Upgrade to 6 core CPU...now desktop hangs?

[Dariusz Piatkowski]

Folks!

I was running the AMD Phenom II X4-965 CPU here for about a year...basically
learning about SMP enabling my machine, etc, etc.

I figured with today's hardware and our beloved OS2 configuration the way it is,
I won't be doing any further major upgrades for some time, and since I have a
working combo I'd just upgrade for a slightly better CPU, so I went to the
Phenom II X6 1100T, which is really the end of the road for the AM2+/AM3 combos.

So...I figured: OK, add 2 more CPUs to my CONFIG.SYS...right?

PSD=OS2APIC.PSD /P=6 /APIC

Well...not so fast...it appears I am running out of memory somewhere...the real
problem is getting stuck on the blue 'PMSHELL' window when my 'startup.cmd' VIO
window pops up, and nothing, I mean nothing get's executed. The VIO window just
sits there...keyboard still seems active, the num-lock flickers, but the VIO
window or the PMSHELL do not accept any inputs.

Throttling this back to a 4 core configuration brings the system back to life on
this X6 Phenom...

Sooo....the big question is: what am I missing here? Any ideas? Do I need to do
a full 'hardware detection' boot next?

Thanks!

[Lars Erdmann]

Does this thing have an NVidia chipset ?

Lars

"Dariusz Piatkowski" <dariusz@_NO-SPAM_mnsi.net> schrieb im Newsbeitrag
news:Qg5I6Bo2seGy-pn2-Kd0UsLfoqjk4@neurobox...

[Dariusz Piatkowski]

Hi Lars!

On Mon, 9 Jan 2012 05:57:25 UTC, "Lars Erdmann" <lars.e...@arcor.de> wrote:

> Does this thing have an NVidia chipset ?
>
> Lars

Nope...I am using the MSI 790X-G45 motherboard
(http://www.msi.com/product/mb/790X-G45.html), this has:

- AMD 790X and SB710 Chipset

Now, here is something strange. After trying repeatedly, playing around with the
device drivers, removing some, putting some back after (choices being: UniAudio,
APM, Unicode, UDF, etc) I finally managed to boot to Desktop with 6 cores
recognized.

This is with "PSD=OS2APIC.PSD /P=6 /APIC" in CONFIG.SYS and the BIOS set to
support MPS 1.4 version (which was my previous setting, working OK with the 4
core CPU).

The strange part is: the boot still hung, however, hitting the combination of
CTRL-ESC, ALT-ESC, and ALT-TAB somehow brought up the "Window List" window,
which was blank, however at this point in time the boot progresses, and soon
enough, I had a working machine.

The remaining issue is that I actually have a continuing CPU spike taking place,
at any point in time 1 of the cores (not the same one continuoulsy, but various
cores) is usually at about 50% load...not sure why...but the behaviour is almost
exact to the previous 'spiking' stuff I was seeing with Firefox usage when
high-res timer was turned ON. Eventually, through some other tests I was able to
narrow this down to what I believe is the cause => libc06b4.dll (of course, just
a suspicion, but removing that DLL caused the spikes to go away). I haven't
tried this on my 6 core configuration yet given that the actual Desktop boot has
been very elusive so far, with success occuring only 3 times tonight.

I am now trying to figure out how the heck to trap/log the delay I am seeing
while executing the normal boot process. Any hints? Haven't tried the FULL
Hardware Detection yet...wanted to narrow this down further first...

--

[Mr. G]

On Tue, 10 Jan 2012 05:01:30 UTC, "Dariusz Piatkowski"
<dariusz@_NO-SPAM_mnsi.net> wrote:

> Now, here is something strange. After trying repeatedly, playing around with the
> device drivers, removing some, putting some back after (choices being: UniAudio,
> APM, Unicode, UDF, etc) I finally managed to boot to Desktop with 6 cores
> recognized.
>
> This is with "PSD=OS2APIC.PSD /P=6 /APIC" in CONFIG.SYS and the BIOS set to
> support MPS 1.4 version (which was my previous setting, working OK with the 4
> core CPU).
>
> The strange part is: the boot still hung, however, hitting the combination of
> CTRL-ESC, ALT-ESC, and ALT-TAB somehow brought up the "Window List" window,
> which was blank, however at this point in time the boot progresses, and soon
> enough, I had a working machine.
>

My vote goes to either Uniaud or USB

[Lars Erdmann]

Yes very well possible. I just found some bug in USBOHCD.SYS that most
likely does bad things to your system.
It would effectively program the wrong physical address into a register
(pointing to nowhere land) and the risk that the USB DMA engine would
subsequently write to that address before it is finally set to the correct
value.
No memory protection with DMA, therefore no traps and orderly machine stop
...
I hope to get out a new set of drivers in about 2 weeks.

On the other hand: I would try to run OS2APIC.PSD without the /APIC
parameter. That is equivalent to
ACPI.PSD /VW (in short: you run your system in "virtual wire" mode).
Some MP table are so grossly broken that it's worth giving it a try. There
are other parameters to OS2APIC.PSD to also try out.

Lars


"Mr. G" <nos...@yahoo.com> schrieb im Newsbeitrag
news:G8df4sdtcWG8-pn2-gMstiV3rddlI@localhost...

[Ilya Zakharevich]

On 2012-01-10, Lars Erdmann <lars.e...@arcor.de> wrote:
> I hope to get out a new set of drivers in about 2 weeks.

Thanks for that. Do you get any updates for not-waking-up bug? On
the Thinkpad T23 here (with USB 1.1), the latest-at-summer drivers
would not behave well after suspend (no other problems except for USB
not visible). I needed to revert to USBUHCD.SYS-10_158.

Yours,
Ilya

[Trevor Hemsley]

On Tue, 10 Jan 2012 05:01:30 UTC in comp.os.os2.setup.misc, "Dariusz Piatkowski"
<dariusz@_NO-SPAM_mnsi.net> wrote:

> This is with "PSD=OS2APIC.PSD /P=6 /APIC" in CONFIG.SYS and the BIOS set to
> support MPS 1.4 version (which was my previous setting, working OK with the 4
> core CPU).

/APIC never got much beyond the beta stage so you might try it without.

Also, I have got past many issues like you describe by using SET
RESTARTOBJECTS=NO in config.sys

--
Trevor Hemsley, Brighton, UK
Trevor dot Hemsley at ntlworld dot com

[Lars Erdmann]

Suspend/Resume is on the todo list. The code is pretty messy. But before I
tackle that I will need to reach the stage where normal operation is
reliable on all flavours: UHCI, OHCI, EHCI.

Lars

"Ilya Zakharevich" <nospam...@ilyaz.org> schrieb im Newsbeitrag
news:slrnjgp2so.6k...@panda.math.berkeley.edu...

[Dariusz Piatkowski]

Tried both, separately AND together...no luck...at least not visible luck...the
boot still gets stuck when the STARTUP.CMD VIO window pops-up, no text is
displayed in it however...

[Dariusz Piatkowski]

Hi Lars!

On Tue, 10 Jan 2012 18:41:05 UTC, "Lars Erdmann" <lars.e...@arcor.de> wrote:

> Yes very well possible. I just found some bug in USBOHCD.SYS that most
> likely does bad things to your system.
> It would effectively program the wrong physical address into a register
> (pointing to nowhere land) and the risk that the USB DMA engine would
> subsequently write to that address before it is finally set to the correct
> value.
> No memory protection with DMA, therefore no traps and orderly machine stop
> ...
> I hope to get out a new set of drivers in about 2 weeks.

OK, looking forward to giving these a go. As-is, I attempted to rule out the USB
stuff already (for what it's worth)...I removed ALL the USB drivers, that did
not appear to have any impact on the situation.

I also removed the 'Legacy' support in BIOS...that appeared to have hung my
system hard (no keyboard response, etc).

> On the other hand: I would try to run OS2APIC.PSD without the /APIC
> parameter. That is equivalent to
> ACPI.PSD /VW (in short: you run your system in "virtual wire" mode).
> Some MP table are so grossly broken that it's worth giving it a try. There
> are other parameters to OS2APIC.PSD to also try out.
>
> Lars

Yes, tried this approach as well. Hard to tell if it had any impact, the total
number of complete boots today was about 2-3...so no different from last night.
Mind you, I am trying such a large combination of stuff it's getting hard to
track it all...even when I'm writing this down...LOL

>
>
> "Mr. G" <nos...@yahoo.com> schrieb im Newsbeitrag
> news:G8df4sdtcWG8-pn2-gMstiV3rddlI@localhost...
> > On Tue, 10 Jan 2012 05:01:30 UTC, "Dariusz Piatkowski"
> > <dariusz@_NO-SPAM_mnsi.net> wrote:
> >
> >> Now, here is something strange. After trying repeatedly, playing around
> >> with the
> >> device drivers, removing some, putting some back after (choices being:
> >> UniAudio,
> >> APM, Unicode, UDF, etc) I finally managed to boot to Desktop with 6 cores
> >> recognized.
> >>
> >> This is with "PSD=OS2APIC.PSD /P=6 /APIC" in CONFIG.SYS and the BIOS set
> >> to
> >> support MPS 1.4 version (which was my previous setting, working OK with
> >> the 4
> >> core CPU).
> >>
> >> The strange part is: the boot still hung, however, hitting the
> >> combination of
> >> CTRL-ESC, ALT-ESC, and ALT-TAB somehow brought up the "Window List"
> >> window,
> >> which was blank, however at this point in time the boot progresses, and
> >> soon
> >> enough, I had a working machine.
> >>
> > My vote goes to either Uniaud or USB
>

--

[Dariusz Piatkowski]

Hi Trevor!

On Tue, 10 Jan 2012 21:36:30 UTC, "Trevor Hemsley"
<Trevor....@mytrousers.ntlworld.com> wrote:

> On Tue, 10 Jan 2012 05:01:30 UTC in comp.os.os2.setup.misc, "Dariusz Piatkowski"
> <dariusz@_NO-SPAM_mnsi.net> wrote:
>
> > This is with "PSD=OS2APIC.PSD /P=6 /APIC" in CONFIG.SYS and the BIOS set to
> > support MPS 1.4 version (which was my previous setting, working OK with the 4
> > core CPU).
>
> /APIC never got much beyond the beta stage so you might try it without.
>
> Also, I have got past many issues like you describe by using SET
> RESTARTOBJECTS=NO in config.sys

Yup, as Lars suggested, I tried dropping the /APIC option, however that did not
appear to improve teh situation.

Now, regarding the follow-up suggestion of yours, I attempted the following
(based on some previous Intenet searches):

SET RUNWORKPLACE=G:\OS2\CMD.EXE
SET RESTARTOBJECTS=NO
SET SHELLEXCEPTIONHANDLER=OFF
SET SHAPIEXCEPTIONHANDLER=OFF

...the result was a hard-lock however...what I haven't tried was leaving the full
CONFIG.SYS setup as-is in my 4-core configuration and attemptign the SET
RESTARTOBJECTS=NO on it's own.

I will try that tomorrow and will follow-up with my results.

So far, strangely enough (to me anyways) it doesn't appear like I'm going to
have much luck moving to the 6-core configuration.

I wish that the eCS ACPI solution could be tried out on some type of a trial
basis...because if it works I would be happy to pay the price and move along to
more happy times...LOL...yes, but I'm just getting 'greedy' here trying to get
the full 6 cores running...

[Dave Yeo]

Dariusz Piatkowski wrote:
> I wish that the eCS ACPI solution could be tried out on some type of a trial
> basis...because if it works I would be happy to pay the price and move along to
> more happy times...LOL...yes, but I'm just getting 'greedy' here trying to get
> the full 6 cores running...

I tried for quite a while to get both my cores working on Warp 4 without
any luck (always crashed when doing anything, even with the most minimal
system). I finally shelled out for eCS and as soon as I enabled SMP
things just worked. Also nice to get rid of 15 years of cruft.
BTW, even with only one core I often (maybe 25% of the time) had boot
failures like you're experiencing. One thing I did that really improved
things here was to take a blower to my box. Huge cloud of dust came out
and a few boot oddities went away. Just have to be careful to not let
your fans over spin if you blow it out.
Dave
ps The ACPI readme also stresses to physically clean out your computer
before even trying it.

[Ilya Zakharevich]

On 2012-01-10, Lars Erdmann <lars.e...@arcor.de> wrote:
> Suspend/Resume is on the todo list. The code is pretty messy. But before I
> tackle that I will need to reach the stage where normal operation is
> reliable on all flavours: UHCI, OHCI, EHCI.

Well, one of the reasons I switched back from your version is that it
DID NOT fix the problems I had with UHCD (even before the first
suspend). I get 30sec timeouts on each read from newer memory sticks;
if a program reads in 5K chunks, then this gives is 1.5KB/sec
throughput. ;-)

Thanks for your work,
Ilya

[Mr. G]

I only mentioned USB as I recently had a USB hang on my OS/2
partition which had not been booted to for quite some time.

I expect you have examined the contents of STARTUP.CMD
to try and find a possible cause even though no text is displayed.

[Dariusz Piatkowski]

Yeah, I am only running the LBMix & UniAudio mixer settings:

===
detach G:\APPS\MULTIMEDIA\LBMIX\PIPEMIX\Mix90.EXE -l 500
detach G:\APPS\MULTIMEDIA\LBMix\LBMIX.EXE -x -i LBMixP
detach g:\mmos2\unimix -loadg:\mmos2\unimix_audio_settings
rem NET START REQ
rem @if errorlevel 2 G:\IBMLAN\NETPROG\lserr.exe

EXIT
===

As is, I had also tried to rule this out as a source of the problem...I re-named
the file, instead I simply got the watch-mouse pointer on a completely BLUE
screen...so it appears that PMSHELL is getting hung somehow.

[Doug Bissett]

This is actually a very old problem, known as a "startup conflict". It
can usually be caused/fixed, by simply changing the order of starting
things in CONFIG.SYS, so that they don't step on each other.
Unfortunately, it is not possible to say "do this" and have positive
results (just as often, the results are negative). You need to play
with things, to work around the problem.

FWIW, one of the main causes of that problem (but not the only one),
is the same thing that causes looping sounds at startup. The fix for
that one is to use the mmfix.exe program, which comes with eCS 2.1.
See "A Note About Repeating System Sounds" at:

> http://svn.netlabs.org/uniaud/

for more information about that.

The latest indication, that I saw, happened after I installed the new
ACPI package to my T43. It worked perfectly (well as perfectly as ACPI
has ever worked) with eCS 2.0, but in eCS 2.1, boot stopped at not
being able to load LANMSGEX.EXE (the message indicated that it
couldn't read the file). The work around turned out to be to use
BASEDEV=BOOTDLY.SYS /D:2
(comes with eCS 2.1) immediately after
PSD=ACPI.PSD

Previous to that, I have seen many other things that have caused the
problem, and it still happens intermittently, without changing
anything. Reboot always works, in that case.

Your problem is probably caused because OS2ACPI.PSD takes a little
longer to get 6 processors going, or, 6 processors can get other
things started sooner than was possible with only 4 processors.

You must also remember that all RUN, and CALL statements, in
CONFIG.SYS, are started at the same time that STARTUP.CMD is executed.
The programs can step on each other. The work around for that is to
take things out of STARTUP.CMD, and start things by making an icon in
the STARTUP folder (you can do the same with RUN and CALL statements
that are in CONFIG.SYS, but that can cause more trouble than it
fixes). That delays the program startup slightly, but STARTUP.CMD is
not stepped on). Of course, then you can have the problem where all
programs in the STARTUP folder get started at the same time. The fix
for that is to use REXX AUTOSTART to start the STARTUP folder
contents, one at a time. Startup may take a few extra seconds, but you
don't have conflicts (once you get it adjusted properly). I find that
it is essential to use REXX AUTOSTART if there are more than about 4
items in the STARTUP folder:

> http://www.7cities.net/~mckinnis/rxautost/index.html

The startup sound is also involved, so turning that off/on can also
change the timing.

Note that startup conflicts can, and do, cause problems at startup,
while running, and at shutdown. It all depends on exactly what the
conflict is.

Hope something here helps...
--
From the eComStation of Doug Bissett
dougb007 at telus dot net
(Please make the obvious changes, to e-mail me)

[Lars Erdmann]

What is the last version you used ? I have changed a lot over time.
You could also write a bug report and do the usual stuff: attach PCI -D
output, attach trace.

Lars

"Ilya Zakharevich" <nospam...@ilyaz.org> schrieb im Newsbeitrag

news:slrnjgqcv9.7o...@panda.math.berkeley.edu...

[Dariusz Piatkowski]

Just a brief update regarding Trevor's suggestion...tried it...no
change...repeated re-boots and seeing the same hang.

On Wed, 11 Jan 2012 04:33:22 UTC, "Dariusz Piatkowski"
<dariusz@_NO-SPAM_mnsi.net> wrote:

--

[Dariusz Piatkowski]

Hi Doug!

Alright...so re-boot, after re-boot, after re-boot...I had tried all the
combinations I could think off, including basically stripping down the
CONFIG.SYS to what seemed like just the bare essentials...basically the OS,
-sound -USB -utilities, what remained was the Network stuff.

I had removed all my entries from the regular STARTUP folder and moved them all
to the XWorkplace startup, which is pre-defined to start up 2 seconds AFTER the
Desktop populates.

I also dwindled down the HPFS386 cache to 64Meg, tried a run with cache turned
OFF, various VIRTUALADDRESSLIMIT combinations (2048 is what my 4 core system
seems to like) and the OS2APCI with, and w/o the /APIC parameter.

At this point in time I am basically giving up on this...maybe eventually the
news will reach me that eCS ACPI is OK, and then I will give it a try.

I'm curious though if anyone out there is running a 6 core combination I would
appreciate hearing from you.

Thanks again for all your help guys...keep the ideas coming...I am always happy
to give it another whirl...but for now, I'm ready to get back to normal
computing that doesn't keep me up until the wee 1-1:30am morning hours...LOL!

[Mr. G]

Just for kicks, have you tried the xWorkplace feature of holding down
the shift key just before WPS is supposed to come up. Does that not
bring up the ifix feature or something like that? Maybe you could find
something in there?

[Lars Erdmann]

try these:

PSD=OS2APIC.PSD /P=6 /PREC=ILD
PSD=OS2APIC.PSD /P=6 /PREC=LID
PSD=OS2APIC.PSD /P=6 /PREC=DLI /NMI=LINT1 /PIC=LINT0
PSD=OS2APIC.PSD /P=6 /PREC=DLI /NMI=LINT0 /PIC=LINT1

Lars

"Dariusz Piatkowski" <dariusz@_NO-SPAM_mnsi.net> schrieb im Newsbeitrag

news:Qg5I6Bo2seGy-pn2-98SSYRk4T3pT@neurobox...

[Dariusz Piatkowski]

Hi Lars!

Well, some good news...and some rather...umm...very bad news....

I had previously tried some of your suggestions in the past as I was trying to
deal with this pesky occasional soft-hang on my machine. It would only happen in
a certain situation...eventually, I was unable to resolve it.

Anyways, I was thrilled to try these latest combinations you posted, and one of
them did hold promise. The "PSD=OS2APIC.PSD /P=6 /PREC=LID" actually brought me
to a fully populated DESKTOP, I was able to run things fine, basically use my
apps. The only problem: I could not do this consistently...meaning, on one boot
things would go OK, still a delay in the boot, but hitting a few CTRL-ESC
sequences allowed the boot to progress.

As it happened I somehow wanted to see if overclocking the hardware would help
out with the issue I was seeing. Not really knowing what was causing the
problems I was seeing I had nothing else to try.

Subsequently I cranked up the CPU vcore, upped the CPU-NB bridge frequency along
with the HTT, also tossed some increased voltage for a good measure...LOL...ouch
indeed. I knew what I was doing was risky, but I've toyed with PC hardware in
the past and understood the risks.

Well....drummroll please....1st, it worked great, the machine booted w/o any
hangs/delays...chugged along rather nicely for about 5-10 mins...then....THINGS
WENT DARK!!!

1st time ever, I had actually fried something...I scrambled a bit, found a new
replacement motherboard as the one I was using wouldn't even stay powered up
(and I tried everything I could think of to clear the CMOS, heck, I even tried
my 'spare' 2core and 4core CPUs...). Bottom line, my 6-core SMP experimentation
is over for the time being.

The replacement board (still MSI as the previous one was pretty well behaved in
OS/2) doesn't quite work that great. Being in a rush I had picked up what was
available at the store, ended up with 880G-E45 (previous had 790X-G45). The new
board has built-in video, only 1 PS2 connector, a ton of built-in stuff (more
USB controllers too), and somewhat different BIOS. Needless to say, attempting
to boot OS2 in the 6-core configuration produces a very HARD hang on the
RESOURCE.SYS driver.

I have:

5-21-02 3:36p 55402 0 RESOURCE.SYS

Signature: @#IBM:10.070#@ Resource Management Driver
Vendor: IBM
Revision: 10.70
File Version: 10.70
Description: Resource Management Driver

...but what it's actually getting stuck on, I have absolutely no idea.

The old board would chug along happily past the RESOURCE.SYS driver and would
only get stuck upon the Desktop load.

Good news is: the latest USB drivers you released (v 179) appear to work very
well. No TRAPs or hangs on boot...I'm using the FULL driver set here, including
your version of USBD.SYS. Unfortunately I can no longer test the bug-fix you
provided on my old board.

Thanks for all the suggestions, much appreciated. Now I need to figure out why
during boot I am getting 'out of memory' complaints...I suspect all the extra
built-in stuff is just getting in the way.

-Dariusz

--

[Paul Ratcliffe]

On 20 Jan 2012 04:26:10 GMT, Dariusz Piatkowski <dariusz@_NO-SPAM_mnsi.net>
wrote:

> As it happened I somehow wanted to see if overclocking the hardware would help
> out with the issue I was seeing. Not really knowing what was causing the
> problems I was seeing I had nothing else to try.

It's blindingly obvious that overclocking something is NEVER going to help
anything. It's also pretty pointless, as well as all the other potential
consequences... as you found out.

[Dariusz Piatkowski]

Hi Paul,

Hmm...given the symptoms I was seeing, namely intermittent no-problem boot into
the 6-core configuration, overclocking certainly had a lot to offer. For
starters, CPU frequency as well as memory settings will have direct impact on
the behaviour of the system. Not to even mention things like the CPU built-in
memory controller, CPU-NB timing, etc, etc...the list goes on, and you know what
I'm talking about if you understand any of the AMD Phenom II CPU and chipset
architecture. In my case, it appears doing so did improve the boot situation,
although at the expense of the hardware itself. As with many other posts on the
subject of 'Desktop boot hang', it all appears to be related to some type of a
timing issue.

Bummer??? Umm...perhaps, but this was an el-cheapo MB (total expense $20 when
purchased NEW over 1 yr ago...so yeah, I think I got my money's worth out of
it). Do I wish that it lived longer? For sure...failure is no fun, no matter
when...lol...live and learn though...and other then the obvious
challenge/nuisance factor, it's a bit fun after all, how else do you plan on
learning and discovering new things?

Just because I'm an OS2 user by choice I have no plans to bow my head down and
accept the OEM supplied settings for my systems. Countless thousands out there
play around with this stuff. Can a successfull overclock work? You bet your
behind...there certainly is some thrill to being able to run a family DSLR 1080p
movie encoding, while I'm watching another 1080 video, serving content over the
web-server, and pulling down some of my kids' favourite TV comics...so yeah,
fun.

If anything, I wish more folks would do this. Since our community is so small,
it's exceedingly difficult to find any real knowledge and insight into these
types of problems. The real know-how of the inner-workings appears to be gone
and/or moved-on. For instance, quite frankly, your post added absolutely nothing
to this discussion...sorry, just being blunt here. I do welcome real
contributions that extend our understanding of the problem and hopefully the
identification of proper way to resolve the problem. So if you feel up to the
challenge, by all means, feel free to re-consider your input and provide a real
suggestion worth carrying on with.

[martinot]

Did it work out?

--
br,
martinot

[ivan]

On Sat, 28 Feb 2015 16:55:47 UTC, martinot <mart...@objectmail.com>
wrote:

Why are you digging up these old 2012 posts?


ivan
--