CPU performance?

[Glenn119]

The thread about a slow down in R@H got me back to thinking about Hyper threading and general CPU performance.
As I wish to get the most I can out of my CPU’s, I ran a series of tests on a few of my PC’s.

P4 3.2 Ghz 512ram 1826 Whetstone 6379 Dhrystone no Hyper threading

P4 1.8 Ghz 128ram 973 Whetstone 2132 Dhrystone no Hyper threading

Xeon Dual 3.6Ghz 2gb ram 1596 Whetstone per CPU 3875 Dhrystone per CPU Hyper threading on 4 logical CPU’s

Test Computer
Xeon Dual 2.2Ghz 1gb ram 875 Whetstone per CPU 2132 Dhrystone per CPU Hyper threading on 4 logical CPU’s

Xeon Dual 2.2Ghz 1gb ram 1108 Whetstone per CPU 3559 Dhrystone per CPU Hyper threading on 2 logical CPU’s

Xeon Dual 2.2Ghz 1gb ram 1188 Whetstone per CPU 3806 Dhrystone per CPU Hyper threading off 2 logical CPU’s

Not that I’m really sure what all this means (I’m no guru) but I have concluded that I get the most work out of using hyper threading and 4 logical CPU’s.

Any comments or help interpreting my results is much appreciated.

:)

[Bok]

Are you using the optimized boinc clients ?

It's hard to tell whether it's better or not from just the benchmarks, because with clients running on logical CPU's (HT), then they will all take longer I would have thought.

Only real way is to have two identical PC's, run one HT and other not and compare the RAC as it goes along.

Unfortunately I don't have any 2 identical Intel machines!

Bok

[Glenn119]

Test Computer
Xeon Dual CPU's 2.2Ghz 1gb ram 875 Whetstone per CPU 2132 Dhrystone per CPU Hyper threading on 4 logical CPU’s

Xeon Dual CPU's 2.2Ghz 1gb ram 1108 Whetstone per CPU 3559 Dhrystone per CPU Hyper threading on

Xeon Dual CPU's 2.2Ghz 1gb ram 1188 Whetstone per CPU 3806 Dhrystone per CPU Hyper threading off

Yes runing the optimized clients.

Only the top example is using HT (4 logical)

from example 2 and 3 it seems just having hyper threading enabled causes a little hit in performance. What I did was change my R@H preferances from use 4 processors to 2 between example 1 and 2 :umm: