PP-DLW Dual Xeon 2.8 Unable to Disable Hyperthreading?

[Guest]

Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

I am running PP-DLW with Dual Xeon 2.8GHz 533FSB processors.

In a nutshell, when I go into the BIOS and disable hyperthreading, the
system fails to boot. Is this normal?

Here is why I want to disable hyperthreading:

I am running into an issue where one of my encoding applications seems
to think that there are only 2 logical processors instead of 2 actual
processors.

Windows device manager correctly sees 4 logical cpu's.

The problem is that the encoding app is only reaching 50% processor
utilization according to Windows XP task manager.

I suspect that the encoding app is only using the first processor in
hyperthreading mode.

I further substantiate there being a problem here because a P4-1.5GHz
can accomplish this encode in 14 hours.

I even further yet substantiate there being a problem here because a
Dual Intel P3-1.4GHz can do the job in 9.5 hours.

The Dual Xeon system is currently only able to accomplish it in 7.5
hours. A time that loosely corresponds to ONE Xeon 2.8GHz processor.

Any ideas anybody?
I built this system for encoding and it now seems I am going to be
severely limited in the applications that will support what I have,
unless I can turn off hyperthreading.


Any comments or suggestions would be greatly appreciated.

Best Regards,
Charles W.

 

[Tim]

Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

Hi,

Fails to boot? Where & when? If it is during XP startup at some early point,
do you get a BSOD? If so what does it say? At a guess, I would suspect that
a special HAL or registry setting kicks in when HT is enabled and you do the
install with HT on. So, try a repair install. If you do a repair, don't
forget to re-install SP1 if its in, the mobo drivers and all updates esp.
critical updates. I vaguely recall that HT became supported with SP1
anyways...

I agree in principle with your conclusion that HT is likely to be the cause
for the slow down. Code that is highly optimised for conventional
architecture X86 (IE non HT) CPU's can suffer with HT enabled. Such code is
not too common, but an encoding application is a likely candidate for such
software tuning. It has been reported for example in SQL Server news groups
that it suffers with HT enabled for this reason.

Effectively HT comes to the fore when either a) the code is optimised in a
more conventional manner giving slight to moderate benefit E.G. normal
programs or b) when the code has been hand optimised using the Intel HT
extensions and also compiled using the Intel C++ compiler (or FORTRAN) -
most benefitial, cited as around 35% improvement. There are no doubt other
compilers that will do this too. There are droves of programming techniques
that can be used to create code that is HT friendly - the reason for the
popularity of the Intel compilers as they do this automatically when
generating code.

So, either try a repair, or post back with the BSOD STOP details. Don't be
shy about turning HT off and doing a bench mark. If you succeed, please let
us know how you get on. Failing all that, you could check to see if there is
an update for your encoder that is compiled for HT enabled systems.

- Tim


"2boxers" <2boxers_at_comcast_dot_net> wrote in message
news:afm3905c1badhscd41o944anlinlilpgbr@4ax.com...
> I am running PP-DLW with Dual Xeon 2.8GHz 533FSB processors.
>
> In a nutshell, when I go into the BIOS and disable hyperthreading, the
> system fails to boot. Is this normal?
>
> Here is why I want to disable hyperthreading:
>
> I am running into an issue where one of my encoding applications seems
> to think that there are only 2 logical processors instead of 2 actual
> processors.
>
> Windows device manager correctly sees 4 logical cpu's.
>
> The problem is that the encoding app is only reaching 50% processor
> utilization according to Windows XP task manager.
>
> I suspect that the encoding app is only using the first processor in
> hyperthreading mode.
>
> I further substantiate there being a problem here because a P4-1.5GHz
> can accomplish this encode in 14 hours.
>
> I even further yet substantiate there being a problem here because a
> Dual Intel P3-1.4GHz can do the job in 9.5 hours.
>
> The Dual Xeon system is currently only able to accomplish it in 7.5
> hours. A time that loosely corresponds to ONE Xeon 2.8GHz processor.
>
> Any ideas anybody?
> I built this system for encoding and it now seems I am going to be
> severely limited in the applications that will support what I have,
> unless I can turn off hyperthreading.
>
>
> Any comments or suggestions would be greatly appreciated.
>
> Best Regards,
> Charles W.
>
>

 

[Guest]

Archived from groups: alt.comp.periphs.mainboard.asus (More info?)

On Fri, 30 Apr 2004 21:08:58 +1200, "Tim" <Tim@NoSpam.com> wrote:

>Hi,
>
>Fails to boot? Where & when?
Fails to boot as in fails to initialize video after powering up the
system.

I have to clear the CMOS memory just to get the system to come back
up.

>. Code that is highly optimised for conventional
>architecture X86 (IE non HT) CPU's can suffer with HT enabled. Such code is
>not too common,
The encoder I am using, TMPGEnc sees 2 logical processors instead of
4. It is capable of supporting hyperthreading and it is capable of
supporting SMP, but apparently it is failing to do both at the same
time as with the case of Dual Xeon.

Fortunately, Windows XP does a good job with balancing the load for
multiple applications. When I run 2 instances of the encoder, I am
using both Xeons fully. i.e. 100% cpu utilization.

>shy about turning HT off and doing a bench mark.
Im not shy, it is just that it isnt working

>us know how you get on. Failing all that, you could check to see if there is
>an update for your encoder that is compiled for HT enabled systems.
I am writing an email to the author today.

Thank you for the reply,
Charles