After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using the Red Hat
RPM downloaded from RH Network, the CPU count on the machine reported by
dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
This has occured on all 4 Dell 2650's that I've installed this patch on. I
don't have any other mult-processor machines available to test this with.
Things seem to run OK, but this is a bit ominous. Has anyone seem this
problem with this or other PC Models?
Thanks,
Bob
DISCLAIMER: The information contained herein is confidential and is intended
solely for the addressee(s). It shall not be construed as a recommendation
to buy or sell any security. Any unauthorized access, use, reproduction,
disclosure or dissemination is prohibited. Neither ISLAND nor any of its
subsidiaries or affiliates shall assume any legal liability or
responsibility for any incorrect, misleading or altered information
contained herein. Thank you.
> After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using
> the Red Hat
> RPM downloaded from RH Network, the CPU count on the machine
> reported by
> dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
>
> This has occured on all 4 Dell 2650's that I've installed
> this patch on. I
> don't have any other mult-processor machines available to
> test this with.
Congratulations, you purchased a fine PowerEdge 2650 with processors which
contain HyperThreading technology. Each physical processor appears as two
logical processors. This behaviour is expected, and correct. :-)
Thanks,
Matt
--
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
#1 US Linux Server provider for 2001 and Q1/2002! (IDC May 2002)
Matt,
Thanks for the reply. First, let me say thank you very much for a super web
site that helped us sort out issues with the 2650 RAID stuff.
I'm not familiar with the HyperThreading concept.
Do you know of any docs that discuss this. I'm particularly concerned with
how this impacts the results of monitoring tools such as top.
Thanks,
Bob
-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Thursday, July 18, 2002 4:01 PM
To: [email protected]; [email protected]
Subject: RE: Wrong CPU count
> After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using
> the Red Hat
> RPM downloaded from RH Network, the CPU count on the machine
> reported by
> dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
>
> This has occured on all 4 Dell 2650's that I've installed
> this patch on. I
> don't have any other mult-processor machines available to
> test this with.
Congratulations, you purchased a fine PowerEdge 2650 with processors which
contain HyperThreading technology. Each physical processor appears as two
logical processors. This behaviour is expected, and correct. :-)
Thanks,
Matt
--
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
#1 US Linux Server provider for 2001 and Q1/2002! (IDC May 2002)
DISCLAIMER: The information contained herein is confidential and is intended
solely for the addressee(s). It shall not be construed as a recommendation
to buy or sell any security. Any unauthorized access, use, reproduction,
disclosure or dissemination is prohibited. Neither ISLAND nor any of its
subsidiaries or affiliates shall assume any legal liability or
responsibility for any incorrect, misleading or altered information
contained herein. Thank you.
>
> I'm not familiar with the HyperThreading concept.
>
Looks like this could be a good place for you to start:
http://cedar.intel.com/cgi-bin/ids.dll/topic.jsp?catCode=CDN
Bruce H.
And doubles the cost of licensing software that uses per cpu licensing while giving marginally better performance.
-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Thursday, July 18, 2002 3:01 PM
To: [email protected]; [email protected]
Subject: RE: Wrong CPU count
> After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using
> the Red Hat
> RPM downloaded from RH Network, the CPU count on the machine
> reported by
> dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
>
> This has occured on all 4 Dell 2650's that I've installed
> this patch on. I
> don't have any other mult-processor machines available to
> test this with.
Congratulations, you purchased a fine PowerEdge 2650 with processors which
contain HyperThreading technology. Each physical processor appears as two
logical processors. This behaviour is expected, and correct. :-)
Thanks,
Matt
--
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
#1 US Linux Server provider for 2001 and Q1/2002! (IDC May 2002)
That's interesting. Can it be disabled?
-----Original Message-----
From: Hubbard, Dwight [mailto:[email protected]]
Sent: Thursday, July 18, 2002 4:30 PM
To: [email protected]; [email protected];
[email protected]
Subject: RE: Wrong CPU count
And doubles the cost of licensing software that uses per cpu licensing while
giving marginally better performance.
-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Thursday, July 18, 2002 3:01 PM
To: [email protected]; [email protected]
Subject: RE: Wrong CPU count
> After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using
> the Red Hat
> RPM downloaded from RH Network, the CPU count on the machine
> reported by
> dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
>
> This has occured on all 4 Dell 2650's that I've installed
> this patch on. I
> don't have any other mult-processor machines available to
> test this with.
Congratulations, you purchased a fine PowerEdge 2650 with processors which
contain HyperThreading technology. Each physical processor appears as two
logical processors. This behaviour is expected, and correct. :-)
Thanks,
Matt
--
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
#1 US Linux Server provider for 2001 and Q1/2002! (IDC May 2002)
* Hubbard, Dwight ([email protected]) wrote:
> And doubles the cost of licensing software that uses per cpu licensing while giving marginally better performance.
I would guess that depends on how the license defines a 'CPU'. A
license which uses Linux's /proc/cpuinfo output would seem rather flawed
to me, though I suppose possible.
Stephen
>
> And doubles the cost of licensing software that uses per cpu
> licensing while giving marginally better performance.
>
??
I don't wan't to start a flamewar here, but while what you say above is
true..
On a GNU/Linux system loaded with free software, I'll take all the added
performance I can get...
Regards,
Bruce H.
> That's interesting. Can it be disabled?
It's an option in the BIOS Setup page (press F2 shortly after power-on) on
Dell servers which support disabling it.
Thanks,
Matt
--
Matt Domsch
Sr. Software Engineer, Lead Engineer, Architect
Dell Linux Solutions http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
#1 US Linux Server provider for 2001 and Q1/2002! (IDC May 2002)
>
> > That's interesting. Can it be disabled?
>
> It's an option in the BIOS Setup page (press F2 shortly after
> power-on) on
> Dell servers which support disabling it.
That does raise a question from me, though. With the ability disabled at
the bios level, will the kernel still be able to recognize the cpu's as
hyperthreaded, and bypass the bios disabling them?
Thanks,
Bruce H.
On Thu, 18 Jul 2002, Holzrichter, Bruce wrote:
>
> That does raise a question from me, though. With the ability disabled at
> the bios level, will the kernel still be able to recognize the cpu's as
> hyperthreaded, and bypass the bios disabling them?
>
And that arises a question for me too, can we enable/disable HT at boot
time? (think in a mobo without the option to enable/disable HT)
Best regards
--
Robinson Maureira Castillo
Asesor DAI
INACAP
On Thu, Jul 18, 2002 at 04:59:14PM -0400, Robinson Maureira Castillo wrote:
> And that arises a question for me too, can we enable/disable HT at boot
> time? (think in a mobo without the option to enable/disable HT)
You can disable it with 'noht'.
Dave
--
| Dave Jones. http://www.codemonkey.org.uk
| SuSE Labs
yes, as bios option.
On my point of view it would be interesting to verify is hyperthreading is
really usefull or not.
Recently I am studying new Athlon XP architecture, and I am programming
to study Xeon as soon, and then to compare them, to decide to which
architecture I should refer for new servers. Point is that instrction unit
management is too different, while real word benchs are 50% pro Athlon,
50% pro Xeon depending on the applications (well, more or less) ...
On Thu, 18 Jul 2002, Robert Sinko wrote:
> Date: Thu, 18 Jul 2002 16:35:29 -0400
> From: Robert Sinko <[email protected]>
> To: "'Hubbard, Dwight'" <[email protected]>, [email protected],
> [email protected]
> Subject: RE: Wrong CPU count
>
> That's interesting. Can it be disabled?
>
> -----Original Message-----
> From: Hubbard, Dwight [mailto:[email protected]]
> Sent: Thursday, July 18, 2002 4:30 PM
> To: [email protected]; [email protected];
> [email protected]
> Subject: RE: Wrong CPU count
>
>
> And doubles the cost of licensing software that uses per cpu licensing while
> giving marginally better performance.
>
> -----Original Message-----
> From: [email protected] [mailto:[email protected]]
> Sent: Thursday, July 18, 2002 3:01 PM
> To: [email protected]; [email protected]
> Subject: RE: Wrong CPU count
>
>
> > After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using
> > the Red Hat
> > RPM downloaded from RH Network, the CPU count on the machine
> > reported by
> > dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
> >
> > This has occured on all 4 Dell 2650's that I've installed
> > this patch on. I
> > don't have any other mult-processor machines available to
> > test this with.
>
> Congratulations, you purchased a fine PowerEdge 2650 with processors which
> contain HyperThreading technology. Each physical processor appears as two
> logical processors. This behaviour is expected, and correct. :-)
>
> Thanks,
> Matt
>
> --
> Matt Domsch
> Sr. Software Engineer, Lead Engineer, Architect
> Dell Linux Solutions http://www.dell.com/linux
> Linux on Dell mailing lists @ http://lists.us.dell.com
> #1 US Linux Server provider for 2001 and Q1/2002! (IDC May 2002)
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
>
> DISCLAIMER: The information contained herein is confidential and is intended
> solely for the addressee(s). It shall not be construed as a recommendation
> to buy or sell any security. Any unauthorized access, use, reproduction,
> disclosure or dissemination is prohibited. Neither ISLAND nor any of its
> subsidiaries or affiliates shall assume any legal liability or
> responsibility for any incorrect, misleading or altered information
> contained herein. Thank you.
>
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at http://www.tux.org/lkml/
>
On Fri, 19 Jul 2002 [email protected] wrote:
>
> yes, as bios option.
>
> On my point of view it would be interesting to verify is hyperthreading is
> really usefull or not.
>
It would be interesting to determine if "hyperthreading" in the CPU
actually exists. It may just be an artifact of dual instruction units,
actually a defect (perhaps harmless), that is hyped as a feature.
For instance, it has long been known that if a CPU were to have as
many instruction units as possible instruction branches, program
jumps upon logical conditions would not slow the machine down. The
hardware just continues using the instruction unit that contains the
correct program-flow while the others are re-loaded.
I guess that this is what is happening. After all, the processor only
has "so-many" connections to the outside world so it can't actually
function as two processors but, as Clinton said; "It depends upon what
is is..."
Cheers,
Dick Johnson
Penguin : Linux version 2.4.18 on an i686 machine (797.90 BogoMips).
The US military has given us many words, FUBAR, SNAFU, now ENRON.
Yes, top management were graduates of West Point and Annapolis.
On Thu, 18 Jul 2002, Holzrichter, Bruce wrote:
> >
> > I'm not familiar with the HyperThreading concept.
> >
>
> Looks like this could be a good place for you to start:
> http://cedar.intel.com/cgi-bin/ids.dll/topic.jsp?catCode=CDN
I did a research paper for my computer architecture class. The main two
references for hyperthreading (aka simultaneous multithreading) are:
http://www.cs.washington.edu/research/smt/index.html
http://developer.intel.com/technology/hyperthread/
Both sites have numerous links to other resources on the subject.
On Thursday 18 July 2002 01:07 pm, Robert Sinko wrote:
> Matt,
>
> Thanks for the reply. First, let me say thank you very much for a super web
> site that helped us sort out issues with the 2650 RAID stuff.
>
> I'm not familiar with the HyperThreading concept.
>
> Do you know of any docs that discuss this. I'm particularly concerned with
> how this impacts the results of monitoring tools such as top.
>
> Thanks,
> Bob
Take a look at http://www.intel.com and search for Hyperthreading, should find
an article that will help...
Dont know what 'impact' you're concerned about, top will report the
two instruction units as two processors and show you what they
are doing.
Long as you run a kernel with the appropriate support in it you'll
be fine :)
> -----Original Message-----
> From: [email protected] [mailto:[email protected]]
> Sent: Thursday, July 18, 2002 4:01 PM
> To: [email protected]; [email protected]
> Subject: RE: Wrong CPU count
>
> > After upgrading from kernel 2.4.7-10smp to 2.4.9-34smp using
> > the Red Hat
> > RPM downloaded from RH Network, the CPU count on the machine
> > reported by
> > dmesg and listed in /proc/cpuinfo was 4 rather than the actual 2.
> >
> > This has occured on all 4 Dell 2650's that I've installed
> > this patch on. I
> > don't have any other mult-processor machines available to
> > test this with.
>
> Congratulations, you purchased a fine PowerEdge 2650 with processors which
> contain HyperThreading technology. Each physical processor appears as two
> logical processors. This behaviour is expected, and correct. :-)
--
Jack F. Vogel IBM Linux Technology Center
[email protected] (work) || [email protected] (home)
On Thu, 18 Jul 2002, Hubbard, Dwight wrote:
> And doubles the cost of licensing software that uses per cpu licensing while giving marginally better performance.
You can turn it off if you don't want it.
--
bill davidsen <[email protected]>
CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.
On Fri, 19 Jul 2002, Richard B. Johnson wrote:
> On Fri, 19 Jul 2002 [email protected] wrote:
>
> >
> > yes, as bios option.
> >
> > On my point of view it would be interesting to verify is hyperthreading is
> > really usefull or not.
> >
>
> It would be interesting to determine if "hyperthreading" in the CPU
> actually exists. It may just be an artifact of dual instruction units,
> actually a defect (perhaps harmless), that is hyped as a feature.
Clearly not, it requires another set of registers including instruction
counter.
> For instance, it has long been known that if a CPU were to have as
> many instruction units as possible instruction branches, program
> jumps upon logical conditions would not slow the machine down. The
> hardware just continues using the instruction unit that contains the
> correct program-flow while the others are re-loaded.
That is not correct, it certainly can slow the machine down. Speculative
execution is used, but it's not free, since it requires fetching
instructions not used through a limited bandwidth to memory. Much on this
in comp.arch, there is a tradeoff between avoiding stalls and causing
them, separate branch target cache, etc. Details probably a lot better to
be discussed there.
--
bill davidsen <[email protected]>
CTO, TMR Associates, Inc
Doing interesting things with little computers since 1979.