-type disk at first and when it unfortunately died,
later on on the ~72 Gbyte IBM 2½" 'SFF' Ultra320 SCSI disk that I
still have in it, with my own 'custom' cabling. (Works fine, although,
running Tru64 UNIX now...)

> Are you talking about the sources?
> The binaries are available on the Freeware V5 CD.

I only recall the boot floppy images, odd! I'll soon take a look at
it then. (Maybe you should also inform John Malmberg?)

> The one at home has a 233 MHz CPU, 152 MB of memory (added 128 to the
> original 24), an internal 518 MB Toshiba SCSI disk drive with Windows NT
> SP6 (several years ago, this was the PDC of my home Windows domain), and
> an external disk enclosure with two more disks - one with additional
> Windows file, and one with OpenVMS on it.

That sounds like a very nice system! I've never felt what the 233 MHz
one 'feels' like, is it (very) noticeably faster?


> <http://b.it.co.at/cluster/IT-Symp-2006_Cluster.png>
>
> We built this cluster during the 2006 DECUS symposium in Düsseldorf,

> Germany. Half a dozen folks brought hardware and we just connected it to
> a three architecture, mixed physical/virtual, grossly mixed version,
> highly unsupported cluster :-)

That is an amazing cluster, nice to see you took a screenshot of that
terminal session! I don't recall if I ever attempted to cluster my
Multia/UDB with the rest, back when I was still running a cluster.

- MG

[MG]

On 30-jan-2013 2:31, MG wrote:
> I must say, the dual-port [Compaq NC3131] mentioned looks quite
> interesting.

Note to self: Scratch that, it's 64-bit; that's in itself not a
problem, but the card is too long to physically fit in the system,
I fear.

- MG

[Phillip Helbig---undress to reply]

In article <ke9no...@usenet.bachner.priv.at>, Hans Bachner

<Ha...@Bachner.priv.at> writes:

> But you need a manageable switch where you can set the port to 10/FD. An
> autonegotiating switch can sense the speed, but not duplex settings, and
> will usually fall back to half duplex if negotiation doesn't work out. I
> don't believe that autonegotiation properly works on this card.

OK. But, presumably, one could later set it in LANCP.

[Hans Vlems]

On 29 jan, 21:51, Hans Bachner <H...@Bachner.priv.at> wrote:

The Multia 7.2 driver kit works on VMS v7.3 too.
The Multia runs under V7.3 for at least one hour before something
nasty happens that crashes the system.
I do not know whether that is a heat problem or VMS instability on an
unsupported platform.
Since I have enough other Alpha's the Multia is not often used.
Hans

[Hans Vlems]

On 30 jan, 01:44, Hans Bachner <H...@Bachner.priv.at> wrote:

> We built this cluster during the 2006 DECUS symposium in Düsseldorf,

> Germany. Half a dozen folks brought hardware and we just connected it to
> a three architecture, mixed physical/virtual, grossly mixed version,
> highly unsupported cluster :-)
>
> Hans.

When I got my Multia I did not have an AXP/VMS V7.2 CD so I tried a
V7.3 installation cd in stead.
And that worked: the Multia boots and runs for at least an hour. It is
the only system I own that is
not running VMS reliably: it crashes eventually after about an hour,
sometimes it lasts longer.
The system came with 40 MB memory and that was replaced with 256 MB.
High memory boards, the
lid just could be closed so now the system runs with 128 MB. Why the
system crashes I do not know.
Could be a heat issue or a VMS support issue. Using the graphics
console seems to reduce the up time
a lot more than a network login (CTERM, LAT or telnet).

VMS was installed with VMS on CDrom and the Multia drivers on a
floppy. The target disk was in a BA35x
cabinet. After the installation was complete the system disk was
transferred to the internal 2.5" hard disk.
It is not difficult to make an ISO file (LD file) from that V7.3
system disk.
Hans

PS
HECnet users: the image will be available on TITAAN:: next Saturday.

[Dennis Grevenstein]

Steven Schweda <sms.an...@gmail.com> wrote:

> One with an empty flash memory chip socket says "COMPAQ
> NETELLIGENT" "10/100 TX PCI Intel UTP Controller". One with
> a soldered-in flash memory chip says "COMPAQ NC3121".

I think this card was also sold as a DE600, the official replacement for
the DE500. I remember getting one of ebay for a € and it just worked. It is
based on an Intel etherpro chip I think. It did not require any firmware
hacks.

Dennis

--
under construction