Gromacs nears full release

[Bruce]

If you're contributing to team 2136075668 you should ignore this message and keep doing what you're doing.

If you're contributing to team 131, read and heed. (exception: if you've got an old CPU - - like a K6-II or Celeron 500 you won't be able to do this.)

The Gromacs core has once again finished beta and reached a possible candidate release version. If you restart your client with the flags "-advmethods -forceasm" on your command line, the current WU will finish normally, and the next one should be a Gromacs WU. A new core will download and you'll start producing about 2.5x as much work as you have been.

There is still a slight risk of a crash, but these WUs have been beat up really hard in beta test, so unless you've got a machine with really aggressive RAM clocking or one with compromised cooling, you should be fine. The new core will fire up the (almost) unused section of your CPU that runs SSE or 3DNow+ instructions and work it harder than you've ever worked that machine. Both your team and science will be thankful at your increased production.

For further information, ask here - - or visit me at http://forum.folding-community.org/ when I'm not here.

[zhotfire]

Thanks Bruce! Done... and done! :)

[veggyhed]

XP2000 = switched over
XP2100 = switched over

Thanks.

[Louis_Berding2002]

Switched....lets see how it works....

[Virus]

Bruce, can you give a write up as if I was a dummy, because I am on how to add the lines to the command line.

[Bruce]

Well, you didn't specify, but I'll assume you're running Windows and ;) don't have a copy of "Windows for Dummies" ;)

There'll be a shortcut for FAH in your Startup folder. I'll assume that's how you get the client started - - though you can do this to any of the icons - - or you can copy one and work only on the copy.

Right click on the icon and select Properties. In the Target field, you'll find the full command that starts the client. (Something like "C:\Program files\FAH\FAH3Console.exe" or maybe "C:\Program files\FAH\winFAH.exe") Depending on whether the name contains a space, you may or may not have quotation marks.

On the extreme right end of that line (without changing anything that is already there!) add a space and whatever parameter(s) you wish to add. It may now look like this:
"C:\Program files\FAH\FAH3Console.exe" -advmethods

Stop the client (if the console, use CTRL-C; if the winGUI client, use the context menu) and then start it using the shortcut you just modified.

If it's still not clear, ask again.

[veggyhed]

dont forget to add -forceasm

C:\XP2100FAH>FAH3Console.exe -forceasm -advmethods

Note: Please read the license agreement (FAH3Console.exe -license). Further
use of this software requires that you have read and accepted this agreement.

No directory settings found in registry. Using current directory...


--- Opening Log file [March 10 10:30:11]


# Windows Console Edition ################################################## ###
################################################## #############################

Folding@home Client Version 3.24

http://foldingathome.stanford.edu
email:help@foldingathome.stanford.edu

################################################## #############################
################################################## #############################

Arguments: -forceasm -advmethods

[10:30:11] - Ask before connecting: No
[10:30:11] - User name: veggyhed (Team 131)
[10:30:11] - User ID = 35F722F12C0764981
[10:30:11] - Machine ID: 1
[10:30:11]
[10:30:11] Loaded queue successfully.
[10:30:11] + Benchmarking ...
[10:30:14]
[10:30:14] + Processing work unit
[10:30:14] Core required: FahCore_78.exe
[10:30:14] Core found.
[10:30:14] Working on Unit 06 [March 10 10:30:14]
[10:30:14] + Working ...
[10:30:14]
[10:30:14] *------------------------------*
[10:30:14] Folding@home Gromacs Core
[10:30:14] Version 1.43 (March 4, 2003)
[10:30:14]
[10:30:14] Preparing to commence simulation
[10:30:14] - Assembly optimizations manually forced on.
[10:30:14] - Not checking prior termination.
[10:30:14] - Go method
[10:30:15] - Expanded 417561 -> 2495176 (decompressed 597.5 percent)
[10:30:15]
[10:30:15] Project: 531 (Run 55, Clone 4, Gen 3)
[10:30:15]
[10:30:15] Assembly optimizations on if available.
[10:30:15] Entering M.D.
[10:30:49] (Starting from checkpoint)
[10:30:49] Protein: p531_BBA5 in water
[10:30:49]
[10:30:49] Writing local files
[10:30:49] Completed 69000 out of 100000 steps (69)

[Bruce]

<opps> :o

Yes - - It's good to have -forceasm there, in case your UPS doesn't ride out the power losses, etc.

Alice: You should have reminded me to include BOTH advmethods and forceasm

[Louis_Berding2002]

http://www.louisberding.no-ip.com/Box1.jpg

:eek:

I can't believe this!! I've completed about two units in one day!!!! This kicks ass!!!! :D :cool:
Hmmm...I was just looking though....only 7 points for this unit. Are these units smaller, or is it just my luck that I'm getting sm,alle units with Gromacs. Are there larger Gromac units?


EDIT: Why the hell can't I insert a picture?! :banghead: It only puts a hyper link to the picture. :confused:

[Bruce]

Gromacs has had both large and small WUs - - though both GAH and Tinker have had some extremely large WUs recently. Gromacs does kick ass, though (when they finally get all the bugs out of it).

[Louis_Berding2002]

Okay...I hope my next one tonight is a much larger one. I need more points damn it! :D I need to get to the top 10 quickly before everyone else starts crunching for 131. :D :cool:

[veggyhed]

Originally posted by Louis_Berding2002
Okay...I hope my next one tonight is a much larger one. I need more points damn it! :D I need to get to the top 10 quickly before everyone else starts crunching for 131. :D :cool:

We are starting to move. We are at 129 and going up. Out daily output is way up. Come on Team 131 lets go :baa:

[Louis_Berding2002]

I'm trying my best...now with three machines crunching. :cool:

[Bruce]

There are some new Gromacs proteins.

Like I said earlier, if you have a relatively modern CPU (with SSE or 3DNow+) and don't have "-advmethods -forceasm" set as your command line switches, you should do it now. You'll get about double your production toward team 131. :D

[veggyhed]

I am loving this gromacs stuff.

Is FAH going to be switching to gromacs 100% ?

If u take a tinker WU and a gromac WU of the same protein ( if this is possible) would the output be the same even if the method of calculation is different ?

[Louis_Berding2002]

Words cannot describe how fast it's running for me at the moment. Yet, EM3 says unknown for the points for this current unit...I hope it's a lot. ;)

[Virus]

I have 4 machines running and only 2 of them are running gromacs. I made sure they are all setup the same. The two machines that it's working on are XP Professional and Win2k. 98 and XP home aren't working. Are gromac wu's random?

[veggyhed]

The server dishes them out . If u get good at reading this page http://folding.stanford.edu/serverstat.html and this page http://folding.stanford.edu/psummary.html u can sort of see what u will be folding.
My 3 machines at work are folding GAH WU`s and 1 is folding a 400 frame FAH WU.
My 2 machines at home have been nothing but gromac

[TDKozan]

Stupid question: What CPUs precisely will this work with?

TK ;>

[Bruce]

Originally posted by veggyhed
Is FAH going to be switching to gromacs 100% ?

If u take a tinker WU and a gromac WU of the same protein ( if this is possible) would the output be the same even if the method of calculation is different ?

Probably.

If you run the same simulation in Tinker and Gromacs, you will get the same result, only faster.

In fact, even if you fold the same protein, it's unlikely you would run the same simulation. Gromacs simulates the solvent explicitly. That means in most cases, a lot of water molecules are simulated, right near the protein. Tinker uses what is called an "implicit solvent" which means they don't actually simulate individual molecules, but they have equations that allow for their effects.

[Bruce]

Originally posted by Louis_Berding2002
Words cannot describe how fast it's running for me at the moment. Yet, EM3 says unknown for the points for this current unit...I hope it's a lot. ;)

There are new Gromacs proteins coming out on a regular basis. TheWeatherMan updates the files with the figures for the new proteins frequently, but he's having a tough time keeping up. You can try downloading the information and see if it's covered - - or you can look up the numbers at Stanford and edit them into the EM3 file yourself. When in doubt, http://folding.stanford.edu/psummary.html is always up-to-date.

[Bruce]

Originally posted by Virus
I have 4 machines running and only 2 of them are running gromacs. I made sure they are all setup the same. The two machines that it's working on are XP Professional and Win2k. 98 and XP home aren't working. Are gromac wu's random?

As Veggyhed said, the servers dish out whatever they have. That makes all WUs somewhat random.

If you're using the -advmethods flag, you've expressed a preference for Gromacs WU (see note below). I've seen instances where the servers dispensing Gromacs WUs run low on new WUs or a problem is identified and a project is temporarily suspended. Either way, if you can't get your preferred type of WU, you'll be sent to another server for a different type of WU.

All advanced methods testing is currently on the Gromacs core. When the GAH core was being tested, it was also done with the -advmethods flag, so it may not ALWAYS mean Gromacs - - it might change from time to time, but it is rather unlikely, and not in the near future.

[Bruce]

Originally posted by TDKozan
Stupid question: What CPUs precisely will this work with?

TK ;>

Not a stupid question. I've got a list somewhere but I'll need to find it to be really accurate.

Gromacs will work with any CPU. Only those which support the SSE instructions or the 3DNow-extended instructions will be able to use the optimizations. These instruction sets allow a single instruction to work on a large number of data elements - - hence the name SIMD instructions. Single Instruction, Multiple Data.

The Pentium 233 came out with MX instructions which were insufficient (SIMD, but for integers only). About the same time the K6-II came out with 3DNow which covered most floating point instructions - - which is what Gromacs needs. Later AMD extended it to 3DNow+ which helped even more. Meanwhile, Intel was countering with SSE and SSE+. I'm not sure if the K6-II's version is sufficient, but 3DNow-extended is, and so is SSE.

If you run Sandra or WCPUID or other benchmarking tools and it says you've got either 3DNow+ or SSE, you're in. Otherwise you'll have to wait until I find my list.

EDIT: List is at http://forum.folding-community.org/viewtopic.php?p=28470&highlight=#28470

optimizations: PIII and above and Duron and above.

Celerons are mixed: 533MHz Celerons may or may not have SSE. Faster Celerons do; slower do not.

I'm still checking on K6-II and K6-III

The P4 finally lives up to some of the original advertising claims, because good SSE programming can keep the pipeline full.

[TDKozan]

Thanks Bruce! You're a treasure.

TK ;>