What flags to use? Client v. 4.0
Geeky1
University of the Pacific (Stockton, CA, USA)
What flags should I be using on my systems?
Right now:
1800+ has -advmethods, -forceasm
Dual 2500s have -advmethods, -forceasm, -forceSSE
1.3GHz Celeron has none
2.4GHz P4 has -advmethods, -forceasm, -forceSSE
The Celeron is a Tualatin core CPU, so it's got the same extensions etc. as a P3.
So, what flags should I use?
Right now:
1800+ has -advmethods, -forceasm
Dual 2500s have -advmethods, -forceasm, -forceSSE
1.3GHz Celeron has none
2.4GHz P4 has -advmethods, -forceasm, -forceSSE
The Celeron is a Tualatin core CPU, so it's got the same extensions etc. as a P3.
So, what flags should I use?
0
Comments
edit:\ -forcesSSE is no longer case sensitive in Fah4Console you you can use -forcesse instead if you want.
also ...a better explantation of -advmethods from the stanford forums:
Using "-advmethods", as mentioned, means you agree to accept work units that are in the late beta stages of testing. They cannot be guaranteed stable; by definition, they are still under testing. If you cannot process "-advmethods" units, please do everyone involved a favor and remove the flag.
-forceSSE
-forceasm
-advmethods
They seem to work on a decent P4 and on a Barton, ALL of them.
I have a WU template generator running now, with Core_ca, and on the P4 it should do its 20 iteration by 30 seed work in 45-46 hours from start time. On an older client, same P4 box, no -forceSSE switch available, other USED, same basic WU family took about 60 hours-- CALCULATED guestimate, it was 2 days in when I ABORTED the thing at about 80% finished.
The Barton box did a project 1000+ (1023) in about 15.5 hours, and a p914 in about 9.5 hours so far, all three switches used. Right now the Barton box is on ASK ME connect mode simply so I can watch it turn in and intitate a dialup before letting it get on the net. After the router actually is here, will see how it does on autoconnect. so, some time is wasted unless I calc time real close and set alarm clock....:D
These are both for the 4.00 final client, the windows graphical one is running with the display minimized for client, no screensaver at all, and the Linux one is running on a console as less copies of it get spwned that way on my newer kernel (It has 5 copies of core, and 4 of client from one instance right now, up and eating RAM, and running from KDE in a Konsole doubles that numebr of "false" copies that never get a time run count), and the Linux kernel keeps trying to give it MORE CPU time slices than it is taking by attempting to renice it almost continuously to a higher priority).
John.
Warning:
By using the -forceSSE flag, you are overriding program
safeguards that monitor the stability of SSE
instructions on your system. If you did not intend
to do this, please restart the program without
-forceSSE. If work units are not completing fully,
then please discontinue use of the flag.
just a warning that f@h may malfunction and this usually only happens when the cpu/system gets too hot from the overclock and extra sse boost stress.
any amd capable of sse should benefit from this -forcesse flag somewhat when it comes to increasing perfomance ...stability may be somewhat jeopardized but you can get over that with cooling techniques.
prof this is what I recommend to you ...take the system with the coolest cpu/sys temp and start with that one ...run the -advmethods and -forcesse flags.
Check the fahlog.txt often or monitor with EMIII for a few wu's until you feel confident that f@H is running stable.
Mostly what you are concerned with is the temps under load. If you begin to have problems the wu will cease part way with a return of a partial wu usually and begin a new one (sometimes it will even fold most of the way like 88% or something).
At this point try to lower the cpu temp by cutting back as little as 1 mhz on the fsb and try again ...if you can drop the vcore a notch this also helps a lot. At this point it becomes a balance between f@h stability and cpu/system stability.
An example ...I was running my system stable 2.55ghz/442 2.0v ...the ambient temps began to get warm last week and f@h4 started killing wu's and sending them back and getting new ones. Sometimes the wu's would finish 1 step or 2 or even 35 or 60 ...just depending on when/how/why the temps increased beyond the breaking point.
So ...after a few days of no -forcesse I decided to cool the system down a bit by lowering to 2.54ghz/440 1.95v (also I increased my hsf fan rpm from 5000 to 5400) and viola ...that's where I am now.
Now granted this will probably not be sufficient enough to take care of the summer temps but I will cross that bridge when I get to it ...but for now I game and surf etc . and the warmest my cpu temp has gotten is 51C since.
Good luck!
csimon
edit:\ your success/failure temps may vary.
also ...I am not teh uber cooling expert ...you may want consult others more advanced in that arena ...I usually tend to slap a big loud fan and copper fins on it w/ AS#!!!
I'm going to try your advice. I've also copied this to a doc file for future reference - great summation!
Intel: -advmethods
AMD: -forcesse -advmethods
That is assuming that the AMD has SSE otherwise forceasm should work.
As always csimon is on top of things.:)