Hyper-Threading Technology and SW

I posted this here some time ago (thread was "Hyper-Threading Technology"):

So your 50% / 100% measures are perfectly consistent with Intel's marketing,
and with the planned future processors that will actually have 2 processing
units on the chip. My guess is you should keep hyper-threading on if your
computer runs other software in parallel with SW.

Besides, some people argued that SW actually does support multiprocessors,
or rather that Parasolid uses it in a parallelizable jobs such as the
(dreaded) shells, but I could never verify it on my dual AMD machine...

I also suggested once that recent stability problems in SW *could* be
related to *possible* work to rewrite it (thread-safe...) to take advantage
of multiprocessing in the future. Time will tell...

 

I just bought a new system. A Shuttle with a intel 3 Ghz, 1 gig ram
a(hyperx ddr433) a FX500 with a Zippy keyboard. The system has
Hyper-threading technology, which simulate a second cpu by software. Since
SW doesn't support multi cpu my assumption is that turning off this
capability will speed up the system. When I rebuild, the cpu never exceed
50%. Now with the hyper-threading off the cpu goes to 100%. Anyone has
seen this?


--
Robin B
Mech Eng, Montréal, Canada
Self employed
SW2004 SP2.1
Shuttle SB75G
Intel 3 Ghz, 1 gig ram HyperX ddr433
FX500, Spaceball 4000FLX

 

I vaguely remember someone reporting about a 5-10% slowdown running SW (or
maybe it was PW) when running with HT on.


Jerry Steiger
Tripod Data Systems

 

I remember that post and it was slower. PW is supposedly a little faster
but the only way to benchmark PW, AFAIK, is to use a stopwatch and that's a
little subjective. Since we're talking about the speed "processing" here,
maybe Mike's Ship in the Bottle benchmark could shed some light on it. If
there's no difference or an improvement, leave HT on. If it rebuilds slower
then leave it off.

- Eddy

 

Hi

I posted just before christmas after I did a benchmarktest which
showed that Hyper-Threading is faster. Therefore I worked for a while
with hyper-threading enabled. But I disabled it a month ago and it
feels faster. I write feels because I have nothing that proves it.

Klaus

 

Yes, I remember. I know your HT test showed otherwise but the conclusion in
the thread was that it's slower. Or maybe there was no conclusion
depending how you look at it

- Eddy

 

Ok, I did my own test.

First test is a rebuild on arraying a hole for a 38*27 holes. I just click
ok and look at the watch. Hyper-threading on take 33 sec and HT off takes
29 sec.

Second test is archiving in rar format (best compression) a directory of
pictures (including many sub-directories). Result: 46 sec in both cases.

I'll leave it off since it seems equal or faster and I don't have to bother
with a dual cpu in the task manager that never reach 100%.

Robin