Page 2 of 3
Re: Late hand-ins still useful?
Posted: Tue Sep 14, 2010 5:07 pm
by MtM
I thought -advmethods ( eg late beta ) units tended to be larger then current production work units, but those things are always changing so probably you are right about the current situation. I can find references of it being mentioned that selecting no advmethotds will tend to give smaller wu's, but those might deal with older client/work unit mixes.
Also, the wu size field has an influence on what units you recieve, probably not by wu size but AS logic and project weight? Again, I can find references but I'm not aware if they are still valid.
Re: Late hand-ins still useful?
Posted: Tue Sep 14, 2010 11:53 pm
by bruce
The client for Intel OS-X machines only runs SMP WUs and they have fairly short deadlines so the assumption is that you'll run 24x7, especially on a Core(2)Dual. Deadlines vary with the complexity of the WU, so there's no particular advantage to selecting advmethods or Small/Normal/Big WUs. They'll all need roughly the same number of processing hours per day.
I'm not a Mac person, myself, but you mentioned Boot Camp. If that's an option for you, I'd try downloading either of the 6.23 Windows Uniprocessor clients. ** The uniprocessor WUs have proportionally longer deadlines so you should be able to complete useful work if you fold fewer hours per day, though I don't know how many hours would be required.
The faster you return any WU, the better, but anything that beats the Preferred Deadline will be considered useful. Between the Preferred and the Final Deadlines, there is still some value, but the work will have been reissued to someone else, so there's a high probability that it will be duplicated.
If you fold fewer hours per day than it takes to make either deadline for these WUs, then I recommend that you not participating at all.
You probably already discovered the project summary page, but all deadlines can be found here:
http://fah-web.stanford.edu/psummary.html
** Other notes:
The console client uses the CLI window. You can open that screen and see a text description of the progress
The systray client runs in the Windows Systray. It comes with a viewer, but I don't recommend using the viewer. Progress can be noted from the tooltip (text from a mouse flyover of the icon in the systray) or by viewing the log (same as the CLI output).
Most people who check on progress regularly use one of the third-party monitor programs. but that's not required.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 12:36 am
by 7im
MtM wrote:I thought -advmethods ( eg late beta ) units tended to be larger then current production work units, but those things are always changing so probably you are right about the current situation. I can find references of it being mentioned that selecting no advmethotds will tend to give smaller wu's, but those might deal with older client/work unit mixes.
Also, the wu size field has an influence on what units you recieve, probably not by wu size but AS logic and project weight? Again, I can find references but I'm not aware if they are still valid.
If you can find references about adv=bigger on the Fah web site, they are incorrect, and I will fix them. Please provide links.
WU size does affect AS logic for CPU clients, does NOT for SMP clients. Please don't mix the two.
If you have FAH references that do mix the two, please provide links, and again, I will correct the mistakes.
Thanks.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 4:39 am
by lanbrown
PantherX wrote:Last time I checked, Macs are only capable of folding SMP WUs. There isn't a Classic Client or WUs for Mac.
@lanbrown: Can you please tell me where you came across -smp 1 because I was informed by a well known user that -smp 1 isn't a valid argument (and that is correct). From my personal experimenting, I found that -smp 1 = -smp so if you are interested in the details, here is my
post.
The Mac client has always been a bit different and they have had multiprocessor/core machines longer and more mainstream than everyone else.
There IS a classic client for the Mac as it was for the PPC and the Core Solo machines; the v5 client. The v6 client was Intel only but it didn't support the Core Solo.
Not the most recent information from Standford, but:
Can I mix and match SMP and non-SMP work units?
The integrated client will run either SMP or non-SMP work units on eligible architectures. The version 6 client allows you to request either SMP or non-SMP work units from the command line; however, there is no change in the architectures for which we have no SMP work units (e.g. OSX/PPC) and for which we have no uni-processor work units (e.g. OSX/Intel). Also, on Windows the SMP client is not yet available.
So it looks like the OP would need to use the PPC client and let Rosetta do the conversion.
The v6 client is a native SMP client, the CPU threshold settings don't work, so how would one limit the usage?
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 5:07 am
by PantherX
For limiting the SMP CPU Usage, use -smp X where X>=2 but not more than the total CPUs your system has.
BTW PPC is not longer supported by F@H.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 9:32 am
by MtM
7im wrote:MtM wrote:I thought -advmethods ( eg late beta ) units tended to be larger then current production work units, but those things are always changing so probably you are right about the current situation. I can find references of it being mentioned that selecting no advmethotds will tend to give smaller wu's, but those might deal with older client/work unit mixes.
Also, the wu size field has an influence on what units you recieve, probably not by wu size but AS logic and project weight? Again, I can find references but I'm not aware if they are still valid.
If you can find references about adv=bigger on the Fah web site, they are incorrect, and I will fix them. Please provide links.
WU size does affect AS logic for CPU clients, does NOT for SMP clients. Please don't mix the two.
If you have FAH references that do mix the two, please provide links, and again, I will correct the mistakes.
Thanks.
Only thing I feel like putting on the forums instead of pm, only because it's the best example I can find and well, you said earlier in this thread that I didn't know what I was talking about ( you hurt my pride boohoo
).
Install guide wrote:This is the same as using the -advmethods command line flag. This option states a preference to request newer late stage beta work units. Because newer work units tend to be larger, this setting my slightly increase your points per day. But because the mix of work units changes from week to week, using this setting might also reduce your PPD, or have no affect at all.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 1:59 pm
by lanbrown
PantherX wrote:For limiting the SMP CPU Usage, use -smp X where X>=2 but not more than the total CPUs your system has.
BTW PPC is not longer supported by F@H.
But they still are handing out WU's though!!!!
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 2:47 pm
by codysluder
They do hand out PPC WUs a lot of the time but there have also been pretty long periods when there were no WUs to be had. No longer supported doesn't mean it won't work. They often keep things running well after they are no longer supported if the results are still useful, but you need to be advised that it may stop working at any time without notice and if it does stop, please don't complain.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 2:47 pm
by PantherX
lanbrown wrote:...But they still are handing out WU's though!!!!
Now that is strange
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 4:02 pm
by lanbrown
PantherX wrote:lanbrown wrote:...But they still are handing out WU's though!!!!
Now that is strange
Why? The client is not being actively being supported, but there are still projects that can use it. It all depends on the project owner and what core needs to be used. Amber is a no on the PPC but other projects are a go.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 4:10 pm
by PantherX
The reason is that I didn't expect it to last that long:
http://folding.typepad.com/news/2010/01 ... pport.html
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 4:35 pm
by 7im
MtM wrote:
Only thing I feel like putting on the forums instead of pm, only because it's the best example I can find and well, you said earlier in this thread that I didn't know what I was talking about ( you hurt my pride boohoo
).
Install guide wrote:This is the same as using the -advmethods command line flag. This option states a preference to request newer late stage beta work units. Because newer work units tend to be larger, this setting my slightly increase your points per day. But because the mix of work units changes from week to week, using this setting might also reduce your PPD, or have no affect at all.
Fixed another place that Myth was mistakenly perpetuated.
Set -advmethods flag always, requesting new advanced
scientific cores and/or work units if available (no/yes) [no]?
This is the same as using the -advmethods command line flag, except -advmethods will not appear on the arguments line in the fahlog. This option states a preference to request late stage beta work units. Do not assume using -advmethods will enable FAH clients to get larger WUs or WUs with higher PPD. Because the mix of work unit projects change weekly or monthly, using this setting might reduce your PPD, increase your PPD, or have no affect at all. The same applies for work unit sizes, increasing, decreasing, or not. As the F@h project progresses, WUs tend to get bigger over time, so the work unit sizes will tend to increase whether this flag is used or not.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 5:21 pm
by lanbrown
This is even
NEWER.
http://en.fah-addict.net/news/news-0-22 ... d-2973.php
Xuhui Huang has just released four new uniprocessor projects: 2968, 2969, 2972, and 2973.
These projects simulate the trpzip2 protein system at different temperatures in order to test the sampling algorithms which have been developed to limit and control temperature jumps.
The projects utilise the Gromacs core, and are distributed to clients using the -advmethods flag through a new server at 143.89.28.70, hosted by the Hong Kong University of Science and Technology.
Specific details on these projects, such as deadlines, number of atoms, points, etc. are currently unknown, but we will update this news with the specifics as soon as they emerge.
Updated on 22/05/2010 :
These projects have been released to the entire FAH userbase, so you should get them without needing to start your client with any special flags.
They have also appeared on the psummary page with the following characteristics: 8,991 atoms, preferred deadline of 16.7 days, final deadline of 24 days, and worth 80.7 points.
The most important aspect:
Updated on 22/05/2010 :
These projects have been released to the
entire FAH userbase, so you should get them without needing to start your client with any special flags.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 5:31 pm
by PantherX
7im wrote:...As the F@h project progresses, WUs tend to get bigger over time, so the work unit sizes will tend to increase whether this flag is used or not.
Does it mean:
A) As the Gen # increases, the WU size increases?
B) As F@H releases more Projects, the general trend is larger WU Size?
Agreed, but where does it specify PPC? AFAIK, uniprocessor means Classic Client which is currently on Windows only.
Re: Late hand-ins still useful?
Posted: Wed Sep 15, 2010 5:32 pm
by Tobit
PantherX wrote:AFAIK, uniprocessor means Classic Client which is currently on Windows only.
and Linux