Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
But it also "doesn't" happen to some apparently (not sure about windows
though), so would be useful to try and at least isolate if there's any
difference why some get the issue and others don't. I.e could just be the
latest kernel, or some adjustment that prevents the issue, but is evident in
all other cases.

On 10/14/05, Gary <[EMAIL PROTECTED]> wrote:
>
> Not an OS issue, happens on both *NIX and windows variants.
>
> At 02:38 PM 10/13/2005, ScratchMonkey wrote:
> >--On Thursday, October 13, 2005 10:37 AM +0100 Ian mu <
> [EMAIL PROTECTED]>
> >wrote:
> >
> >>Is it possible for those with the problem to confirm what dist and
> kernel
> >>version on any machines that have the problem, just may help give some
> >>ideas as well?
> >
> >Version is sufficient *if* the kernel is stock for that distro. Otherwise
> >we need to know precise customizations, ie. what you changed in the
> config.
> >
> >
> >___
> >To unsubscribe, edit your list preferences, or view the list
> >archives, please visit:
> >http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
>
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Gary]

Not an OS issue, happens on both *NIX and windows variants.

At 02:38 PM 10/13/2005, ScratchMonkey wrote:

--On Thursday, October 13, 2005 10:37 AM +0100 Ian mu <[EMAIL PROTECTED]>
wrote:


Is it possible for those with the problem to confirm what dist and kernel
version on any machines that have the problem, just may help give some
ideas as well?


Version is sufficient *if* the kernel is stock for that distro. Otherwise
we need to know precise customizations, ie. what you changed in the config.


___
To unsubscribe, edit your list preferences, or view the list
archives, please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[ScratchMonkey]

--On Thursday, October 13, 2005 10:37 AM +0100 Ian mu <[EMAIL PROTECTED]>
wrote:


Is it possible for those with the problem to confirm what dist and kernel
version on any machines that have the problem, just may help give some
ideas as well?


Version is sufficient *if* the kernel is stock for that distro. Otherwise
we need to know precise customizations, ie. what you changed in the config.


___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
Is it possible for those with the problem to confirm what dist and kernel
version on any machines that have the problem, just may help give some ideas
as well?

On 10/13/05, e-Plutonia <[EMAIL PROTECTED]> wrote:
>
> --
> [ Picked text/plain from multipart/alternative ]
> Since I am not on the Intel platform anymore, I can't help out directly,
> however Gary, if you want, I can bring a few of my old Dual Xeon 3.2 / 3.0/
> 2.8 boxes back online, as well as some P4s.
>
> On 10/12/05, Gary <[EMAIL PROTECTED]> wrote:
> >
> > I'm beginning to wonder if certain stepping Xeon's/P4's have this HT
> > problem. I will investigate this further, but I don't think that is
> > the case since I have multiple servers with different stepping xeons
> > and they exhibit the same issues with large servers.. Perhaps it's
> > cache coherency problems, or one of the many pitfalls of hyperthreading.
> >
> >
> > >If they are both still present you can diff the configurations.
> > >
> > >More to the point though, a stock kernel may not support HT, whereas
> > >unless moronically configured, a newly compiled kernel should use most
> > >of the target architecture, instead of a generic architecture.
> > >
> > >Next, the issue of HT as an architecture. Many of the situations where
> > >HT has been reported to perform badly are in cases where the server is
> > >processing two or more SRCDS instances throughout 80% or above of it's
> > >active processing time. In this scenario you will be suffering a great
> > >deal of cache hits, and a fair few (but quite regularly timed) context
> > >switches. What is interesting to note about those who claim better
> > >results, is they tend to have servers loaded up with other/different
> > >applications. Careful analysis of other performance data might lead to
> > >a real conclusion regarding the cause of this issue. How many cache
> > >misses are occuring with HT on, as opposed to off? What about context
> > >switches? How do these change if you lock processes to specific
> > >processors? The list of questions goes on. Sadly, I'm not using any HT
> > >capable test hardware at present.
> >
> >
> >
> > ___
> > To unsubscribe, edit your list preferences, or view the list archives,
> > please visit:
> > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> >
>
>
>
> --
> DJ Fadyeyev
> Founder
> e-Plutonia
> --
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[e-Plutonia]

--
[ Picked text/plain from multipart/alternative ]
Since I am not on the Intel platform anymore, I can't help out directly,
however Gary, if you want, I can bring a few of my old Dual Xeon 3.2 / 3.0 /
2.8 boxes back online, as well as some P4s.

On 10/12/05, Gary <[EMAIL PROTECTED]> wrote:
>
> I'm beginning to wonder if certain stepping Xeon's/P4's have this HT
> problem. I will investigate this further, but I don't think that is
> the case since I have multiple servers with different stepping xeons
> and they exhibit the same issues with large servers.. Perhaps it's
> cache coherency problems, or one of the many pitfalls of hyperthreading.
>
>
> >If they are both still present you can diff the configurations.
> >
> >More to the point though, a stock kernel may not support HT, whereas
> >unless moronically configured, a newly compiled kernel should use most
> >of the target architecture, instead of a generic architecture.
> >
> >Next, the issue of HT as an architecture. Many of the situations where
> >HT has been reported to perform badly are in cases where the server is
> >processing two or more SRCDS instances throughout 80% or above of it's
> >active processing time. In this scenario you will be suffering a great
> >deal of cache hits, and a fair few (but quite regularly timed) context
> >switches. What is interesting to note about those who claim better
> >results, is they tend to have servers loaded up with other/different
> >applications. Careful analysis of other performance data might lead to
> >a real conclusion regarding the cause of this issue. How many cache
> >misses are occuring with HT on, as opposed to off? What about context
> >switches? How do these change if you lock processes to specific
> >processors? The list of questions goes on. Sadly, I'm not using any HT
> >capable test hardware at present.
>
>
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>



--
DJ Fadyeyev
Founder
e-Plutonia
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Gary]

I'm beginning to wonder if certain stepping Xeon's/P4's have this HT
problem. I will investigate this further, but I don't think that is
the case since I have multiple servers with different stepping xeons
and they exhibit the same issues with large servers.. Perhaps it's
cache coherency problems, or one of the many pitfalls of hyperthreading.



If they are both still present you can diff the configurations.

More to the point though, a stock kernel may not support HT, whereas
unless moronically configured, a newly compiled kernel should use most
of the target architecture, instead of a generic architecture.

Next, the issue of HT as an architecture. Many of the situations where
HT has been reported to perform badly are in cases where the server is
processing two or more SRCDS instances throughout 80% or above of it's
active processing time. In this scenario you will be suffering a great
deal of cache hits, and a fair few (but quite regularly timed) context
switches. What is interesting to note about those who claim better
results, is they tend to have servers loaded up with other/different
applications. Careful analysis of other performance data might lead to
a real conclusion regarding the cause of this issue. How many cache
misses are occuring with HT on, as opposed to off? What about context
switches? How do these change if you lock processes to specific
processors? The list of questions goes on. Sadly, I'm not using any HT
capable test hardware at present.




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[James Tucker]

On 10/13/05, ScratchMonkey <[EMAIL PROTECTED]> wrote:
> --On Wednesday, October 12, 2005 9:50 AM -0400 Mahmoud Foda
> <[EMAIL PROTECTED]> wrote:

> To know why your system works better, we'd need to know exactly what the
> "standard" kernel was (presumably you don't mean an unadorned kernel with
> default configuration direct from Linus' build tree) and what you did to
> compile your own.

If they are both still present you can diff the configurations.

More to the point though, a stock kernel may not support HT, whereas
unless moronically configured, a newly compiled kernel should use most
of the target architecture, instead of a generic architecture.

Next, the issue of HT as an architecture. Many of the situations where
HT has been reported to perform badly are in cases where the server is
processing two or more SRCDS instances throughout 80% or above of it's
active processing time. In this scenario you will be suffering a great
deal of cache hits, and a fair few (but quite regularly timed) context
switches. What is interesting to note about those who claim better
results, is they tend to have servers loaded up with other/different
applications. Careful analysis of other performance data might lead to
a real conclusion regarding the cause of this issue. How many cache
misses are occuring with HT on, as opposed to off? What about context
switches? How do these change if you lock processes to specific
processors? The list of questions goes on. Sadly, I'm not using any HT
capable test hardware at present.

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[ScratchMonkey]

--On Wednesday, October 12, 2005 9:50 AM -0400 Mahmoud Foda
<[EMAIL PROTECTED]> wrote:


Ohh wow guys. I think i know why HT works better on my servers?
 Compile your linux kernals on your servers. Dont use the kernal that
comes with your distro.

I tried some expirements and HLDS doesn't run good at all on a standard
kernal running HT. If i you compile it you get just better results.


The problem is that we don't know why your custom compile works any better.
A "standard" Fedora kernel is compiled on a Fedora system so there's no
reason a custom kernel would be any better, unless you've chosen a
different set of patches or config options.

To know why your system works better, we'd need to know exactly what the
"standard" kernel was (presumably you don't mean an unadorned kernel with
default configuration direct from Linus' build tree) and what you did to
compile your own.


___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[ScratchMonkey]

--On Wednesday, October 12, 2005 3:59 PM +0200 Nic Strix
<[EMAIL PROTECTED]> wrote:


thanks for the advice. now to figure out how to do that..


For Red Hat systems (including Fedora and Centos), download and install the
kernel source RPM (from the SRPMS directory). This puts the source in
/usr/src/redhat. chown that directory hierarchy to your favorite mortal,
and rebuild the RPM from there:

chown -R scratch.scratch /usr/src/redhat
su scratch
rpmbuild -ba /usr/src/redhat/SPECS/kernel.spec



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

RE: [hlds_linux] Hyperthreading and DoD:Source

[Nic Strix]

HT is a hardware function = hyperthreading, and your proccessor needs to
support it. You can only properly turn it on and off via the BIOS. It is
also possible to enable/disable HT in linux but it is advised to do it via
the BIOS and seeing as both methods require a restart I would go for the
BIOS method. Check your motherboard's manual to see how to access the BIOS
and to enable/disable HT if yor system supports it

Nic



From: "Vapok" <[EMAIL PROTECTED]>
Reply-To: hlds_linux@list.valvesoftware.com
To: 
Subject: RE: [hlds_linux] Hyperthreading and DoD:Source
Date: Wed, 12 Oct 2005 09:10:12 -0500

Noob question:

How do you turn HT on in DoD: Source?

Thanks



___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

RE: [hlds_linux] Hyperthreading and DoD:Source

[Vapok]

Noob question:

How do you turn HT on in DoD: Source?

Thanks



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Nic Strix]

thanks for the advice. now to figure out how to do that..



From: Mahmoud Foda <[EMAIL PROTECTED]>
Reply-To: hlds_linux@list.valvesoftware.com
To: hlds_linux@list.valvesoftware.com
Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
Date: Wed, 12 Oct 2005 09:50:34 -0400
[ Picked text/plain from multipart/alternative ]
Ohh wow guys. I think i know why HT works better on my servers?
 Compile your linux kernals on your servers. Dont use the kernal that
comes
with your distro.

I tried some expirements and HLDS doesn't run good at all on a standard
kernal running HT. If i you compile it you get just better results.

 On 10/10/05, Ian mu <[EMAIL PROTECTED]> wrote:
>
> --
> [ Picked text/plain from multipart/alternative ]
> I think this is why its good to get to the bottom of where the problem
is,
> genuinely happy that the problem has gone, just that its evident some
> people
> don't have the problem and some do. Some have been helped by setting
> affinity on servers (we do on a couple of larger ones). I guess
personally
> I'm just a fan of if you pay for something (like ht) in a cpu, you
should
> be
> able to make the best of it, and its worth getting to the bottom of
either
> making the most of it, or fully understanding why you can't if others
can.
> One thing that does seem to make a difference is kernel version (I
think),
> certainly some older ones report cpu usage very differently, and
wondering
> if the scheduler acts differently on it. No expert in that field, just
> something I've wondered as 2 linux servers we have we treat very
> differently
> to the others, so might be interesting for people to quote what linux
dist
> and kernel version they run just to see if there's anything in that or
> not.
>
>
> On 10/10/05, Mahmoud Foda <[EMAIL PROTECTED]> wrote:
> >
> > --
> > [ Picked text/plain from multipart/alternative ]
> > wow. Dude we have HT on our server. It performs better then it off. I
> dont
> > know why i'm different. But HT has helped a lot. I can't imagine going
> > less.
> > 3.2GHz P4 w/ HT. 1 26slot DOD server. 1 18 Slot dod server. 1 20 slot
> dod
> > server. 1 16 slot HLDM server. 1 16slot CS:S. 1 26slot DOD:S server.
> > We normally have about 80 slots full. NO LAG!!! Server CPU Usage is
80%
> > constant. Servers have either Mani or AMX. With lots of plugins. I
dont
> > know
> > why i'm different. And i dont know why people say HT off is better.
When
> i
> > turn HT off, i see lag everywhere. Its unbarible to play. But with HT
on
> > its
> > like smooth and a dream.
> >
> > On 10/8/05, Nic Strix <[EMAIL PROTECTED]> wrote:
> > >
> > > HT turned off and now dod:s runs on 32 with minimum to no lag!
Thanks
> > all.
> > >
> > >
> > >
> > > >From: Gary <[EMAIL PROTECTED]>
> > > >Reply-To: hlds_linux@list.valvesoftware.com
> > > >To: hlds_linux@list.valvesoftware.com,
> hlds_linux@list.valvesoftware.com
> > > >Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
> > > >Date: Thu, 06 Oct 2005 14:33:58 -0400
> > > >
> > > >This topic has nothing to do with what are your thoughts on HT...
> > > >Besides, do those benchmarks happen to include game servers? valves
> > > engine?
> > > >
> > > >Try and generate the same problem. Load a >26 slot server similar
to
> > > >the original posters configuration. I can generate similar results
on
> > > >a couple of non production machines, disabling HT does fix the
> > > >problem (it's been verified by others on this list before.)
> > > >
> > > >/ges
> > > >
> > > >At 02:05 PM 10/6/2005, Ian mu wrote:
> > > >>--
> > > >>[ Picked text/plain from multipart/alternative ]
> > > >>For those who think HT is the problem, can they explain why
> benchmarks
> > > >>show
> > > >>most none-multithreaded apps run faster (albeit only slightly) ?
> Don't
> > > >>really want to turn this into a ht argument, but I'd like to see
> some
> > > >>supporting evidence of the various criticisms of ht we see so
often.
> > > >> As an example...from here (has single threaded benchmarks with ht
> > > >>enabled)
> > > >>
> http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
> > > >>Quote:
> > > >> >>
> > > >
> > > >
> > > >
> > > >_

Re: [hlds_linux] Hyperthreading and DoD:Source

[Mahmoud Foda]

--
[ Picked text/plain from multipart/alternative ]
Ohh wow guys. I think i know why HT works better on my servers?
 Compile your linux kernals on your servers. Dont use the kernal that comes
with your distro.

I tried some expirements and HLDS doesn't run good at all on a standard
kernal running HT. If i you compile it you get just better results.

 On 10/10/05, Ian mu <[EMAIL PROTECTED]> wrote:
>
> --
> [ Picked text/plain from multipart/alternative ]
> I think this is why its good to get to the bottom of where the problem is,
> genuinely happy that the problem has gone, just that its evident some
> people
> don't have the problem and some do. Some have been helped by setting
> affinity on servers (we do on a couple of larger ones). I guess personally
> I'm just a fan of if you pay for something (like ht) in a cpu, you should
> be
> able to make the best of it, and its worth getting to the bottom of either
> making the most of it, or fully understanding why you can't if others can.
> One thing that does seem to make a difference is kernel version (I think),
> certainly some older ones report cpu usage very differently, and wondering
> if the scheduler acts differently on it. No expert in that field, just
> something I've wondered as 2 linux servers we have we treat very
> differently
> to the others, so might be interesting for people to quote what linux dist
> and kernel version they run just to see if there's anything in that or
> not.
>
>
> On 10/10/05, Mahmoud Foda <[EMAIL PROTECTED]> wrote:
> >
> > --
> > [ Picked text/plain from multipart/alternative ]
> > wow. Dude we have HT on our server. It performs better then it off. I
> dont
> > know why i'm different. But HT has helped a lot. I can't imagine going
> > less.
> > 3.2GHz P4 w/ HT. 1 26slot DOD server. 1 18 Slot dod server. 1 20 slot
> dod
> > server. 1 16 slot HLDM server. 1 16slot CS:S. 1 26slot DOD:S server.
> > We normally have about 80 slots full. NO LAG!!! Server CPU Usage is 80%
> > constant. Servers have either Mani or AMX. With lots of plugins. I dont
> > know
> > why i'm different. And i dont know why people say HT off is better. When
> i
> > turn HT off, i see lag everywhere. Its unbarible to play. But with HT on
> > its
> > like smooth and a dream.
> >
> > On 10/8/05, Nic Strix <[EMAIL PROTECTED]> wrote:
> > >
> > > HT turned off and now dod:s runs on 32 with minimum to no lag! Thanks
> > all.
> > >
> > >
> > >
> > > >From: Gary <[EMAIL PROTECTED]>
> > > >Reply-To: hlds_linux@list.valvesoftware.com
> > > >To: hlds_linux@list.valvesoftware.com,
> hlds_linux@list.valvesoftware.com
> > > >Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
> > > >Date: Thu, 06 Oct 2005 14:33:58 -0400
> > > >
> > > >This topic has nothing to do with what are your thoughts on HT...
> > > >Besides, do those benchmarks happen to include game servers? valves
> > > engine?
> > > >
> > > >Try and generate the same problem. Load a >26 slot server similar to
> > > >the original posters configuration. I can generate similar results on
> > > >a couple of non production machines, disabling HT does fix the
> > > >problem (it's been verified by others on this list before.)
> > > >
> > > >/ges
> > > >
> > > >At 02:05 PM 10/6/2005, Ian mu wrote:
> > > >>--
> > > >>[ Picked text/plain from multipart/alternative ]
> > > >>For those who think HT is the problem, can they explain why
> benchmarks
> > > >>show
> > > >>most none-multithreaded apps run faster (albeit only slightly) ?
> Don't
> > > >>really want to turn this into a ht argument, but I'd like to see
> some
> > > >>supporting evidence of the various criticisms of ht we see so often.
> > > >> As an example...from here (has single threaded benchmarks with ht
> > > >>enabled)
> > > >>
> http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
> > > >>Quote:
> > > >> >>
> > > >
> > > >
> > > >
> > > >___
> > > >To unsubscribe, edit your list preferences, or view the list
> archives,
> > > >please visit:
> > > >http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> > >
> > >
> > >
> > > ___
> > > To unsubscribe, edit your list preferences, or view the list archives,
> > > please visit:
> > > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> > >
> > --
> >
> > ___
> > To unsubscribe, edit your list preferences, or view the list archives,
> > please visit:
> > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> >
> --
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
I think this is why its good to get to the bottom of where the problem is,
genuinely happy that the problem has gone, just that its evident some people
don't have the problem and some do. Some have been helped by setting
affinity on servers (we do on a couple of larger ones). I guess personally
I'm just a fan of if you pay for something (like ht) in a cpu, you should be
able to make the best of it, and its worth getting to the bottom of either
making the most of it, or fully understanding why you can't if others can.
 One thing that does seem to make a difference is kernel version (I think),
certainly some older ones report cpu usage very differently, and wondering
if the scheduler acts differently on it. No expert in that field, just
something I've wondered as 2 linux servers we have we treat very differently
to the others, so might be interesting for people to quote what linux dist
and kernel version they run just to see if there's anything in that or not.


 On 10/10/05, Mahmoud Foda <[EMAIL PROTECTED]> wrote:
>
> --
> [ Picked text/plain from multipart/alternative ]
> wow. Dude we have HT on our server. It performs better then it off. I dont
> know why i'm different. But HT has helped a lot. I can't imagine going
> less.
> 3.2GHz P4 w/ HT. 1 26slot DOD server. 1 18 Slot dod server. 1 20 slot dod
> server. 1 16 slot HLDM server. 1 16slot CS:S. 1 26slot DOD:S server.
> We normally have about 80 slots full. NO LAG!!! Server CPU Usage is 80%
> constant. Servers have either Mani or AMX. With lots of plugins. I dont
> know
> why i'm different. And i dont know why people say HT off is better. When i
> turn HT off, i see lag everywhere. Its unbarible to play. But with HT on
> its
> like smooth and a dream.
>
> On 10/8/05, Nic Strix <[EMAIL PROTECTED]> wrote:
> >
> > HT turned off and now dod:s runs on 32 with minimum to no lag! Thanks
> all.
> >
> >
> >
> > >From: Gary <[EMAIL PROTECTED]>
> > >Reply-To: hlds_linux@list.valvesoftware.com
> > >To: hlds_linux@list.valvesoftware.com,hlds_linux@list.valvesoftware.com
> > >Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
> > >Date: Thu, 06 Oct 2005 14:33:58 -0400
> > >
> > >This topic has nothing to do with what are your thoughts on HT...
> > >Besides, do those benchmarks happen to include game servers? valves
> > engine?
> > >
> > >Try and generate the same problem. Load a >26 slot server similar to
> > >the original posters configuration. I can generate similar results on
> > >a couple of non production machines, disabling HT does fix the
> > >problem (it's been verified by others on this list before.)
> > >
> > >/ges
> > >
> > >At 02:05 PM 10/6/2005, Ian mu wrote:
> > >>--
> > >>[ Picked text/plain from multipart/alternative ]
> > >>For those who think HT is the problem, can they explain why benchmarks
> > >>show
> > >>most none-multithreaded apps run faster (albeit only slightly) ? Don't
> > >>really want to turn this into a ht argument, but I'd like to see some
> > >>supporting evidence of the various criticisms of ht we see so often.
> > >> As an example...from here (has single threaded benchmarks with ht
> > >>enabled)
> > >>http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
> > >>Quote:
> > >> >>
> > >
> > >
> > >
> > >___
> > >To unsubscribe, edit your list preferences, or view the list archives,
> > >please visit:
> > >http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> >
> >
> >
> > ___
> > To unsubscribe, edit your list preferences, or view the list archives,
> > please visit:
> > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> >
> --
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Mahmoud Foda]

--
[ Picked text/plain from multipart/alternative ]
wow. Dude we have HT on our server. It performs better then it off. I dont
know why i'm different. But HT has helped a lot. I can't imagine going less.
3.2GHz P4 w/ HT. 1 26slot DOD server. 1 18 Slot dod server. 1 20 slot dod
server. 1 16 slot HLDM server. 1 16slot CS:S. 1 26slot DOD:S server.
 We normally have about 80 slots full. NO LAG!!! Server CPU Usage is 80%
constant. Servers have either Mani or AMX. With lots of plugins. I dont know
why i'm different. And i dont know why people say HT off is better. When i
turn HT off, i see lag everywhere. Its unbarible to play. But with HT on its
like smooth and a dream.

 On 10/8/05, Nic Strix <[EMAIL PROTECTED]> wrote:
>
> HT turned off and now dod:s runs on 32 with minimum to no lag! Thanks all.
>
>
>
> >From: Gary <[EMAIL PROTECTED]>
> >Reply-To: hlds_linux@list.valvesoftware.com
> >To: hlds_linux@list.valvesoftware.com,hlds_linux@list.valvesoftware.com
> >Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
> >Date: Thu, 06 Oct 2005 14:33:58 -0400
> >
> >This topic has nothing to do with what are your thoughts on HT...
> >Besides, do those benchmarks happen to include game servers? valves
> engine?
> >
> >Try and generate the same problem. Load a >26 slot server similar to
> >the original posters configuration. I can generate similar results on
> >a couple of non production machines, disabling HT does fix the
> >problem (it's been verified by others on this list before.)
> >
> >/ges
> >
> >At 02:05 PM 10/6/2005, Ian mu wrote:
> >>--
> >>[ Picked text/plain from multipart/alternative ]
> >>For those who think HT is the problem, can they explain why benchmarks
> >>show
> >>most none-multithreaded apps run faster (albeit only slightly) ? Don't
> >>really want to turn this into a ht argument, but I'd like to see some
> >>supporting evidence of the various criticisms of ht we see so often.
> >> As an example...from here (has single threaded benchmarks with ht
> >>enabled)
> >>http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
> >>Quote:
> >> >>
> >
> >
> >
> >___
> >To unsubscribe, edit your list preferences, or view the list archives,
> >please visit:
> >http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
>
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Nic Strix]

HT turned off and now dod:s runs on 32 with minimum to no lag! Thanks all.




From: Gary <[EMAIL PROTECTED]>
Reply-To: hlds_linux@list.valvesoftware.com
To: hlds_linux@list.valvesoftware.com,hlds_linux@list.valvesoftware.com
Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
Date: Thu, 06 Oct 2005 14:33:58 -0400

This topic has nothing to do with what are your thoughts on HT...
Besides, do those benchmarks happen to include game servers? valves engine?

Try and generate the same problem. Load a >26 slot server similar to
the original posters configuration. I can generate similar results on
a couple of non production machines, disabling HT does fix the
problem (it's been verified by others on this list before.)

/ges

At 02:05 PM 10/6/2005, Ian mu wrote:

--
[ Picked text/plain from multipart/alternative ]
For those who think HT is the problem, can they explain why benchmarks
show
most none-multithreaded apps run faster (albeit only slightly) ? Don't
really want to turn this into a ht argument, but I'd like to see some
supporting evidence of the various criticisms of ht we see so often.
 As an example...from here (has single threaded benchmarks with ht
enabled)
http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
Quote:
>>




___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Marcelo Bezerra]

Hope it was easy.
Gave up complaining it a long, long time ago... :(

On Fri, 2005-10-07 at 12:11 +0100, Ben Kennish wrote:
> Hi,
>
> I know this sounds rather picky of me but PLEASE can people not start a
> new topic by replying to an existing one and then changing the subject
> and message text.
>
> Cos your email client sends out a header saying that it's in reference
> to the e-mail you replied to and when I view messages in "threaded
> view", the new thread gets tacked on to the end of the old one.
>
> Cheers,
>
> Ben
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives, please 
> visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux


___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ben Kennish]

Hi,

I know this sounds rather picky of me but PLEASE can people not start a
new topic by replying to an existing one and then changing the subject
and message text.

Cos your email client sends out a header saying that it's in reference
to the e-mail you replied to and when I view messages in "threaded
view", the new thread gets tacked on to the end of the old one.

Cheers,

Ben

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[William Warren]

HT is p-4 intel only..if you are running AMD you aren't using HT.

Deacon @[dgx] wrote:

From what I've seen, and other can probably comment in more detail,

hyperthreading can lag a server and we've seen many posts to the list
asking about this.  I would recommend searching the list archieve.

That being said, your lag can also be caused by the server config itself.
We have serveral servers that support and are running hyperthreading, but
most are AMD processors so I might be way off, but here's what I've seen.

I have several machines that run 18-24 players w/o lag.  If I take the
same config and change the maxplayer value to 26-32 people complain of
lag, choke, etc.

So, I would recommend tweaking your config for 32 players (sv_maxrate,
updaterate, tic_rate/tickrate, pingboost, etc.) prior to disabling
hyperthreading support.  Feel free to email me at [EMAIL PROTECTED] for info
on DOD:S and DOD server configs or just reply here.

Also, check for our DOD:S threads from about a week ago.  A few of us
posted our DOD:S configs and startup options.

Thanks,
Deacon


On Thu, 6 Oct 2005, Nic Strix wrote:



Hi

I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon box
(dell poweredge sc1425)  since shortly after the game was released. While
running a default server with 32 slots the server lags out with one of the
virtual cpu's maxing out. It seems to run fine with 24 slots. I read
somewhere that this is caused by the hyperthreading on the box and the only
way to solve this is to turn hyperthreading off.

My question is - is this true? Will I gain a 32 slot (hopefully many more
slots) without the cpu maxing by turning hyperthreading off? Or is there a
way to make the OS do this? Or is there a fix on its way? I dont live close
to the box (amsterdam--london) so a quick bus journey to play with the BIOS
is not really possible.

Any and all help will be highly appreciated! This mailinglist has already
helped me out a lot and I hope it will continue!!

TIA
Nic



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux





___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux



--
My "Foundation" verse:
Isa 54:17  No weapon that is formed against thee shall prosper; and
every tongue that shall rise against thee in judgment thou shalt
condemn. This is the heritage of the servants of the LORD, and their
righteousness is of me, saith the LORD.

-- carpe ductum -- "Grab the tape"
CDTT (Certified Duct Tape Technician)

Linux user #322099
Machines:
206822
256638
276825
http://counter.li.org/

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[ScratchMonkey]

--On Thursday, October 06, 2005 2:33 PM -0400 Gary
<[EMAIL PROTECTED]> wrote:


Besides, do those benchmarks happen to include game servers? valves
engine?


I used to run a couple of Tribes Vengeance servers on my dual Xeon and got
abysmal performance when they filled until I locked each to a single
logical CPU. Note that you can do this on a per-process basis and don't
have to deny all other processes the advantages of SMP. Use the taskset
command to launch a process limited to a bitmap of CPU's it's allowed to
use.

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
Hiya, not going to waffle any more on ht debate as will just clog up the
list...
 Just remember that depending on what you are specifically looking at ht
will often report incorrect cpu usage (or as correct as it can know) because
there is simply no way to tell how much "spare" ht can be used for the 2nd
virtual cpu so, so you may in effect "see" 200% but not have it all
available as absurd as that may sound. It literally thinks you have 4 cpu's
to calculate how much is used, when in reality you have 2 + whatever bits ht
can clutch for you.
 It does seem to vary from kernel/top versions though as to how its
reported, and not entirely sure how "clever" recent versions are when it
comes to giving a decent idea of spare cpu. Main thing I would do is watch
fps through the stats command and watch to see if its dropping as a
comparison of actual gameserver performance each time rather than cpu %age.

 On 10/6/05, Nic Strix <[EMAIL PROTECTED]> wrote:
>
> Sorry I should have said!
> we have a dual xeon 2.8 (2 x cpu's) with 2 gig ram
> also we do run multiple servers and total cpu usage is around 50%. The 32
> dods maxes out 1/2 of 2.8 so by turning off HT i could gain full usage of
> 2.8 for the 32 player server (possibly 40 slots) and leave the other games
> in place. At least that is what I hope to achieve!
>
>
> >From: Ian mu <[EMAIL PROTECTED]>
> >Reply-To: hlds_linux@list.valvesoftware.com
> >To: hlds_linux@list.valvesoftware.com
> >Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
> >Date: Thu, 6 Oct 2005 20:12:59 +0100
> >
> >--
> >[ Picked text/plain from multipart/alternative ]
> >I know, no one has said they are multithreaded, multitasking though it
> can
> >help in (no not huge amounts but the question is more is it a problem,
> >rather than does it help loads). Now it could well be there is a specific
> >valve issue with some setups where ht causes problems for their servers
> >which by the sounds of it some seem to get (in which case valve should
> look
> >into it), but I haven't experienced it so its possible they may not do
> >also,
> >and as I said on a none ht server, a 32 slot dod:s server eats up cpu
> like
> >I
> >eat pies.
> > If the poster has a 2.4G dual xeon for example I would expect them to be
> >maxed out whatever and would say 24 slots is the way to go, if a 3.8 then
> I
> >would say yes something is amiss. I'm just a little surprised no one has
> >even asked what cpus he has before recommending he goes and starts
> messing
> >with the bios or kernel params etc.
> >
> >
> > On 10/6/05, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
> > >
> > > Game servers are not multithreaded. Therefore, they'll only run on one
> >of
> > > the virtual CPUs rather than both. It might help if you're running
> > > multiple gameservers but certainly not just one.
> > >
> > > ~dvander
> > >
> > >
> >--
> >
> >___
> >To unsubscribe, edit your list preferences, or view the list archives,
> >please visit:
> >http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
>
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Nic Strix]

Sorry I should have said!
we have a dual xeon 2.8 (2 x cpu's) with 2 gig ram
also we do run multiple servers and total cpu usage is around 50%. The 32
dods maxes out 1/2 of 2.8 so by turning off HT i could gain full usage of
2.8 for the 32 player server (possibly 40 slots) and leave the other games
in place. At least that is what I hope to achieve!



From: Ian mu <[EMAIL PROTECTED]>
Reply-To: hlds_linux@list.valvesoftware.com
To: hlds_linux@list.valvesoftware.com
Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
Date: Thu, 6 Oct 2005 20:12:59 +0100

--
[ Picked text/plain from multipart/alternative ]
I know, no one has said they are multithreaded, multitasking though it can
help in (no not huge amounts but the question is more is it a problem,
rather than does it help loads). Now it could well be there is a specific
valve issue with some setups where ht causes problems for their servers
which by the sounds of it some seem to get (in which case valve should look
into it), but I haven't experienced it so its possible they may not do
also,
and as I said on a none ht server, a 32 slot dod:s server eats up cpu like
I
eat pies.
 If the poster has a 2.4G dual xeon for example I would expect them to be
maxed out whatever and would say 24 slots is the way to go, if a 3.8 then I
would say yes something is amiss. I'm just a little surprised no one has
even asked what cpus he has before recommending he goes and starts messing
with the bios or kernel params etc.


 On 10/6/05, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
>
> Game servers are not multithreaded. Therefore, they'll only run on one
of
> the virtual CPUs rather than both. It might help if you're running
> multiple gameservers but certainly not just one.
>
> ~dvander
>
>
--

___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
I know, no one has said they are multithreaded, multitasking though it can
help in (no not huge amounts but the question is more is it a problem,
rather than does it help loads). Now it could well be there is a specific
valve issue with some setups where ht causes problems for their servers
which by the sounds of it some seem to get (in which case valve should look
into it), but I haven't experienced it so its possible they may not do also,
and as I said on a none ht server, a 32 slot dod:s server eats up cpu like I
eat pies.
 If the poster has a 2.4G dual xeon for example I would expect them to be
maxed out whatever and would say 24 slots is the way to go, if a 3.8 then I
would say yes something is amiss. I'm just a little surprised no one has
even asked what cpus he has before recommending he goes and starts messing
with the bios or kernel params etc.


 On 10/6/05, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote:
>
> Game servers are not multithreaded. Therefore, they'll only run on one of
> the virtual CPUs rather than both. It might help if you're running
> multiple gameservers but certainly not just one.
>
> ~dvander
>
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[dvander]

Game servers are not multithreaded.  Therefore, they'll only run on one of
the virtual CPUs rather than both.  It might help if you're running
multiple gameservers but certainly not just one.

   ~dvander

> --
> [ Picked text/plain from multipart/alternative ]
> For those who think HT is the problem, can they explain why benchmarks
> show
> most none-multithreaded apps run faster (albeit only slightly) ? Don't
> really want to turn this into a ht argument, but I'd like to see some
> supporting evidence of the various criticisms of ht we see so often.
>  As an example...from here (has single threaded benchmarks with ht
> enabled)
> http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
> Quote:
>>>
> We also wanted to examine if leaving HT enabled had any performance
> penalty
> with certain games or applications. This amounts to a concern that
> competition for CPU resources could actually lower the performance when
> compared to a standard 3.06 GHz Pentium
> 4with
> HT disabled. We tested all of our benchmark applications in both
> modes,
> and while the majority of the performance increases were nominal, all of
> the
> benchmarks performed better with HT enabled. This is big news, as the
> design
> obviously allows full CPU partitioning, and gives users 3.06 GHz of power
> for a demanding games, dual-threading for supported application, yet still
> allows multi-threading and multi-tasking advantages in standard business
> use.
>>>
>  On 10/6/05, Marcel <[EMAIL PROTECTED]> wrote:
>>
>> Hi,
>>
>> afaik you just need to put a "noht" to your kernel parameters to disable
>> hyperthreading. Another way would be to disable HT completely in your
>> kernel config.
>>
>> Google should have some information for this matter.
>>
>> - Marcel
>>
>>
>> Nic Strix schrieb:
>> > Hi
>> >
>> > I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon
>> box
>> > (dell poweredge sc1425) since shortly after the game was released.
>> While
>> > running a default server with 32 slots the server lags out with one of
>> the
>> > virtual cpu's maxing out. It seems to run fine with 24 slots. I read
>> > somewhere that this is caused by the hyperthreading on the box and the
>> only
>> > way to solve this is to turn hyperthreading off.
>> >
>> > My question is - is this true? Will I gain a 32 slot (hopefully many
>> more
>> > slots) without the cpu maxing by turning hyperthreading off? Or is
>> there
>> a
>> > way to make the OS do this? Or is there a fix on its way? I dont live
>> close
>> > to the box (amsterdam--london) so a quick bus journey to play with the
>> BIOS
>> > is not really possible.
>> >
>> > Any and all help will be highly appreciated! This mailinglist has
>> already
>> > helped me out a lot and I hope it will continue!!
>> >
>> > TIA
>> > Nic
>> >
>> >
>> >
>> > ___
>> > To unsubscribe, edit your list preferences, or view the list archives,
>> please visit:
>> > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>> >
>> >
>>
>> ___
>> To unsubscribe, edit your list preferences, or view the list archives,
>> please visit:
>> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>>
> --
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
A 32 slot dod:s server is pretty nigh on maxing out most cpu's ~ 3G area is
my point which no one seems to have bothered mentioning (and no one seems to
have asked what cpus are in the dell), HT or not. Mine were hitting ~ 90%
cpu usage on both ht and none ht servers when full (see earlier mails when
dod:s was released for cpu usage reported).

 On 10/6/05, Gary <[EMAIL PROTECTED]> wrote:
>
> This topic has nothing to do with what are your thoughts on HT...
> Besides, do those benchmarks happen to include game servers? valves
> engine?
>
> Try and generate the same problem. Load a >26 slot server similar to
> the original posters configuration. I can generate similar results on
> a couple of non production machines, disabling HT does fix the
> problem (it's been verified by others on this list before.)
>
> /ges
>
> At 02:05 PM 10/6/2005, Ian mu wrote:
> >--
> >[ Picked text/plain from multipart/alternative ]
> >For those who think HT is the problem, can they explain why benchmarks
> show
> >most none-multithreaded apps run faster (albeit only slightly) ? Don't
> >really want to turn this into a ht argument, but I'd like to see some
> >supporting evidence of the various criticisms of ht we see so often.
> > As an example...from here (has single threaded benchmarks with ht
> enabled)
> >http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
> >Quote:
> > >>
>
>
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Nic Strix]

Thanks gary,

Pretty much what I wanted to hear. Now only to figure out how to turn HT off
on my linux box (fedora 3 64+) without accessing the BIOS

best regards
Nic



From: Gary <[EMAIL PROTECTED]>
Reply-To: hlds_linux@list.valvesoftware.com ges
To: hlds_linux@list.valvesoftware.com,hlds_linux@list.valvesoftware.com
Subject: Re: [hlds_linux] Hyperthreading and DoD:Source
Date: Thu, 06 Oct 2005 14:33:58 -0400

This topic has nothing to do with what are your thoughts on HT...
Besides, do those benchmarks happen to include game servers? valves engine?

Try and generate the same problem. Load a >26 slot server similar to
the original posters configuration. I can generate similar results on
a couple of non production machines, disabling HT does fix the
problem (it's been verified by others on this list before.)

/ges

At 02:05 PM 10/6/2005, Ian mu wrote:

--
[ Picked text/plain from multipart/alternative ]
For those who think HT is the problem, can they explain why benchmarks
show
most none-multithreaded apps run faster (albeit only slightly) ? Don't
really want to turn this into a ht argument, but I'd like to see some
supporting evidence of the various criticisms of ht we see so often.
 As an example...from here (has single threaded benchmarks with ht
enabled)
http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
Quote:
>>




___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Gary]

This topic has nothing to do with what are your thoughts on HT...
Besides, do those benchmarks happen to include game servers? valves engine?

Try and generate the same problem. Load a >26 slot server similar to
the original posters configuration. I can generate similar results on
a couple of non production machines, disabling HT does fix the
problem (it's been verified by others on this list before.)

/ges

At 02:05 PM 10/6/2005, Ian mu wrote:

--
[ Picked text/plain from multipart/alternative ]
For those who think HT is the problem, can they explain why benchmarks show
most none-multithreaded apps run faster (albeit only slightly) ? Don't
really want to turn this into a ht argument, but I'd like to see some
supporting evidence of the various criticisms of ht we see so often.
 As an example...from here (has single threaded benchmarks with ht enabled)
http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
Quote:
>>




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

RE: [hlds_linux] Hyperthreading and DoD:Source

[JCHost.net - Support]

I think we all know that HT is one of the best things Intel has ever done as
a company.  All my servers use HT technology

-Original Message-
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Ian mu
Sent: Thursday, October 06, 2005 1:06 PM
To: hlds_linux@list.valvesoftware.com
Subject: Re: [hlds_linux] Hyperthreading and DoD:Source

--
[ Picked text/plain from multipart/alternative ]
For those who think HT is the problem, can they explain why benchmarks show
most none-multithreaded apps run faster (albeit only slightly) ? Don't
really want to turn this into a ht argument, but I'd like to see some
supporting evidence of the various criticisms of ht we see so often.
 As an example...from here (has single threaded benchmarks with ht enabled)
http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
Quote:
>>
We also wanted to examine if leaving HT enabled had any performance penalty
with certain games or applications. This amounts to a concern that
competition for CPU resources could actually lower the performance when
compared to a standard 3.06 GHz Pentium
4<http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__7#>wit
h
HT disabled. We tested all of our benchmark applications in both
modes,
and while the majority of the performance increases were nominal, all of the
benchmarks performed better with HT enabled. This is big news, as the design
obviously allows full CPU partitioning, and gives users 3.06 GHz of power
for a demanding games, dual-threading for supported application, yet still
allows multi-threading and multi-tasking advantages in standard business
use.
>>
 On 10/6/05, Marcel <[EMAIL PROTECTED]> wrote:
>
> Hi,
>
> afaik you just need to put a "noht" to your kernel parameters to disable
> hyperthreading. Another way would be to disable HT completely in your
> kernel config.
>
> Google should have some information for this matter.
>
> - Marcel
>
>
> Nic Strix schrieb:
> > Hi
> >
> > I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon
> box
> > (dell poweredge sc1425) since shortly after the game was released. While
> > running a default server with 32 slots the server lags out with one of
> the
> > virtual cpu's maxing out. It seems to run fine with 24 slots. I read
> > somewhere that this is caused by the hyperthreading on the box and the
> only
> > way to solve this is to turn hyperthreading off.
> >
> > My question is - is this true? Will I gain a 32 slot (hopefully many
> more
> > slots) without the cpu maxing by turning hyperthreading off? Or is there
> a
> > way to make the OS do this? Or is there a fix on its way? I dont live
> close
> > to the box (amsterdam--london) so a quick bus journey to play with the
> BIOS
> > is not really possible.
> >
> > Any and all help will be highly appreciated! This mailinglist has
> already
> > helped me out a lot and I hope it will continue!!
> >
> > TIA
> > Nic
> >
> >
> >
> > ___
> > To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> >
> >
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux


___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Ian mu]

--
[ Picked text/plain from multipart/alternative ]
For those who think HT is the problem, can they explain why benchmarks show
most none-multithreaded apps run faster (albeit only slightly) ? Don't
really want to turn this into a ht argument, but I'd like to see some
supporting evidence of the various criticisms of ht we see so often.
 As an example...from here (has single threaded benchmarks with ht enabled)
http://www.sharkyextreme.com/hardware/cpu/article.php/3261_1500631__1
Quote:
>>
We also wanted to examine if leaving HT enabled had any performance penalty
with certain games or applications. This amounts to a concern that
competition for CPU resources could actually lower the performance when
compared to a standard 3.06 GHz Pentium
4with
HT disabled. We tested all of our benchmark applications in both
modes,
and while the majority of the performance increases were nominal, all of the
benchmarks performed better with HT enabled. This is big news, as the design
obviously allows full CPU partitioning, and gives users 3.06 GHz of power
for a demanding games, dual-threading for supported application, yet still
allows multi-threading and multi-tasking advantages in standard business
use.
>>
 On 10/6/05, Marcel <[EMAIL PROTECTED]> wrote:
>
> Hi,
>
> afaik you just need to put a "noht" to your kernel parameters to disable
> hyperthreading. Another way would be to disable HT completely in your
> kernel config.
>
> Google should have some information for this matter.
>
> - Marcel
>
>
> Nic Strix schrieb:
> > Hi
> >
> > I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon
> box
> > (dell poweredge sc1425) since shortly after the game was released. While
> > running a default server with 32 slots the server lags out with one of
> the
> > virtual cpu's maxing out. It seems to run fine with 24 slots. I read
> > somewhere that this is caused by the hyperthreading on the box and the
> only
> > way to solve this is to turn hyperthreading off.
> >
> > My question is - is this true? Will I gain a 32 slot (hopefully many
> more
> > slots) without the cpu maxing by turning hyperthreading off? Or is there
> a
> > way to make the OS do this? Or is there a fix on its way? I dont live
> close
> > to the box (amsterdam--london) so a quick bus journey to play with the
> BIOS
> > is not really possible.
> >
> > Any and all help will be highly appreciated! This mailinglist has
> already
> > helped me out a lot and I hope it will continue!!
> >
> > TIA
> > Nic
> >
> >
> >
> > ___
> > To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> > http://list.valvesoftware.com/mailman/listinfo/hlds_linux
> >
> >
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives,
> please visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>
--

___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Marcel]

Hi,

afaik you just need to put a "noht" to your kernel parameters to disable
 hyperthreading. Another way would be to disable HT completely in your
kernel config.

Google should have some information for this matter.

- Marcel


Nic Strix schrieb:

Hi

I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon box
(dell poweredge sc1425)  since shortly after the game was released. While
running a default server with 32 slots the server lags out with one of the
virtual cpu's maxing out. It seems to run fine with 24 slots. I read
somewhere that this is caused by the hyperthreading on the box and the only
way to solve this is to turn hyperthreading off.

My question is - is this true? Will I gain a 32 slot (hopefully many more
slots) without the cpu maxing by turning hyperthreading off? Or is there a
way to make the OS do this? Or is there a fix on its way? I dont live close
to the box (amsterdam--london) so a quick bus journey to play with the BIOS
is not really possible.

Any and all help will be highly appreciated! This mailinglist has already
helped me out a lot and I hope it will continue!!

TIA
Nic



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux




___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[aXeR \(AmmoBOX\)]

I'm assuming that you haven't set the tickrate to 33 and your leaving it to
its own devices (which will churn away at 66) if that's the case you have 2
choices.

1.) Set -tickrate 33 in the start line OR 2.) Turn off hyperthreading.

Is it not possible for you to get your ISP / Datacentre staff to do the work
on your behalf?

Regards, Jon

- Original Message -
From: "Nic Strix" <[EMAIL PROTECTED]>
To: 
Sent: Thursday, October 06, 2005 3:44 PM
Subject: [hlds_linux] Hyperthreading and DoD:Source



Hi

I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon box
(dell poweredge sc1425)  since shortly after the game was released. While
running a default server with 32 slots the server lags out with one of the
virtual cpu's maxing out. It seems to run fine with 24 slots. I read
somewhere that this is caused by the hyperthreading on the box and the
only
way to solve this is to turn hyperthreading off.

My question is - is this true? Will I gain a 32 slot (hopefully many more
slots) without the cpu maxing by turning hyperthreading off? Or is there a
way to make the OS do this? Or is there a fix on its way? I dont live
close
to the box (amsterdam--london) so a quick bus journey to play with the
BIOS
is not really possible.

Any and all help will be highly appreciated! This mailinglist has already
helped me out a lot and I hope it will continue!!

TIA
Nic



___
To unsubscribe, edit your list preferences, or view the list archives,
please visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Deacon @[dgx]]

>From what I've seen, and other can probably comment in more detail,
hyperthreading can lag a server and we've seen many posts to the list
asking about this.  I would recommend searching the list archieve.

That being said, your lag can also be caused by the server config itself.
We have serveral servers that support and are running hyperthreading, but
most are AMD processors so I might be way off, but here's what I've seen.

I have several machines that run 18-24 players w/o lag.  If I take the
same config and change the maxplayer value to 26-32 people complain of
lag, choke, etc.

So, I would recommend tweaking your config for 32 players (sv_maxrate,
updaterate, tic_rate/tickrate, pingboost, etc.) prior to disabling
hyperthreading support.  Feel free to email me at [EMAIL PROTECTED] for info
on DOD:S and DOD server configs or just reply here.

Also, check for our DOD:S threads from about a week ago.  A few of us
posted our DOD:S configs and startup options.

Thanks,
Deacon


On Thu, 6 Oct 2005, Nic Strix wrote:

> Hi
>
> I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon box
> (dell poweredge sc1425)  since shortly after the game was released. While
> running a default server with 32 slots the server lags out with one of the
> virtual cpu's maxing out. It seems to run fine with 24 slots. I read
> somewhere that this is caused by the hyperthreading on the box and the only
> way to solve this is to turn hyperthreading off.
>
> My question is - is this true? Will I gain a 32 slot (hopefully many more
> slots) without the cpu maxing by turning hyperthreading off? Or is there a
> way to make the OS do this? Or is there a fix on its way? I dont live close
> to the box (amsterdam--london) so a quick bus journey to play with the BIOS
> is not really possible.
>
> Any and all help will be highly appreciated! This mailinglist has already
> helped me out a lot and I hope it will continue!!
>
> TIA
> Nic
>
>
>
> ___
> To unsubscribe, edit your list preferences, or view the list archives, please 
> visit:
> http://list.valvesoftware.com/mailman/listinfo/hlds_linux
>


___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Re: [hlds_linux] Hyperthreading and DoD:Source

[Gary]

This is correct.

At 10:44 AM 10/6/2005, Nic Strix wrote:

Hi

I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon box
(dell poweredge sc1425)  since shortly after the game was released. While
running a default server with 32 slots the server lags out with one of the
virtual cpu's maxing out. It seems to run fine with 24 slots. I read
somewhere that this is caused by the hyperthreading on the box and the only
way to solve this is to turn hyperthreading off.


G. Stanley
Engineering
Velocity Servers, DBA
[EMAIL PROTECTED]



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

[hlds_linux] Hyperthreading and DoD:Source

[Nic Strix]

Hi

I have had a DoD:S server running on Linux Fedora 3 64 on a dual xeon box
(dell poweredge sc1425)  since shortly after the game was released. While
running a default server with 32 slots the server lags out with one of the
virtual cpu's maxing out. It seems to run fine with 24 slots. I read
somewhere that this is caused by the hyperthreading on the box and the only
way to solve this is to turn hyperthreading off.

My question is - is this true? Will I gain a 32 slot (hopefully many more
slots) without the cpu maxing by turning hyperthreading off? Or is there a
way to make the OS do this? Or is there a fix on its way? I dont live close
to the box (amsterdam--london) so a quick bus journey to play with the BIOS
is not really possible.

Any and all help will be highly appreciated! This mailinglist has already
helped me out a lot and I hope it will continue!!

TIA
Nic



___
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux