FaD with hypherthreading

[tgxiii]

I noticed that with my P4C system with hyperthreading I can run two processes of THINK.EXE due to the two logical processors. I was wondering, would I get better performance if I set each of the processes on a different affinty? For example, run THINK on CPU 0 only and THINK-1 on CPU 1 only. Or, would it be better if both processes were running with both CPUs? THINK and THINK-1 allowed to run on CPU 0 and CPU 1.

If it's better to separate them, how would I go about making it so that it does that whenever I start THINK?

[PresterJohn]

there has been some old threads about hyper-threading (HT)and running more FAD threads then there are processors.

the consensus is that for the overwhelming majority of the WU's being crunched, you would get the best results from running only one FAD thread per cpu. in cases of HT, you should disable the auto-sense and set it manually to run just one FAD thread because you really don't have 2 physical cpus in your machine.

http://www.xtremepccentral.com/forums/showthread.php?s=&threadid=9667

[tgxiii]

Ah, I see. So only one instance per physical CPU is what I got out of it.

Reading through that other thread led me to some other questions. Did I understand correctly that running the Queue window hidden is less efficient than if it were open? What's the reasoning in this?

[PresterJohn]

Ah, I see. So only one instance per physical CPU is what I got out of it.

bingo...give the man a prize! :)

Did I understand correctly that running the Queue window hidden is less efficient than if it were open? What's the reasoning in this?

also corrrect! :)

for optimal crunching efficiency, run the queue window visible (not hidden) and just minimize it to the program bar. the reason for this is that when the queue runs as hidden, in some cases......it may take up to ~15 minutes for the server to allocate a work unit (WU) to the waiting cpu.

[tgxiii]

Thanks, PJ.

P.S.
What's my prize? :D

:beer:

[PresterJohn]

Originally posted by tgxiii
P.S.
What's my prize? :D

:beer:

ask TT or Sharder...they're the prize-givers around here. me...i'm just a humble FAD worker ant. ;)

[dannychuckle]

But the other stuff you run will get out of think's way faster cos you've got HT... which is nice :D

[sharder8]

Originally posted by PresterJohn
ask TT or Sharder...they're the prize-givers around here. me...i'm just a humble FAD worker ant. ;) Well . . . . . I agree with PJ on the Intel processors! :P

But, I'm one of the few that continue to run 2 think instances on a single AthlonXP processor! :rolleyes:

So-o-o-o-o, No prize from me, but check with TT, he's the Celery man! :bunny:

Harder

[JSkorna]

I've tried it both ways on my XPs and am ready to go back to 2 instances also. Jobs just seem to take forever.

[serlv ( Zippy )]

As I understand it, it is more beneficial to run it non-minimized, too. I cannot remember why that is, but there is supposed to be a good reason for this. I do. Anything I have open will cover it anyway. If I have nothing open, but want to see my wallpaper or something I just drag it to a corner or down to where only an inch or so is showing.

[PresterJohn]

Originally posted by serlv ( Zippy )
As I understand it, it is more beneficial to run it non-minimized, too. I cannot remember why that is, but there is supposed to be a good reason for this.

from an OS pov, it shouldn't matter if it is minimized to the bar or if it is windowed, it accomplishes the same thing.

nt/w2k/xp will default to providing thread focus to the foreground app/window. thus it does not matter if is windowed or minimized, it is still getting secondary time-slices from the cpu.

[Twinkletoes]

Originally posted by serlv ( Zippy )
As I understand it, it is more beneficial to run it non-minimized, too. Just for the record, I certainly misunderstood this on first reading, Zippy. I think ! So to aimiably demonstrate my stupidity in these things, is this a fair summary of what you guys are saying:

It's better to run it minimized down on the task bar, but it's just as good, or maybe even better to drag it down to the smallest size possible and leave it tucked away in a corner of the main screen ?

tgxiii/Sharder Sorry, no prize from me either. And also just for the record, I certainly ain't a Celeron man. I'll give you some comparative figures later once "Fred", my new XP2000+, has been chugging along for a while, but as far as Think is concerned, the two 2GHz Celerons were not the best buy for crunching. In fact I'm a bit disappointed in them. I'd swop them for Athlons any day. One lives and learns, even when one is approaching Geobot's age.





(For tgxiii's benefit: Geobot is a day older than I.)

[gino x]

I still consider you the old fart TT :D

[sharder8]

Originally posted by Twinkletoes
I'll give you some comparative figures later once "Fred", my new XP2000+, [/SIZE] Dang it TT!!! :rolleyes:

Last week I repaired and added a HP Vectra, 350MHz PII, to my herd and named it "Fred"!! :bunny: It's actually the replacement for "TwinkleToes". :rolleyes:

Harder

[Twinkletoes]

Originally posted by gino x
I still consider you the old fart TT Well, Young Thinker, I believe that reflects more badly on you and your brainpower, than on me and my gut-power. :D


Sorry to hear about your faux-pas Sharder me old mate. My heart bleeds for you. I will inform you in good time when I start applying my mighty brain to finding a suitable name beginning with "G". (The wife is going on holiday at the end of February. I might decide to have a second birthday then.)

[tgxiii]

New information about FAD with HT here: http://forum.find-a-drug.org.uk/viewtopic.php?t=4266&start=0&postdays=0&postorder=asc&highlight=

What do you guys think?

[prujohn]

Ok so I have an HT processor... the verdict is 2 FADs?

[prujohn]

Interesting results when I run 2 FADS with HT...

Well first, when I run only one instance of FAD.. my processor shows varying utlization (between 70-100%)

When i run 2 instances of FAD, both sides of my processor peg out at 100% solid... PC runs normally though and I can do other stuff on it... no major overheatage either...