SMP Affinity Changer

[zorzyk]

A-Ch with one SMP client on Pentium D915 (2800 MHz stock, dual core, 1 MB RAM) gives about 5% decrease of folding time, which means about 2.5 hours less needed to complete WU (p2653: was ca. 52 h, with A-Ch now 49h 15mins), and PPD raised from 815 to 855.

[thriftyj*]

Hi Guys

Can someone make the need of this affinity changer stupidly simple for me please?

I'm now on an Asus Intel Q6600 box running two instances of F@H.

Will affinity changer be of any use, or not?

Thank you.

hondo^
The need for it will be up to you. I just installed it and tested it myself and am seeing it work. What rilian did was make the computer work better.

the SMP divides the work into two cores - but sometimes one core is working hard and the other core is lagging.

Affinity- (as I understand it and grossly oversimplify) checks the two cores every so often. Even outs the work load making the core with less work get more, working the computer work more efficiently.

Voila' - more efficient code. Will it work for you? [install it where it is supposed to go, correctly, and I think it will - but I am still a new believer - although, if it is good enough for uncle fungus? I would stake my fungus on it! ]

thanks again, rilian.>>increased speed 11% on WU 2653

[él Mero]

Although rilian does a great job advertising this software, the guy you really should thank is nik4eva, the developer.

[rilian]

Although rilian does a great job advertising this software, the guy you really should thank is nik4eva, the developer.

agree

[thriftyj*]

Many thanks to nik4eva.

rilian is to be thanked as the one who answered my questions and the one I was speaking with.

[MoneyGuyBK]

Although rilian does a great job advertising this software, the guy you really should thank is nik4eva, the developer.

Many Thanx to nik4eva, rilian and Folding Forums for developing and promoting this great tool.
A-Ch has netted me [and humanity and the science] a good 3.5 hours per SMP.... Kudos.


Peace

[*hondo*]

Sori for asking maybe a stupid question, but how do you know if Affinity Changer is working or not?

[Jeannie]

Sori for asking maybe a stupid question, but how do you know if Affinity Changer is working or not?

By comparing how long it takes to complete a particular work unit without the affinity changer to how long it takes to complete another WU of the same project with affinity changer, on the same client.
If you're using something like FAHMON to monitor your client(s), the monitor will also display Points Per Day (PPD) for each client for the projects you are running, which gives you a quicker 'answer' to how the affinity changer affects your client. Again, you need to compare the same project, and allow for the fact that there are normal variations in the 'speed'. For example, one of my computers almost always gets WU2653, using Windows SMP, and it usually gets around 990 PPD, but sometimes gets as low as 900 PPD if something else is running on the computer (but usually only changes to as low as 950). When I tried the Affinity Changer, my PPD went down into the 700's, so I don't use the Affinity Changer (my processor is an AMD; the Affinity Changer was designed to work with Intel processors, so I'm not complaining.)

[butc8]

Sori for asking maybe a stupid question, but how do you know if Affinity Changer is working or not?

By comparing how long it takes to complete a particular work unit without the affinity changer to how long it takes to complete another WU of the same project with affinity changer, on the same client.
If you're using something like FAHMON to monitor your client(s), the monitor will also display Points Per Day (PPD) for each client for the projects you are running, which gives you a quicker 'answer' to how the affinity changer affects your client. Again, you need to compare the same project, and allow for the fact that there are normal variations in the 'speed'. For example, one of my computers almost always gets WU2653, using Windows SMP, and it usually gets around 990 PPD, but sometimes gets as low as 900 PPD if something else is running on the computer (but usually only changes to as low as 950). When I tried the Affinity Changer, my PPD went down into the 700's, so I don't use the Affinity Changer (my processor is an AMD; the Affinity Changer was designed to work with Intel processors, so I'm not complaining.)

Or go into task manager under processes there should be a "FahSmpAffinityChanger.exe" there, and dont worry if it says it uses 00:00:00 cpu time.

[*hondo*]

Thanks for the information you Guy's, I don't know if I've done something wrong but to be honest I can't see a major difference, with or without Aff/Ch.
On the same WUs P3050 & P2653 for 2 Hours I ran with Aff /ch, after which I un-installed it and for a further 2 hours ran without it.
The result was

With A/F PPD= 3317

Without A/F PPD = 3322
The CPU is a P Q6600.

Thanks again all you guy's for the feedback.

[rilian]

*hondo* you did use two FAH SMP ?

[*hondo*]

*hondo* you did use two FAH SMP ?

Yes the same one's, P3050 & P2653 Did I do something wrong?

[bruce]

Affinity changer seems to work best with a single copy of FAH-SMP (four copies of FahCore_a1) on 4-core hardware or two copies of FAH-SMP (8 copies of FahCore_a1) on 8-core hardware. If you have more copies of FahCore_a1 running than you have physical cores, the benefit diminishes rapidly.

[rilian]

bruce please check your last post. A-Ch does not work on 8-core hardware so far. i guess you need to divide by 2 all core-related part

[*hondo*]

I'm sorry you guy's I can imagine this must be frustrating for you, I'm clueless to what you are saying.

My CPU is an Oced Pentium Q6600 which I thought is a 4 core. I'm running two SMP WUs, which are completed within usually < 24 Hrs. Is there a more effective way to run this CPU either with, or without A/Ch ?

Thanks again.