Subject Does 1.5 Classic on Windows suffer from the HT problem?
Author cg_bland
Hi there,

We are aware that SuperServer on Windows can suffer from
hyperthreading "see-saw" problems, and have experienced them ourselves
at a customer on an 8-way W2003 Enterprise box. Unfortunately, it will
be difficult for us to test on this machine for a while.

So, we had some questions on Classic.

Does the HT issue also apply to FB1.5 Classic on Windows?

My reading of HT is this (and could be wrong):
HT causes a "see saw" problem previously only occurred on SMP
machines. Setting the CPU affinity prevented this from happening by
tieing the processors to a physical CPU. Unfortunately, with HT, you
get see-sawing within that physical CPU, and you can't lock the
process to a virtual CPU. So HT has to be turned off.

Now, how does this sit with Classic (assuming that CPU affinity is set
to allow multiple CPUs to be used)? Is HT a "non problem"?

Or, is the approach to set the CPU affinity to use multiple CPUs, but
turn off HT so that the processes don't get bounced within a virtual CPU?

Finally, if things go ahead, we will be deploying Windows Classic in a
production environment. We are testing at the moment - however, has
anyone else had good/bad experiences with Windows Classic?

Many thanks
Chris