Hi Jim,
>I must disagree with you, Jordan, based on the thread:
>
A tip on hardware to speed up VFP Thread #
472486 Message #
472486I've read that message as well and was a bit sceptical back than.
As I had only limited exposure to multicore - machines back then and
the author mentioned his expirience going against "conventional wisdom",
I kept it in the back of my mind as well.
Since then I had the chance to run some very taxing vfp apps on such machines:
2 P IV Hyperthreading enabled
1 Dell Xeon double processor (each single core but Hyper-threading capable)
3 different Athlon X2
1 Meron Dual Core (only short spin, not full exposure yet)
and I have
not been able to recreate his findings.
In my tests searching for "pure vfp speed" I disabled almost all other
running processes/services (including firewall, virus scan and so on
after pulling the LAN cable). The very best CPU utilization I got was
around 65% on Athlon double cores when running table updates on
compressed files with complex vfp-udf processing.
The kernel time was highly correlated to the percentage above 50%.
The same apps on HT single cores showed only CPU utilization of about 54%,
and the Dell showed about 28%, which is about the same since the OS considers it
to be a quad core machine.
When there are other processes running, I get about 50% CPU load for the vfp app
but the machine is handling the background tasks on the other cores and feels much
more responsive - but more on pure double cores than on HT single cores. Multiple
HT cores didn't help for my uses at all.
I nowadays think the message you reminded of was either a VERY freak special situation
or a measurement error including other OS-processes - perhaps virus scan or a SQL server.
For taxing workstation work WITH a human user hacking at the machine I recommend
nowadays non-HT double cores, for pure vfp-batch-processing machines WITHOUT
human keyboard banging I recommend single cores, as the second CPU is underutilized.
Server machines are definitely to be measured by other yardsticks <bg>.
My 0.03 EUR (a tiny bit more than usual <g>)
thomas