hyperthreaded processors running on win2k

My company is currently deploying new hardware for our new hosting site. We
have all new servers. 4 of them are running Win2k standard server and are
web servers, and 3 of them are win2k advanced servers running SQL Server
2000.
These servers have dual Xeons, so if hyperthreading is on, win2k sees 4
procs.
We have been working on tracking down speed issues for quite a while and
recently realized (by accident) , that the webservers run WAY faster when
hyperthreading is turned off. We are using MS Application Center Test for
load testing, so that's how we determined the performance gain.
I've been trying to find a clear cut answer about whether or not you should
use hyperthreading on win2k server, but I haven't found one. Although I have
seen several references to this issue and seen posts of people saying to
shut if off. It seems win2k doesn't know the difference between logical and
physical processors, so maybe that's why it's not good for win2k. ?
Does anyone know where I could find more info on this? We're going to turn
off hyperthreading on the webservers obviously because they perform better,
but I'm wondering if we will see performance gains by also shutting of
hyperthreading on the win2k SQL Servers. ?
Any insight into this would be great.
Thanks.Personally, I have had fairly good experience with Hyperthreading. First,
make sure you are on build 816 or later. There were some significant
HT-friendly modifications in that build. The easiest way is to download the
818 security hotfix and apply it. That way, you do not have to open a case
to get it. Regardless, you must be on SP2 or higher to get SQL to count the
processors correctly for licensing purposes.
Finally, 2 virtual HT processors run about 20-30% faster using SQL than the
same processer set to non HT. SQL still counts them as full processors for
parallelism, meaning that you can over-parallel a query and drag system
performance down. I manually set MAX degree OF PARALLELISM (MAXDOP) to no
higher than the actual physical processor count.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Mark" <maxmann@.kc.rr.com> wrote in message
news:uyOJwARJEHA.1312@.TK2MSFTNGP12.phx.gbl...
> My company is currently deploying new hardware for our new hosting site.
We
> have all new servers. 4 of them are running Win2k standard server and are
> web servers, and 3 of them are win2k advanced servers running SQL Server
> 2000.
> These servers have dual Xeons, so if hyperthreading is on, win2k sees 4
> procs.
> We have been working on tracking down speed issues for quite a while and
> recently realized (by accident) , that the webservers run WAY faster when
> hyperthreading is turned off. We are using MS Application Center Test for
> load testing, so that's how we determined the performance gain.
> I've been trying to find a clear cut answer about whether or not you
should
> use hyperthreading on win2k server, but I haven't found one. Although I
have
> seen several references to this issue and seen posts of people saying to
> shut if off. It seems win2k doesn't know the difference between logical
and
> physical processors, so maybe that's why it's not good for win2k. ?
> Does anyone know where I could find more info on this? We're going to
turn
> off hyperthreading on the webservers obviously because they perform
better,
> but I'm wondering if we will see performance gains by also shutting of
> hyperthreading on the win2k SQL Servers. ?
>
> Any insight into this would be great.
> Thanks.
>