Folding@Home Triggers...

[CupCak3]

My PPD have been shooting around like crazy lately so I've started to try putting new triggers on my F@H startup.

On two boxes I put -forceasm -advmethodes but this has given me some even worse PPD WUs so far

Any suggestions for some others?

I've read somewhere that there's a force SSE trigger but I haven't been able to get it to work

Also, my newest sempron has SSE, SSE2 and SSE3, is there any way to trick F@H into thinking its a intel to get some of this huge 1k+ bonus WU's?


Thanks for any suggestions Hopefully this can help quite a few people on the team boost their PPD


Thanks!


Also, all my person boxes right now have 1163 except for the two which I put the flags... its kinda odd b/c I used to only get 1808 1809 and 1850

 

[kb3edk]

Hi CupCak3, I've been on the F@H TeAm here almost two years and from Day 1, I have always used the -forceasm and -advmethods flags with the client. Yes you may be getting some bad assignments now, but you will also not be able to get some of the good WUs (like my favorite 364-pointers that finish in 20 hours ) without these flags on.

As far as "tricking" F@H into thinking it's running on an Intel, I recently heard someone hacked the client (just by switching the CPU identifier flags) to get AMDs to run the Intel-only WUs. Unfortunately, the AMD performance on these WUs was reported to be mediocre.

 

[CupCak3]

thnx for the info kb3edk




glad to see everyone else could be just as helpfull :thumbsdown:

 

[Rattledagger]

Originally posted by: CupCak3
glad to see everyone else could be just as helpfull :thumbsdown:

Well... my last folding-wu was back in August, and if there's any "triggers" they've automatically added by folding-application and not by me...


Anyway, did see something about Folding@home has currently pulled the QMD-work and is doing some evaluations, and at the same time it looks like they pulled the other big-bonus-wu.

So, everyones PPD is likely down now they're crunching "normal" work...

 

[kb3edk]

Originally posted by: Rattledagger

Originally posted by: CupCak3
glad to see everyone else could be just as helpfull :thumbsdown:

Well... my last folding-wu was back in August, and if there's any "triggers" they've automatically added by folding-application and not by me...


Anyway, did see something about Folding@home has currently pulled the QMD-work and is doing some evaluations, and at the same time it looks like they pulled the other big-bonus-wu.

So, everyones PPD is likely down now they're crunching "normal" work...

Yes Rattledagger, you can sort of "add" the triggers inside the application if you're using the graphical F@H client... just right click the little protein icon in your system tray and configure the client to allow receipt of WUs greater than 5MB in size.

But if you're using the command line version (like me) you have to specify the triggers when the client starts up.

As far as other TeAmmates not speaking up, well things have been pretty quiet on the F@H TeAM since the excitement of the race last December. Also we have been stuck in 25th place for an entire year so people haven't exactly had to come back here for any interesting news...

And yes, I heard the QMDs were on hiatus too. I didn't know about the other "bonus" WUs though. Anyway the news doesn't surprise me in the least, the scoring system on F@H is so wildly inconsistent compared to every other DC project I have done. Who knows, maybe there are a lot more people out there like Insidious, taking a stand, and then Stanford would have no choice but to fix the system or risk losing increasing numbers of participants.

 

[CupCak3]

just to clarify, I wasn't complaining b/c no one was answering me personally but it was frustrating me a bit b/c this is something that could help the team out quite a bit.

if 3/4's of the active members machines could pump out even an extra 10 PPD, that would be huge for us. I would say it could be at least an easy 4k per day for the team. (if it actually worked, but at least the potential is there )

You hit on a great point kb3edk, the massive amount in PPD variance of the WU's is a problem. It definately annoys me, especially that I don't have the time like some of the other big folders to really keep up on the best ways to keep getting the most efficient PPD WUs. I know it can be a crap shoot, but from what i've seen around, there are ways of increasing your chances for the "good" WUs.

 

[Rattledagger]

Originally posted by: kb3edk
Yes Rattledagger, you can sort of "add" the triggers inside the application if you're using the graphical F@H client... just right click the little protein icon in your system tray and configure the client to allow receipt of WUs greater than 5MB in size.

But if you're using the command line version (like me) you have to specify the triggers when the client starts up.

Not an option, since isn't using either of these folding-clients.

Preferences is set on the web-site, but AFAIK there aren't any currently that changes the folding-behaviour, except for "Write to disk at most every N seconds".

 

[Spacehead]

I was gunna ask about this anyway so i might as well ask it here. During the setup as a service, one of the settings is "8h) Set -advmethods flag always". I set it to yes but it doesn't seem to take effect. I guess i'll have to add that on my own?

 

[biodoc]

Cupcak3,

From what I've read in the F@H forum, the QMD WUs are only distributed to Intel CPUs because FahCore_96.exe was compiled using the Intel fortran compiler/libraries. Thus if FahCore_96 is run on Intel, optimizations are turned on. If the app is run on AMD, the optimizations are not turned on. Now if Stanford or anyone else hacks around this they are violating the Intel licensing agreement. The bottom line is that supposedly the QMDs run significantly slower on AMD cpus although some have tranferred these WUs to AMD boxes and claim they run at a resonable speed.

Since the above is from my relative short term memory please correct me if I'm inaccurate

Check out the the following thread at the F@H forum:

http://forum.folding-community.org/viewtopic.php?t=14308

 

[biodoc]

Originally posted by: kb3edk

As far as "tricking" F@H into thinking it's running on an Intel, I recently heard someone hacked the client (just by switching the CPU identifier flags) to get AMDs to run the Intel-only WUs. Unfortunately, the AMD performance on these WUs was reported to be mediocre.

I found a link to the "hacks".

http://distributed.ru/forum/?a=topic&topic=727

Of course you'll have to wait until QMD comes back on line

 

[GLeeM]

How did I miss this thread! Sorry!!

This is my third try at posting. I am doing it in Notepad and then will paste it into a Reply window. IE :|

There are only a few way to influence what WUs the Assignment Server (AS) sends you:
Configure client to download WUs > 5 MB = yes or no
Add (or configure v504beta client) -advmethods flag

One other way: configure for deadlineless WUs (this negates the above settings)

There is only one "Intel only" WU, the QMD. It runs with decent ppd on AMD, I have heard.

About SSE2/3:
WUs that use SSE2/3 are sent to any computer
I think -advmethods may help to get WUs that use SSE2/3
the AS does not know if your computer has this capability
these WUs will run like a normal WU ppd-wise on a computer without SSE2/3
these WUs will run like a bonus WU ppd-wise on a computer with SSE2/3
the upcoming V6 client hopefully will tell AS if computer has this and give suitable WUs

About -forceasm flag/switch:
This tells the client to NOTcheck how it was last shutdown. If the computer crashes, you lose power, or the client causes the computer to crash, then without this flag the client will run (until restarted) without optimizations.
If you have run F@H for long enough to KNOW that your computer is F@H stable, add this flag.

About the scoreing system:
Stanford uses a P4 2.8 with SSE2/3 turned off to benchmark the WUs, if you had this computer you would have a much steadier ppd
as WUs progress down the line, they can become easier or harder to crunch, and so the ppd you get can change
different CPUs will do better than the benchmark machine on some WUs and worse on others
some projects' WUs that use SSE2/3 will do better in ppd, others worse, depends on things in the WU

Bonus WUs really throw a wrench into the works:
some bonus WUs might use SSE2/3 and get even more ppd
different CPUs get widely different ppd on these depending on ram, cache, memory bandwidth, etc.
sometimes you can get these, sometimes you can't, which causes the most ppd differences that people notice, but if there were no bonus for these WUs people would complain even more or elect to not fold them and then the science would not progress with technology. Should F@H stay in the 1990's or do you want it to do research on the cutting edge?
All of these WUs are newer technology!
Stanford's goal is to do successful science to find cures as fast as possible and so they are trying out these new technologies that use more computer resources, for which they must give bonus points to entice enough people to fold them.

Right now it appears that there are NO bonus WUs being sent out! As far as I can tell.