Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Alastair Stevens]

Hi folks - I'm having real problems getting our new dual CPU server
going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
an Adaptec SCSI controller and 512Mb DDR SDRAM. I'm not a Linux newbie
at all, but I've never tried running it on such exotic hardware before,
and it doesn't seem happy....

I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
type, can't reboot, can't do anything. Single user mode _does_ let me
in, however, and this is the only progress so far.

I then tried building a custom 2.4.15-pre4 (on another machine), which
compiled perfectly happily, and I installed this on the server. It
panic'd due to failure to mount the root filesystem. I made an
initrd.img, and it then got further (detecting and initialising the SCSI
controller), but still panic'd with the same message.

I then threw down the gauntlet and installed the rawhide Athlon
SMP kernel (based on 2.4.13) which also booted fine but HUNG at the
login prompt, as above. Finally, I tried the i686 version, which spewed
out tons of error messages regarding "invalid symbols" in the ext3
module.

Either way, I'm stumped. Am I up against an Athlon / chipset problem
here, or is something else wrong? What do I need to do to get my
custom-built 2.4.15-pre4 rolling - why can't it mount the root
partition?

Cheers
Alastair

_____________________________________________
Alastair Stevens
MRC Biostatistics Unit
Cambridge UK
---------------------------------------------
phone - 01223 330383
email - alastair.stevens@mrc-bsu.cam.ac.uk
web - www.mrc-bsu.cam.ac.uk

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Arjan van de Ven]

Alastair Stevens wrote:
> 
> Hi folks - I'm having real problems getting our new dual CPU server
> going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with

Ehm you know that XP cpu's don't support SMP configuration ?

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Alastair Stevens]

> > Hi folks - I'm having real problems getting our new dual CPU server
> > going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
>
> Ehm you know that XP cpu's don't support SMP configuration ?

Erm, no....

If this really is the case, then obviously my supplier doesn't know
either, as he's quite definitively stuck two of them on my mobo!
Linux happily detects two XP 1800+ CPUs on boot, but then both SMP
and UP kernels fail in this strange way. If they didn't boot at all, it
would almost be more helpful ;-)

Cheers
Alastair

_____________________________________________
Alastair Stevens
MRC Biostatistics Unit
Cambridge UK
---------------------------------------------
phone - 01223 330383
email - alastair.stevens@mrc-bsu.cam.ac.uk
web - www.mrc-bsu.cam.ac.uk

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Charles Cazabon]

Alastair Stevens <alastair.stevens@mrc-bsu.cam.ac.uk> wrote:
> >
> > Ehm you know that XP cpu's don't support SMP configuration ?
> 
> Erm, no....
> 
> If this really is the case, then obviously my supplier doesn't know
> either, as he's quite definitively stuck two of them on my mobo!

Vendors generally can't tell their arse from their elbow.  The XP does
not officially support SMP operation, and AMD makes no guarantees that
putting them in an SMP configuration won't cause your motherboard to
melt, your dog to get pregnant, etc.

If you want Athlon in an SMP configuration, buy the Athlon MP.  That's
what it's for.

Charles
-- 
-----------------------------------------------------------------------
Charles Cazabon                            <linux@discworld.dyndns.org>
GPL'ed software available at:  http://www.qcc.sk.ca/~charlesc/software/
-----------------------------------------------------------------------

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Matt Bernstein]

At 14:55 -0000 Arjan van de Ven wrote:

>Alastair Stevens wrote:
>>
>> Hi folks - I'm having real problems getting our new dual CPU server
>> going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
>
>Ehm you know that XP cpu's don't support SMP configuration ?

I hope they do; I've just set up a very similar beast (looks like the same
mobo and same CPUs). Is the RAM "registered" ECC? Are your CPUs the same
stepping? One problem we were bitten by was the Radeon DRI, so we disabled
it (in XF86Config-4) and it now seems to at least boot into X. However,
it's not any faster than a dual PIII (1GHz) at the task it's meant to
perform :( both CPUs report 75% usage, and vmstat 1 doesn't show the IO
systems being slugged. Very strange. We're wondering if we've hit memory
bandwidth as the tasks involve some hard sums with big matrices.

$ cat /proc/cpuinfo
processor	: 0
vendor_id	: AuthenticAMD
cpu family	: 6
model		: 6
model name	: AMD Athlon(tm) Processor
stepping	: 2
cpu MHz		: 1526.519
cache size	: 256 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 1
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov
pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow
bogomips	: 3047.42

processor	: 1
vendor_id	: AuthenticAMD
cpu family	: 6
model		: 6
model name	: AMD Athlon(tm) Processor
stepping	: 2
cpu MHz		: 1526.519
cache size	: 256 KB
fdiv_bug	: no
hlt_bug		: no
f00f_bug	: no
coma_bug	: no
fpu		: yes
fpu_exception	: yes
cpuid level	: 1
wp		: yes
flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr pge mca cmov
pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow
bogomips	: 3047.42

RE: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[M. Edward Borasky]

Hard sums with big matrices? Are you using Atlas? It's tuned (or tunable)
for Athlons. There are some tricks -- you need to use the gcc 2.95.x
compiler; 2.96.x and 3.x apparently slow things down. But I've seen close to
peak rates on my 1.333 GHz Athlon Thunderbird with DDR RAM (1P Asus
motherboard, if it matters). In fact, if you can use 3DNOW2 arithmetic
without your answers going bogus on you, it's even better. Try

https://sourceforge.net/project/showfiles.php?group_id=23725

for the source.

--
"Suppose that tonight, while you sleep, a miracle happens - you wake up
tomorrow with what you have longed for. How will you discover that a miracle
happened? How will your loved ones? What will be different? What will you
notice? What do you need to explode into tomorrow with grace, power, love,
passion and confidence?" -- Michael Hall, PhD

M. Edward (Ed) Borasky
Relax! Run Your Own Brain with Neuro-Semantics!
http://www.borasky-research.net/Flyer.htm
mailto:znmeb@borasky-research.net
http://groups.yahoo.com/group/pdx-neuro-semantics


> -----Original Message-----
> From: linux-kernel-owner@vger.kernel.org
> [mailto:linux-kernel-owner@vger.kernel.org]On Behalf Of Matt Bernstein
> Sent: Wednesday, November 14, 2001 7:08 AM
> To: Arjan van de Ven
> Cc: Alastair Stevens; linux-kernel@vger.kernel.org
> Subject: Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]
>
>
> At 14:55 -0000 Arjan van de Ven wrote:
>
> >Alastair Stevens wrote:
> >>
> >> Hi folks - I'm having real problems getting our new dual CPU server
> >> going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP
> chipset, with
> >
> >Ehm you know that XP cpu's don't support SMP configuration ?
>
> I hope they do; I've just set up a very similar beast (looks like the same
> mobo and same CPUs). Is the RAM "registered" ECC? Are your CPUs the same
> stepping? One problem we were bitten by was the Radeon DRI, so we disabled
> it (in XF86Config-4) and it now seems to at least boot into X. However,
> it's not any faster than a dual PIII (1GHz) at the task it's meant to
> perform :( both CPUs report 75% usage, and vmstat 1 doesn't show the IO
> systems being slugged. Very strange. We're wondering if we've hit memory
> bandwidth as the tasks involve some hard sums with big matrices.
>
> $ cat /proc/cpuinfo
> processor	: 0
> vendor_id	: AuthenticAMD
> cpu family	: 6
> model		: 6
> model name	: AMD Athlon(tm) Processor
> stepping	: 2
> cpu MHz		: 1526.519
> cache size	: 256 KB
> fdiv_bug	: no
> hlt_bug		: no
> f00f_bug	: no
> coma_bug	: no
> fpu		: yes
> fpu_exception	: yes
> cpuid level	: 1
> wp		: yes
> flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr
> pge mca cmov
> pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow
> bogomips	: 3047.42
>
> processor	: 1
> vendor_id	: AuthenticAMD
> cpu family	: 6
> model		: 6
> model name	: AMD Athlon(tm) Processor
> stepping	: 2
> cpu MHz		: 1526.519
> cache size	: 256 KB
> fdiv_bug	: no
> hlt_bug		: no
> f00f_bug	: no
> coma_bug	: no
> fpu		: yes
> fpu_exception	: yes
> cpuid level	: 1
> wp		: yes
> flags		: fpu vme de pse tsc msr pae mce cx8 apic sep mtrr
> pge mca cmov
> pat pse36 mmx fxsr sse syscall mmxext 3dnowext 3dnow
> bogomips	: 3047.42
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
>

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Stephen C. Tweedie]

Hi,

On Wed, Nov 14, 2001 at 03:08:25PM +0000, Matt Bernstein wrote:

> I hope they do; I've just set up a very similar beast (looks like the same
> mobo and same CPUs). Is the RAM "registered" ECC? Are your CPUs the same
> stepping? One problem we were bitten by was the Radeon DRI, so we disabled
> it (in XF86Config-4) and it now seems to at least boot into X.

There are known problems in AMD760+Radeon setups, and a workaround is
to avoid asserting RADEON_SOFT_RESET_HBP during init.  The latest
kernels have that fix in the radeon drm.  Using that in conjunction
with an X server containing the same fix, I've finally got a stable
761+radeon setup here.

I think the X server fix went in on the 4.1.99 branch, but I know that
at least the Red Hat XFree86-4.1 rpms have got the patch back-ported.

> it's not any faster than a dual PIII (1GHz) at the task it's meant to
> perform :( both CPUs report 75% usage, and vmstat 1 doesn't show the IO
> systems being slugged. Very strange. We're wondering if we've hit memory
> bandwidth as the tasks involve some hard sums with big matrices.

If the CPUs were bottlenecked on memory then they would still be
pegged at 100% according to the OS.  They'd just get less work done in
a given interval.

Cheers,
 Stephen

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Andreas Boman]

On Wed, 14 Nov 2001 14:55:19 +0000
Arjan van de Ven <arjanv@redhat.com> wrote:

> Alastair Stevens wrote:
> > 
> > Hi folks - I'm having real problems getting our new dual CPU server
> > going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
> 
> Ehm you know that XP cpu's don't support SMP configuration ?
> -

_Officially_. However afaik the XP cpu's are identical to the MP ones. All AMD K7 cpu's 'work' in SMP.

	Andreas
	
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[John Jasen]

On Wed, 14 Nov 2001, Andreas Boman wrote:

> Arjan van de Ven <arjanv@redhat.com> wrote:
> > Ehm you know that XP cpu's don't support SMP configuration ?
>
> _Officially_. However afaik the XP cpu's are identical to the MP ones. All
> AMD K7 cpu's 'work' in SMP.

www.2cpu.com (I think?) had a discussion on this relatively recently. For
the moment, the only difference between MP and XP chips appears to be that
the MP chips are more easily unlocked for overclocking, if I recall the
article correctly.

So, I doubt that the original posters problems are because he's using XP
chips.

--
-- John E. Jasen (jjasen1@umbc.edu)
-- In theory, theory and practise are the same. In practise, they aren't.

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Christian Bornträger]

> > Ehm you know that XP cpu's don't support SMP configuration ?
>
> _Officially_. However afaik the XP cpu's are identical to the MP ones. All
> AMD K7 cpu's 'work' in SMP.

Correct. the EV6-Bus is just a point-to-point connection tho the northbridge. 
SO it is a bus that scales very well in SMP-systems.The most important parts 
of the SMP-logic are in the northbridge of the chipset, and not in the CPU 
(different to the Intel-solution) . Even Durons are SMP-capable.

greetings

Christian Bornträger

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Lars Magne Ingebrigtsen]

Arjan van de Ven <arjanv@redhat.com> writes:

>> Hi folks - I'm having real problems getting our new dual CPU server
>> going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
> 
> Ehm you know that XP cpu's don't support SMP configuration ?

The XP processors aren't certified by AMD to work in SMP
configurations, but the XP and MP processors are identical.

I've got a Tyan Tiger w/ 2x Athlon XP 1800+ running 2.2.19, and it
works just fine.  (And compiles its kernel blindingly fast.  :-)  The
only Tyan-related problem I've heard people having is that it's quite
picky about what type of RAM it tolerates.  It must be registered ECC
RAM.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Alastair Stevens]

> I've got a Tyan Tiger w/ 2x Athlon XP 1800+ running 2.2.19, and it
> works just fine.  (And compiles its kernel blindingly fast.  :-)  The
> only Tyan-related problem I've heard people having is that it's quite
> picky about what type of RAM it tolerates.  It must be registered ECC
> RAM.

Forgive me, but what does 'registered' strictly refer to? It is ECC RAM,
in a single 512Mb module, but more than that, I don't know.

But the key question is still this: is this purely a hardware issue? My
understanding is that with recent 2.4 kernels, Athlon optimisations and
AMD 760 issues are sorted - am I right?

Perhaps 2.4.15-pre4 would actually work, if I could fix the initrd
problem...? Since most -ac stuff is in there, I opted for that rather
than 2.4.13-acX for now, but maybe I should drop back to that?

Cheers
Alastair

_____________________________________________
Alastair Stevens
MRC Biostatistics Unit
Cambridge UK
---------------------------------------------
phone - 01223 330383
email - alastair.stevens@mrc-bsu.cam.ac.uk
web - www.mrc-bsu.cam.ac.uk

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Lars Magne Ingebrigtsen]

Alastair Stevens <alastair.stevens@mrc-bsu.cam.ac.uk> writes:

> Forgive me, but what does 'registered' strictly refer to? It is ECC RAM,
> in a single 512Mb module, but more than that, I don't know.

Registered RAM sticks look somewhat different from normal (ECC or not)
RAM sticks.  They have a few more ICs on the stick, and are usually
bigger than their normal counterparts.

(The "register" bit refers to, I believe, those buffers on the
sticks.)

> But the key question is still this: is this purely a hardware issue? My
> understanding is that with recent 2.4 kernels, Athlon optimisations and
> AMD 760 issues are sorted - am I right?

If you compile a kernel without any Athlon optimizations, it should
work on the Tyan Tiger.  Try that first, and then try out the
optimizations when you've got that working.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Dan Hollis]

On Wed, 14 Nov 2001, Alastair Stevens wrote:
> But the key question is still this: is this purely a hardware issue? My
> understanding is that with recent 2.4 kernels, Athlon optimisations and
> AMD 760 issues are sorted - am I right?

The 'athlon optimizations' is a VIA KT133A bug, afaik AMD760MP is immune
(at least my tests indicate it is)

-Dan
-- 
[-] Omae no subete no kichi wa ore no mono da. [-]

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Matthew Sell]

We're using the 1.4 GHz Athlons (non-MP) in a Tyan S2460 with no problems.

Check some reviews, such as from Tom's Hardware or AnandTech to see some 
specifics regarding the use of non-MP processors in SMP mode. It definitely 
is possible and it works fine, but check those websites for details from 
someone who knows a heck of a lot more about SMP than I do.....     : )


         - Matt



At 03:04 PM 11/14/2001 +0000, you wrote:
> > > Hi folks - I'm having real problems getting our new dual CPU server
> > > going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
> >
> > Ehm you know that XP cpu's don't support SMP configuration ?
>
>Erm, no....
>
>If this really is the case, then obviously my supplier doesn't know
>either, as he's quite definitively stuck two of them on my mobo!
>Linux happily detects two XP 1800+ CPUs on boot, but then both SMP
>and UP kernels fail in this strange way. If they didn't boot at all, it
>would almost be more helpful ;-)
>
>Cheers
>Alastair
>
>_____________________________________________
>Alastair Stevens
>MRC Biostatistics Unit
>Cambridge UK
>---------------------------------------------
>phone - 01223 330383
>email - alastair.stevens@mrc-bsu.cam.ac.uk
>web - www.mrc-bsu.cam.ac.uk
>
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at  http://www.tux.org/lkml/



Matthew Sell
Programmer
On Time Support, Inc.
www.ontimesupport.com
(281) 296-6066

Join the Metrology Software discussion group METLIST!
http://www.ontimesupport.com/cgi-bin/mojo/mojo.cgi


"One World, One Web, One Program" - Microsoft Promotional Ad
"Ein Volk, Ein Reich, Ein Fuhrer" - Adolf Hitler

Many thanks for this tagline to a fellow RGVAC'er...

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Stefan Smietanowski]

Hi.

 >>Hi folks - I'm having real problems getting our new dual CPU server
 >>going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
 >>
 >
 > Ehm you know that XP cpu's don't support SMP configuration ?

Not totally true. AMD doesn't support them in SMP configuration. Ie, 
they run, but if it breaks you get to keep the pieces.

On the MPs, if it breaks, AMD might help you out.

// Stefan

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Dan Hollis]

On Wed, 14 Nov 2001, Arjan van de Ven wrote:
> Alastair Stevens wrote:
> > Hi folks - I'm having real problems getting our new dual CPU server
> > going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
> Ehm you know that XP cpu's don't support SMP configuration ?

Tell that to my dual 1800 XP's. They seem to have completely forgot.

-Dan
-- 
[-] Omae no subete no kichi wa ore no mono da. [-]

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Matthew Sell]

Guys,


Just my $.02 worth on the dual Athlons...

We just finished putting together what was for us a pretty big box using 
the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome 
heating problems.

We had similar problems of hard lockups, usually soon after starting 
SETI@Home.....

We put some fans into the case to circulate air around the surface of the 
board, as well as additional fans on the front and back of the case for 
additional intake/exhaust.

On the S2460, the IC with the black heatsink was reaching 120 deg. F, at 
which point it promptly locked solid. We mounted a miniature fan directly 
on that heatsink, and with the additional circulation, we're keeping it to 
about 95 deg. F, which seems to keep it happy.

I agree with a previous post that cases today are pretty poorly designed 
with respects to air circulation. Of course, this was a server that we made 
using "department store"-style parts, and we expected to get what we paid 
for. It's been doing fine for the last several days with the case 
re-assembled and placed in the server rack with it's brothers.

Hope this helps a little. BTW: We are using the "Enterprise" SMP kernel 
supplied with RedHat 7.2.  It's not optimal for our arrangement, but it 
gets things going until I have a chance to get back to it to compile a 
custom kernel.


         - Matt




At 02:35 PM 11/14/2001 +0000, you wrote:
>Hi folks - I'm having real problems getting our new dual CPU server
>going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
>an Adaptec SCSI controller and 512Mb DDR SDRAM. I'm not a Linux newbie
>at all, but I've never tried running it on such exotic hardware before,
>and it doesn't seem happy....
>
>I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
>kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
>type, can't reboot, can't do anything. Single user mode _does_ let me
>in, however, and this is the only progress so far.
>
>I then tried building a custom 2.4.15-pre4 (on another machine), which
>compiled perfectly happily, and I installed this on the server. It
>panic'd due to failure to mount the root filesystem. I made an
>initrd.img, and it then got further (detecting and initialising the SCSI
>controller), but still panic'd with the same message.
>
>I then threw down the gauntlet and installed the rawhide Athlon
>SMP kernel (based on 2.4.13) which also booted fine but HUNG at the
>login prompt, as above. Finally, I tried the i686 version, which spewed
>out tons of error messages regarding "invalid symbols" in the ext3
>module.
>
>Either way, I'm stumped. Am I up against an Athlon / chipset problem
>here, or is something else wrong? What do I need to do to get my
>custom-built 2.4.15-pre4 rolling - why can't it mount the root
>partition?
>
>Cheers
>Alastair
>
>_____________________________________________
>Alastair Stevens
>MRC Biostatistics Unit
>Cambridge UK
>---------------------------------------------
>phone - 01223 330383
>email - alastair.stevens@mrc-bsu.cam.ac.uk
>web - www.mrc-bsu.cam.ac.uk
>
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at  http://www.tux.org/lkml/



Matthew Sell
Programmer
On Time Support, Inc.
www.ontimesupport.com
(281) 296-6066

Join the Metrology Software discussion group METLIST!
http://www.ontimesupport.com/cgi-bin/mojo/mojo.cgi


"One World, One Web, One Program" - Microsoft Promotional Ad
"Ein Volk, Ein Reich, Ein Fuhrer" - Adolf Hitler

Many thanks for this tagline to a fellow RGVAC'er...

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Alastair Stevens]

> Just my $.02 worth on the dual Athlons...
>
> We just finished putting together what was for us a pretty big box using
> the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome
> heating problems.
>
> We put some fans into the case to circulate air around the surface of the
> board, as well as additional fans on the front and back of the case for
> additional intake/exhaust.

Well, I haven't done any temperature monitoring yet, but the case is a
monstrous Hudson rackmount chassis, with tons of space and several extra
fans, so I imagine that overheating is unlikely.

It may be hopelessly subjective, but the exhaust air from the back feels
cooler than on most of our other machines - but I know it's the CPUs
that count most.

Taking out one CPU may be an option; I'll try some other kernel options
first....

Cheers
Alastair

_____________________________________________
Alastair Stevens
MRC Biostatistics Unit
Cambridge UK
---------------------------------------------
phone - 01223 330383
email - alastair.stevens@mrc-bsu.cam.ac.uk
web - www.mrc-bsu.cam.ac.uk

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Alastair Stevens]

> We just finished putting together what was for us a pretty big box using
> the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome
> heating problems.

Well, I finally managed to check, and both CPUs are at 76C - sounds
quite hot to me. Is that problematic? I've never run these Athlons
before, so I'm not sure what's supposed to be normal ;-)

Going back to kernel issues - I tried the Red Hat enterprise kernel
(2.4.9-13), but that's no go either. Basically, the summary is:

  - Red Hat Kernel (any) boots fine, hangs at login prompt
  - Red Hat Kernel SINGLE USER runs fine and lets me hack around
  - Custom kernel (2.4.15-pre4 SMP) hangs due to failing to
    mount the root partition (initrd issue I think)

I have experienced the "hanging at login prompt" issue before - with
good old Red Hat 7.0. That was solved by upgrading the broken gcc and
broken glibc, and didn't even appear to be a kernel issue.

Cheers
Alastair

o o o o o o o o o o o o o o o o o o o o o o o o o o o o
Alastair Stevens           \ \
MRC Biostatistics Unit      \ \___________ 01223 330383
Cambridge UK                 \___ www.mrc-bsu.cam.ac.uk

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Momchil Velikov]

>>>>> "Alastair" == Alastair Stevens <alastair.stevens@mrc-bsu.cam.ac.uk> writes:

Alastair>   - Red Hat Kernel (any) boots fine, hangs at login prompt
Alastair>   - Red Hat Kernel SINGLE USER runs fine and lets me hack around
Alastair>   - Custom kernel (2.4.15-pre4 SMP) hangs due to failing to
Alastair>     mount the root partition (initrd issue I think)

Make sure filesystem is ok.  I had hangups at the login prompt and xdm
due to bogus utmp (or wtmp).

Regards,
-velco

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Philippe Amelant]

le mer 14-11-2001 à 17:15, Alastair Stevens a écrit :
> > We just finished putting together what was for us a pretty big box using
> > the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome
> > heating problems.
> 
> Well, I finally managed to check, and both CPUs are at 76C - sounds
> quite hot to me. Is that problematic? I've never run these Athlons
> before, so I'm not sure what's supposed to be normal ;-)
> 

It's hot i think, my dual celeron hang around 60°C.

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Charles Marslett]

Philippe Amelant wrote:
> 
> le mer 14-11-2001 à 17:15, Alastair Stevens a écrit :
> > > We just finished putting together what was for us a pretty big box using
> > > the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome
> > > heating problems.
> >
> > Well, I finally managed to check, and both CPUs are at 76C - sounds
> > quite hot to me. Is that problematic? I've never run these Athlons
> > before, so I'm not sure what's supposed to be normal ;-)
> >
> 
> It's hot i think, my dual celeron hang around 60°C.

I agree, but the Thunderbirds (older 1.4 GHz processors) run hotter than anything
else, so this may be normal.  My dual 1.2 GHz MPs run about 72C, and they are
supposed to be a bit cooler than the Thunderbirds.  I don't have much in the
way of cooling, though....

--Charles 
          /"\                           |
          \ /     ASCII Ribbon Campaign |
           X      Against HTML Mail     |--Charles Marslett
          / \                           |  www.wordmark.org

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Gérard Roudier]

On Wed, 14 Nov 2001, Charles Marslett wrote:

> Philippe Amelant wrote:
> >
> > le mer 14-11-2001 à 17:15, Alastair Stevens a écrit :
> > > > We just finished putting together what was for us a pretty big box using
> > > > the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome
> > > > heating problems.
> > >
> > > Well, I finally managed to check, and both CPUs are at 76C - sounds
> > > quite hot to me. Is that problematic? I've never run these Athlons
> > > before, so I'm not sure what's supposed to be normal ;-)
> > >
> >
> > It's hot i think, my dual celeron hang around 60°C.
>
> I agree, but the Thunderbirds (older 1.4 GHz processors) run hotter than anything
> else, so this may be normal.  My dual 1.2 GHz MPs run about 72C, and they are
> supposed to be a bit cooler than the Thunderbirds.

> I don't have much in the way of cooling, though....

In my opinion, you just need some ketch-up and you will be done. :-)

Athlons can survive at more than 90C, but probably not all the stuff that
is closed to them when temperature is so high, in my opinion. I would not
trust anything higher than 55C.

  Gérard.

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Dan Hollis]

On Wed, 14 Nov 2001, Charles Marslett wrote:
> I agree, but the Thunderbirds (older 1.4 GHz processors) run hotter than anything
> else, so this may be normal.  My dual 1.2 GHz MPs run about 72C, and they are
> supposed to be a bit cooler than the Thunderbirds.  I don't have much in the
> way of cooling, though....

Thats far too hot. If youre using lm78, you need to change the sensor type
from the default to "2" or you get bogus readings.

-Dan
-- 
[-] Omae no subete no kichi wa ore no mono da. [-]

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Scott Russell]

In mailinglists.external.linux-kernel, you wrote:
>le mer 14-11-2001 à 17:15, Alastair Stevens a écrit :
>> > We just finished putting together what was for us a pretty big box using
>> > the Tyan S2460 with 1.4GHz Athlons (not MP) and ran into some troublesome
>> > heating problems.
>> 
>> Well, I finally managed to check, and both CPUs are at 76C - sounds
>> quite hot to me. Is that problematic? I've never run these Athlons
>> before, so I'm not sure what's supposed to be normal ;-)
>> 

Spec for the Thunderbird 1.4GHz is maxed at 90c according to AMDs thermal
docs. Not sure what it is for XP but I doubt it would be much cooler if at
all.

-- 
Regards,
 Scott Russell (lnxgeek@us.ibm.com)
 Linux Technology Center, System Admin, RHCE.
 T/L 441-9289 / External 919-543-9289
 http://bzimage.raleigh.ibm.com/webcam

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Dan Hollis]

On Wed, 14 Nov 2001, Alastair Stevens wrote:
> Well, I finally managed to check, and both CPUs are at 76C - sounds
> quite hot to me.

How did you check?

-Dan
-- 
[-] Omae no subete no kichi wa ore no mono da. [-]

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Dan Hollis]

On Wed, 14 Nov 2001, Alastair Stevens wrote:
> Hi folks - I'm having real problems getting our new dual CPU server
> going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
> an Adaptec SCSI controller and 512Mb DDR SDRAM.

It had better well be registered ECC DDR ram. And hopefully on the tyan
approved list:

http://www.tyan.com/support/html/pc2100_tg_mp.html

-Dan
-- 
[-] Omae no subete no kichi wa ore no mono da. [-]

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Herbert Xu]

Alastair Stevens <alastair.stevens@mrc-bsu.cam.ac.uk> wrote:

> I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
> kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
> type, can't reboot, can't do anything. Single user mode _does_ let me
> in, however, and this is the only progress so far.

Try plugging in a mouse or stop running gpm.
-- 
Debian GNU/Linux 2.2 is out! ( http://www.debian.org/ )
Email:  Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Stephen Frost]

[-- Attachment #1: Type: text/plain, Size: 594 bytes --]

* Herbert Xu (herbert@gondor.apana.org.au) wrote:
> Alastair Stevens <alastair.stevens@mrc-bsu.cam.ac.uk> wrote:
> 
> > I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
> > kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
> > type, can't reboot, can't do anything. Single user mode _does_ let me
> > in, however, and this is the only progress so far.
> 
> Try plugging in a mouse or stop running gpm.

	Yeah, and make sure that silly "PCI Interrupts in MP table" or
	whatever in your BIOS is on or you may have other problems.

		Stephen

[-- Attachment #2: Type: application/pgp-signature, Size: 232 bytes --]

[OT] Athlon SMP blues - SOLVED by gpm

[Alastair Stevens]

> > I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
> > kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
> > type, can't reboot, can't do anything. Single user mode _does_ let me
> > in, however, and this is the only progress so far.
>
> Try plugging in a mouse or stop running gpm.

YES YES YES!!! This was it! After 24 hours of building 17 different
kernels, checking out every inch of my hardware, and trying to build
ramdisk images, it was the humble 'gpm' that caused my headaches!
There's no mouse on the machine. Thanks very much indeed, and boy have I
learned something now....

PS - thanks to all who sent in lots of ideas on this problem! It
actually turns out the machine is *not* overheating at all. The 76C BIOS
CPU temperature was erroneous, and in fact it's more like 42C now, which
is perfectly healthy of course ;-)

Cheers
Alastair

o o o o o o o o o o o o o o o o o o o o o o o o o o o o
Alastair Stevens           \ \
MRC Biostatistics Unit      \ \___________ 01223 330383
Cambridge UK                 \___ www.mrc-bsu.cam.ac.uk

Re: [OT] Athlon SMP blues - SOLVED by gpm

[Paul G. Allen]

As the saying goes, "It's probably some little thing."

It's always the little things that seem to bit you the hardest and where
it most hurts. :)

PGA

Alastair Stevens wrote:
> 
> > > I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
> > > kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
> > > type, can't reboot, can't do anything. Single user mode _does_ let me
> > > in, however, and this is the only progress so far.
> >
> > Try plugging in a mouse or stop running gpm.
> 
> YES YES YES!!! This was it! After 24 hours of building 17 different
> kernels, checking out every inch of my hardware, and trying to build
> ramdisk images, it was the humble 'gpm' that caused my headaches!
> There's no mouse on the machine. Thanks very much indeed, and boy have I
> learned something now....
> 
> PS - thanks to all who sent in lots of ideas on this problem! It
> actually turns out the machine is *not* overheating at all. The 76C BIOS
> CPU temperature was erroneous, and in fact it's more like 42C now, which
> is perfectly healthy of course ;-)
> 
> Cheers
> Alastair
> 

-- 
Paul G. Allen
UNIX Admin II/Network Security
Akamai Technologies, Inc.
www.akamai.com

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[bert hubert]

On Thu, Nov 15, 2001 at 07:48:46AM +1100, Herbert Xu wrote:
> Alastair Stevens <alastair.stevens@mrc-bsu.cam.ac.uk> wrote:
> 
> > I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
> > kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
> > type, can't reboot, can't do anything. Single user mode _does_ let me
> > in, however, and this is the only progress so far.
> 
> Try plugging in a mouse or stop running gpm.

But what causes the problem then?

Regards,

bert
-- 
http://www.PowerDNS.com          Versatile DNS Software & Services
Trilab                                 The Technology People
Netherlabs BV / Rent-a-Nerd.nl           - Nerd Available -
'SYN! .. SYN|ACK! .. ACK!' - the mating call of the internet

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Paul G. Allen]

Alastair Stevens wrote:
> 
> Hi folks - I'm having real problems getting our new dual CPU server
> going. It's a 2x Athlon XP 1800+ on a Tyan mobo, AMD 760MP chipset, with
> an Adaptec SCSI controller and 512Mb DDR SDRAM. I'm not a Linux newbie
> at all, but I've never tried running it on such exotic hardware before,
> and it doesn't seem happy....
> 
> I installed Red Hat 7.2 and the machine boots fine, using SMP or UP
> kernels (Red Hat 2.4.9-7), but totally HANGS at the login prompt. Can't
> type, can't reboot, can't do anything. Single user mode _does_ let me
> in, however, and this is the only progress so far.
> 
> I then tried building a custom 2.4.15-pre4 (on another machine), which
> compiled perfectly happily, and I installed this on the server. It
> panic'd due to failure to mount the root filesystem. I made an
> initrd.img, and it then got further (detecting and initialising the SCSI
> controller), but still panic'd with the same message.
> 
> I then threw down the gauntlet and installed the rawhide Athlon
> SMP kernel (based on 2.4.13) which also booted fine but HUNG at the
> login prompt, as above. Finally, I tried the i686 version, which spewed
> out tons of error messages regarding "invalid symbols" in the ext3
> module.
> 
> Either way, I'm stumped. Am I up against an Athlon / chipset problem
> here, or is something else wrong? What do I need to do to get my
> custom-built 2.4.15-pre4 rolling - why can't it mount the root
> partition?
> 

Long thread, including a some misinformation.

I have a Tyan K7 Thunder with dual 1.4GHz Thunderbirds. It took me a week to get it working, and was the initial reason I joined this list. It seems I had one
of the few Thunders on the planet that was trying to run Linux, condidering I bought the board almost as soon as they hit the stores :).

First of all, all Athlons support SMP. Neither AMD nor MoBo vendors support SMP using non-MP/XP Athlons (I've asked a real person on the phone). I selected dual
1.4GHz Thunderbirds because they were slightly faster than 1.2GHZ MP chips, less money, and I already had one at the time (from an A7V133 system that kept
crashing). XP processors are faster, lower power versions of the MP, so of course they support SMP. They all use the same EV6 bus - a point-to-point bus taken
from the DEC Alpha processor and capable of up to 400MHz operation and twice the bandwidth in SMP configurations per MHz as the Intel bus.

Second, kernels before 2.4.8 did not work worth a damn. As I recall, RH 7.1 hardly installed, let alone booted. The MP chipset was very new (and buggy - as they
all seem to be) and not all the support code was there yet when these kernels came out. The BIOS was even newer and only recently do the BIOS upgrades have
decent AGP and other chipset support in them. The kernel I finally got working in a stable way is 2.4.9ac10 with some minor tweaks to make it friendlier with my
GeForce 3, the 760MP chipset, and IDE. I also tweaked the nVidia driver because it couldn't recognize or work properly with the 760MP AGP harware. Before I
compiled my own kernel, I had no end to problems with lock-ups, often during compile time.

The only problem I still have is IDE. I can not run the IDE drive using DMA or the system will hang HARD, usually with the drive access light on. Even with DMA
disabled it might hang under high IDE usage. I will replace the IDE drive with a SCSI drive soon as the SCSI interface works perfectly and very fast. Early MP
chipsets had AGP and DMA hardware bugs, but according to AMD errata, the revision in my MoBo should not have these bugs (that doesn't mean it doesn't have them
though).

In addition, when compiling my kernel, I made sure I included the IDE and SCSI drivers in the kernel, NOT as modules. This avoids having to make a new
initrd.img every time I made a new kernel (and I made many of them, and have 6 different ones still on the machine :) and solved the issue of not mounting root.


Hope this helps.

PGA
-- 
Paul G. Allen
UNIX Admin II/Programmer
Akamai Technologies, Inc.
www.akamai.com
Work: (858)909-3630

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Jason L Tibbitts III]

>>>>> "PGA" == Paul G Allen <pgallen@randomlogic.com> writes:

PGA> I have a Tyan K7 Thunder with dual 1.4GHz Thunderbirds. It took
PGA> me a week to get it working, and was the initial reason I joined
PGA> this list.

Perhaps the Thunder boards are more difficult, because I just bought a
Tiger board, slapped 2 1.4GHz Athlons in it, filled it up with 3GB of
RAM and tossed in an extra 256MB DIMM just to see if it would work
[1], installed Red Hat 7.1 (running 2.4.3, not the most recent
erratum) and proceeded to run two different large-memory numerical
codes (one cache-heavy uniform-stride with heavy prefetch usage and
one with a completely nonuniform access pattern) and repeated kernel
compiles concurrently for a solid week and had no problems.  Then I
upgraded to 1800+ XP processors and repeated things and it all just
works, even though it's got more than the officially supported 3GB and
isn't using officially supported MP processors.

I'm not running high-performance graphics or the like, however, so AGP
obviously isn't getting heavily tested.  It is running X on a G400,
however, and that's working well enough.

1) Actually I did have a problem getting good memory.  1GB DIMMs seem
   somewhat tough to come by and my supplier only has them from Smart
   Modular Technologies.  Fully three out of six sticks we got in were
   bad.  Not just in the Tiger MP board but also in an Abit KG7.  The
   Tiger board may be finicky about memory, but this Smart Modular
   stuff just seems bad.  (It is on Tyan's "approved list", however.)

 - J<

Re: Athlon SMP blues - kernels 2.4.[9 13 15-pre4]

[Alan Cox]

> The only problem I still have is IDE. I can not run the IDE drive using DMA or the system will hang HARD, usually with the drive access light on. Even with DMA
> disabled it might hang under high IDE usage. I will replace the IDE drive with a SCSI drive soon as the SCSI interface works perfectly and very fast. Early MP
> chipsets had AGP and DMA hardware bugs, but according to AMD errata, the revision in my MoBo should not have these bugs (that doesn't mean it doesn't have them
> though).

The earlier MP chipsets die if you have IDE prefetching enabled (see the
errata doc). I'd be suprised if a BIOS had left that on.