Member discussion of optimal flag settings., Add your questions and findings here.

[susato]

John (314159) sent this report today via email concerning his Pentium 4 rigs:  

-local -forceasm -verbosity 9, with big WU (bigpackets, 5 Mb) set to on, has given me an absolutely steady stream of bonus WUs over about the last several weeks.
A hyper-threaded box running two instances of folding produces at least 12 pph on these

[314159]

Hmmm,

Still getting them and almost exclusively.

In -configonly (advanced options), "always set -advmethods as on" response must be "no" if you wish to try this.

Did I not mention that, Jean?

This strategy could, of course, change quickly. I saw server .162 with the choice WIN WUs the other night. No "F" WUs per serverstats page but a full and active server with W "A's" available.

Added -advmethods at the appropriate time as a test and got what I was looking for immediately.    
(added this to the client launch parameters and NOT the config file.)

This is making these older machines almost usable until I can swap out some Mobos and upgrade them.

The following quote from my wife:

[susato]

Looks like it's time for another edit in the  Optimal Client Settings thread.

Over the past 6 months we've had a lot of discussion of multiple instances of SMP folding on 4 and 8 core machines, the size of work units (small, normal or big),  the flexibility of InCrease compared to the Stanford OSX folding installer, and its user friendliness compared to the barebones console client.

The past six months has also seen the release of the version 6 software, the introduction of passkey codes, and the evolution of InCrease to handle these new features in the software. And the PS3 client upgrade.  Wow!

Please contribute your ideas for the next version of the optimal client settings recommendation post - especially if you are running a PPC Mac or a quad or octo-core computer.   Any other advice?  Bring it on!

[iBozz]

G'day!

Will there be a "preferred settings" update soon?

I'm now running a quad core 27" i7 iMac under OSX.6.2, running 24/7, with the following settings:

Use CPU count of Actual (8)
-advmethods
-forecasm
-verbosity 9
extra args:  -bigadv -smp 8

InCrease shows 1 client 8 cores.

Now I don't begin to understand what all that means (all right, I admit it, what any of that means), but is that the optimum setting right now?

I did read on the f=24&t=10697" target="_blank">F@H forum here (second post, ninth Question):

Q: I have all of that but for some reason I'm still being assigned regular WUs.
A1: When you configured the client, perhaps you didn't accept big WUs. If that's not set properly, several different errors might occur.

A2: If you're using -advmethods or any other flag intended to designate a specific class of WU assignment, remove that flag becuase it can override -bigadv.

that having -advmethods can perhaps override having -bigadv and I infer, from posts earlier in this thread, that there has been some contradiction and/or confusion amongst The Great And The Good about this, so we of The Great Unwashed stand no chance!

According to a MiniUsage, the cores are around 85% to 90% dedicated to Folding and I'm getting around 250 pts/hr on Project 2671.

Incidentally, I notice that the CPU temperature leaps from around 95F to around 150F when InCrease/Folding are running - should I be concerned or can i continue to make toast on the back of the beast?    

Thanks.

[jackrabbit]

Nice upgrade of your machine, maybe it's time to update your sig line

Not a clue, I'm afraid. PM Macaholic about updating the settings recommendation, he should know.

[Macaholic]

Yes.  Use -advmethods if you want regular SMP units only OR -bigadv if you want the whopper SMP units.  Don't use both.  If you use -bigadv the servers should automatically fall back to the regular SMP units (-advmethods setting) if no whopper SMP units are available.

[curby.net]

As I've mentioned I have a friend that's letting me fold on his machine. We've been tweaking the -smp parameter to give him enough of his machine to do his work and letting it fold with the rest. Question is: can the -smp parameter accept arbitrary (i.e. odd) values? We've been using 6 cores which seems to be a decent balance, but he's willing to try using 7 for folding if that would be accepted by the program. Alternately, would 7 be accepted but lead to the client rounding up or down to the next even number? Or should I not even attempt using an odd number of cores? We're doing standard (not huge) smp WUs on this machine.

Thanks!

[Ravage]

-smp can be set using any number of cores (real or virtual) 2 or greater. I am currently running -smp 3 on my quad box cuz ubuntu 9.10 refuses to run smp and gpu's nicely together when smp uses all 4 cores. I have to keep a core free for the gpu's or the performance of the box drops precipitously.

Running an smp # higher than the number of cores available (real and virtual) is possible but not recommended. It just kills the performance of the box.

[curby.net]

Thanks for the quick reply! I'll go give -smp 7 a test.

[Aardvark]

Somewhere in the ff.org Forums, recently, kasson warned against using 'high" prime numbers as smp flags. As I recall, there was no defining as to when high begins. Perhaps someone can educate us further on the dark mysteries of multi-core folding.....