Alpha Server 1200 memory question

[Hans Vlems]

An AlphaServer 1200 5/533 with two cpu's reports no memory errors during the initial hardware tests. It is fitted with 4 GB memory and POST shows:

SROM V3.0 on cpu0
SROM V3.0 on cpu1
XSROM V6.0 on cpu0
XSROM V6.0 on cpu1
BCache testing complete on cpu0
BCache testing complete on cpu1
mem_pair0 - 512 MB
mem_pair1 - 512 MB
mem_pair2 - 512 MB
mem_pair3 - 512 MB
mem_pair4 - 512 MB
mem_pair5 - 512 MB
mem_pair6 - 512 MB
mem_pair7 - 512 MB
20..20..21..21..23..
please wait 96 seconds for T24 to complete
24..24..
Memory testing complete on cpu0
Memory testing complete on cpu1
starting console on CPU 0
sizing memory
0 512 MB DIMM
1 512 MB DIMM
2 512 MB DIMM
3 512 MB DIMM
4 512 MB DIMM
5 512 MB DIMM
6 512 MB DIMM
7 512 MB DIMM
starting console on CPU 1

The 96 seconds reported for T24 is normal.
Now VMS V8.3 displays:

$ sh mem /phys
System Memory Resources on 28-NOV-2015 12:20:30.39

Physical Memory Usage (pages): Total Free In Use Modified
Main Memory (4.00GB) 524288 39508 483596 1184

Bad Pages Total Dynamic I/O Errors Static
454624 0 0 454624

Of the physical pages in use, 9318 pages are permanently allocated to OpenVMS.


The question is: how can I figure out which memory bank(s) are dead?

Hans

[Jan-Erik Soderholm]

Why should anything be "dead" ??

524288 x 8192 /1024 / 1024 /1024 = 4 GB

[Steven Schweda]

> [...] how can I figure out which memory bank(s) are dead?

If there's no good memory diagnostic tool, then I'd
probably start removing modules until the Bad Pages value
falls (ideally to zero).

The SRM console may offer things like memexer or memtest.

http://labs.hoffmanlabs.com/node/1459

[Hans Vlems]

Because VMS finds 464526 bad pages. Which is quite a lot...

[abrsvc]

First I'd check the mem config again. The 1200 only supports 2G not 4.
If there are 512Mb boards in there, then that may be the problem. Remove 4 of the sticks and get back to 2GB. See if that fixes the problem.

Dan

[Steven Schweda]

> [...] The 1200 only supports 2G not 4.

And you think that the worst thing that'll happen when you
install too much memory in a system like this is a few bad
blocks? I know nothing, but...

http://h18002.www1.hp.com/alphaserver/archive/1200/

[...]
Features
o Up to 4GB ECC memory
[...]


According to the manual, an AlphaStation 200 4/233
supported a max of 384MB, but I ran twice that in mine. On
the other hand, so far as I can tell, the 2GB published max
for an XP1000 is the real max.

[Hans Vlems]

Op zondag 29 november 2015 01:58:06 UTC+1 schreef abrsvc:

> First I'd check the mem config again. The 1200 only supports 2G not 4.
> If there are 512Mb boards in there, then that may be the problem. Remove 4 of the sticks and get back to 2GB. See if that fixes the problem.
>
> Dan

Dan,
I'm pretty sure the 1200 supports 4 GB. The User Manual indicates this as well.
The system has been running fine with 16 sticks of 256 MB each for several years now. So I'm pretty sure it is not a configuration issue.

The problem is finding which stick is at error (possibly more).
The SRM command memtest didn't help at all, showed no errors.
VMS just flagged 80% of the memory as bad. I feel it's unlikely that so many
sticks went south all at the same time.

That is why I posted the question: is there a way to pinpoint which memory stick
(s) went bad?

If not then all there is to do is remove all but one pair and repeat the
process until the faulty stick shows up.
What I think, given the large number of bad pages reported by VMS, is that one
of the lower banks is at fault.

The system is open, I removed the lower CPU board, fault finding 101 in progress :-)

Hans

[abrsvc]

My bad... I had an old listing for the model which listed 2GB as max. I have since found the newer specs with the updated 4GB limit.

[Jan-Erik Soderholm]

Den 2015-11-28 kl. 16:47, skrev Hans Vlems:

> Because VMS finds 464526 bad pages. Which is quite a lot...
>

Sorry, I totaly missed the "Bad Pages" line. I'm not sure
I've ever seen that line in SH MEM... :-)

O.T...
Is it just me that never see anything but the actual reply
from Hans Vlems? Or is it you, Hans, that always cut away
the text you are replying to? I would not think so.

Jan-Erik.

[Jan-Erik Soderholm]

Correction. In your reply to "abrsvc", I can actualy see
both the text from abrsvc and your reply. Weird...

Anyway, if you see a thread for the first time with a post that
starts with "Because...", you are lost without the question... :-)

Jan-Erik.

[Hans Vlems]

Op zondag 29 november 2015 13:31:37 UTC+1 schreef Jan-Erik Soderholm:

It depends Jan-Erik on the device where the reply originates.
On my phone (BlackBerry Z10) the application tends to skip the
original text. Which causes confusion at times as you experienced.
On my PC Google Groups includes the previous thread, I have a tendency
to reduce that content!
Hans

[Hans Vlems]

I located the bad memory board the old fashioned way:
put in an additional pair until VMS reported an error.
This became a hard error, VMS didn't even boot any more.
The faulty stick was found and the system now runs with 3.75 GB.
Which is more than sufficient for my needs. I like it though when
a system runs with full memory. Which happened to be very expensive
when this system was new...

Hans

[Michael Moroney]

There was a difference between unsupported because it was never really
tested thoroughly and unsupported because it can't work (address lines or
whatever). I did VMS qualification on the AlphaStation 200 4/233 way back
when. At the time we qualified it to 192MB which was an oh my god amount
of memory, which we figured nobody would actually use because it was worth
so much compared to the rest of the machine. Later on memory got larger
and cheaper and we did qualify it with 384MB. By the time larger memory
came out that line of systems was no longer sold so it never got qualified
with larger amounts.

We did notice with the very large amounts and very long uptimes that
there were occasionally parity errors that would crash the system, due to
the smaller memory cells being more sensitive to cosmic rays or whatever.

[Steven Schweda]

> [...] By the time larger memory came out that line of

> systems was no longer sold so it never got qualified with
> larger amounts.

That's what I always assumed.

> [...] there were occasionally parity errors that would
> crash the system, [...]

I used ECC memory, and never noticed any memory problems,
certainly not a system crash.

[Paul Sture]

It's been a long time since I used Thunderbird but various
news readers have the ability to hide quoted text. I would
check to see if something like this is set on your system.

--
Xscreensaver patch out:
CVE-2015-8025: xscreensaver could be bypassed by disconnecting HDMI cable
<https://cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2015-8025>

[Jan-Erik Soderholm]

Den 2015-11-29 kl. 20:49, skrev Paul Sture:
> On 2015-11-29, Jan-Erik Soderholm <jan-erik....@telia.com> wrote:
>> Den 2015-11-28 kl. 16:47, skrev Hans Vlems:
>>
>>> Because VMS finds 464526 bad pages. Which is quite a lot...
>>>
>>
>> Sorry, I totaly missed the "Bad Pages" line. I'm not sure
>> I've ever seen that line in SH MEM... :-)
>>
>> O.T...
>> Is it just me that never see anything but the actual reply
>> from Hans Vlems? Or is it you, Hans, that always cut away
>> the text you are replying to? I would not think so.
>
> It's been a long time since I used Thunderbird but various
> news readers have the ability to hide quoted text. I would
> check to see if something like this is set on your system.
>

Ah yes.
I have seen that in another cases (see below), but the "issue"
with mr. Vlems posts was sorted out in an earlier post... :-)

In that other case, I can not see posts from *one* specific
poster to the Oracle Rdb mail list. That is "read" using
normal mails using my Office 365 Outlook client. The funny
thing is that whenever someone else replies and quotes
the post from that specific postar, I can see it all. :-)

It s bit of a pity in the above case, since the "specific
poster" in one of the best of the Rdb engineers. :-)

I know from other cases that some mail clients has issues
if the poster starts his signature field with a line with
just three dashes, '---' (or something like that).

Jan-Erik.

[MG]

Op 29-nov-2015 om 10:09 schreef Hans Vlems:

> I'm pretty sure the 1200 supports 4 GB. The User Manual indicates this as well.

Interesting, considering the age of this system and especially
considering the fact that much more recent systems, like the
DS15 I used to own, could maximally take the same amount.

> The problem is finding which stick is at error (possibly more).
> The SRM command memtest didn't help at all, showed no errors.
> VMS just flagged 80% of the memory as bad. I feel it's unlikely that so many
> sticks went south all at the same time.

Does it have ARC/AlphaBIOS? You might be able to run Windows
NT 4.0 or 2000 for AXP and try your luck with a diagnostics/
testing tool that can also exercise and check memory? Just
an idea...

Good luck either way with looking into this. Little is
arguably as annoying (let alone time consuming) as having
to test individual memory SIMMs/DIMMs...

- MG

[Hans Vlems]

Marco, actually the system is the white box equivalent of the AS1200. It underwent brain surgery to run NT only when it was young. Luckily it found a stable home and it happily boots VMS. Since 8.4 it doesn't even need special incantations in nvram to make that happen. HP even corrected the time problem so once it runs 8.4 there's no difference anymore between a DigitalServer 5305 and an AlphaServer 1200. It took 17 years to get there :)

Fault finding in memory boards, especially if there are as many (16) on a 1200, is tiresome. The srm command test cpu1 or test cpu0 did help. It flags the banknumber that has the bad stick.
Then it needs another swap to figure whether the bad stick is on the high or low extender board.

I was very glad to get an early Sinterklaas present last Wednesday : 10 spare memory boards! So the 5305 now runs again with 4GB memory, as much as the DS20.
Only my rx2600 has more available, 20GB.

The other two 5305's have AS1200 cpu's fitted and present themselves as such to the OS. The fourth stays with Phillip in Germany, unless he has tossed it.

Hans

[Phillip Helbig (undress to reply)]

In article <5660dfb5$0$18514$e4fe...@newszilla.xs4all.nl>, MG

<marc...@SPAMxs4all.nl> writes:

> Op 29-nov-2015 om 10:09 schreef Hans Vlems:
> > I'm pretty sure the 1200 supports 4 GB. The User Manual indicates
> > this as well.
>
> Interesting, considering the age of this system and especially
> considering the fact that much more recent systems, like the
> DS15 I used to own, could maximally take the same amount.

The DS15 is a workstation, while the 1200 is a server, hence the
difference: at least in the old days, servers usually supported more
memory.

[Phillip Helbig (undress to reply)]

In article <2a0fea86-13a0-42a2...@googlegroups.com>, Hans

Vlems <hvl...@freenet.de> writes:

> The other two 5305's have AS1200 cpu's fitted and present themselves as

> such h to the OS. The fourth stays with Phillip in Germany, unless he has
> tossed it.

I would never toss it! At most, I might trade it for something I could
use more, to someone who would give it a good home. I'm not using it at
the moment. I used to have it configured as a satellite and booted it
when I needed Mozilla back when the bootservers in the cluster (24×7)
were weaker (2 VAXen and an EV4 or EV45 Alpha). (The 1200 is too power
hungry to run all the time, and since I have only one, I don't have a
hot spare; I always have a few spares of all hardware I keep running.)
Now, even the boot servers have more memory; I've moved to small ALPHAs
like the XP1000 and the PWS. I have a DS10 as a satellite, but only for
convenience in another room.

Thanks again for bringing me the machine!

[Hans Vlems]

Don't trade it Phillip, I'd rather have it back :-)
How much memory does it have?

[Phillip Helbig (undress to reply)]

In article <e47e78c9-4035-4e97...@googlegroups.com>, Hans

Vlems <hvl...@freenet.de> writes:

> Don't trade it Phillip, I'd rather have it back :-)
> How much memory does it have?

512 MB.

I'm not using it at the moment, but might re-activate it in the future!