[Brendonshire]

I know that the current version of the OSX client is new, but does anybody else notice that it is slow?? This thing crawls. I know that the KING has no problems, but I guess that you are not using Macs right?? Maybe Athlon/Linux x() combo, with cheese Anyone else having problems?? This current molecule is kicking my 500MHz cubes ass as well as the TiPowerBook 550MHz. The Ag29a molecule has been crunched for 24Hrs now and I'm only 22% done on the Ti 550, both have 768MB and no other software running. Am I better off waiting for a better client or have they sold their soul to Intel who is backing the study. I wonder if we will ever get a decent client out of this. Is there any leverage that we have??

tyhuffman

[reader50]

G4 350 here.

I got the GNRA Hairpin on both work units that I crunched. About 48 CPU hours each time.

We have not "figured out" the Folding client yet, so there are no optimization pages ... yet. If you or anyone else can get a handle on how to speed it up, we can post it up.

[Brendonshire]

Originally posted by reader50:
<STRONG>G4 350 here.

We have not "figured out" the Folding client yet, so there are no optimization pages ... yet. If you or anyone else can get a handle on how to speed it up, we can post it up. </STRONG>

I know that the optimization does not have anything to do with priority, currently the client can take as much as 80+% of system resources. The speed will have to come from altivec and I think that they won't support that if I remember ViJay correctly. However, isn't it possible to use an algorithm to get a double precision float from a single precision Altivec unit. Like calculate twice and compare results?? Or something like that.
If it is something like that maybe it can be written for them, and they just paste it in, I know it sounds sooo easy! I would think that Apple has examples of this posted, is the G4 that slow?? I wonder if Apple is interested in working with another group on a protein project these guys seem to wish to talk to Intel and that is it. If you have problems with a Wintel system they are right on it, problems with a Mac or Unix and you may as well be shouting at the wind. Have you or others noticed this??

TyHuffman

[Scotttheking]

there is a 3 day timelimit on returns, so if the WU isn't back by then, it doesn't count.

Yep, linux athlon combo
LinAMD machines

But I turned folding off now. Working on ubero now.

[Brendonshire]

Originally posted by Scotttheking:
<STRONG>there is a 3 day timelimit on returns, so if the WU isn't back by then, it doesn't count.

Yep, linux athlon combo
LinAMD machines

But I turned folding off now. Working on ubero now.</STRONG>

Would be so kind as to recommend a machine or a board??

I was running RedHat but gave it up when I bought the Mac.

TyHuffman

[reader50]

Originally posted by Scotttheking:
<STRONG>there is a 3 day timelimit on returns, so if the WU isn't back by then, it doesn't count...</STRONG>

The timelimit varies by the specific protein that you are working on. The GNRA Hairpin that I worked on, for example, has a 6-day time limit. The Folding@home site gives the time limits on each protein that they use, along with how many stats units each protein is worth.

If you do exceed the return time limit, the unit will still count in stats terms, you will receive credit for it. The Folding project will have reissued the unit to someone else by then, so they may not actually use your returned results. Or they may use it, you will likely have still gotten it in before the next person has finished it.

[Osirisis]

I've been considering helping with this and dumping RC5, but it doesn't sound like the software has been very optimized yet. Time to wait for the real release of the software before I dump RC5

[Arkham_c]

This actually sounds like a neat concept. Are they just passing the same data out to everyone, or do they have real datasets?

[Brendonshire]

Originally posted by Osirisis:
<STRONG>I've been considering helping with this and dumping RC5, but it doesn't sound like the software has been very optimized yet. Time to wait for the real release of the software before I dump RC5 </STRONG>

I was about to do the same, dump Folding and jump to RC5, however it seems that we are all doing projects that run 9 days. That's right 9 days. At this rate the TiBook (550) will be able to complete, and the Cube may still make after all of my "help". We will see but there are more optimizations that are low lying fruit, I think that Apple is working on this as well as Alsoft. Alsoft makes the compiler, and it is new to OSX, that usually means that it will take a while to optimize it to OSX and the G4 Altivec, I would think that Apple will help them. I think that Apple is looking to use Altivec as more than a DSP. I mean that the compiler would use it automatically if possible, because we all know that it would not pay for companies to hand tweak their to use Altivec. I think that Apple got a big dose of this with 10.1.0. So I think that Apple would rather optimize the compiler than going through the code. Making tools is more prudent, however, what comes out is only as good as the tools. It would be good if the optimize for Altivec option were not an option, the compiler just did it. Fortran is the language of Folding and other programs like it in BioTech so I think that Apple will be looking to help. Alsoft makes the Fortran compiler that Folding used and maybe the same compiler that Apple used for the BLAST project. What is interesting is that the BLAST libraries are available from Alsoft for a fee, which shows that Apple is interested. Alsoft is headed for a great product in the Fortran compiler.

From the outside it is looking like Apple is getting its ducks in a row to make a run at the workstation market in 2003 and maybe late 2002, but 2003 for sure. That means 3D rendering as well as BioTech as well as movie FX.

So if you are interested in doing something worthwhile join the team, I think that we would all enjoy the company. I look for the client to be recompiled in 3 to 6 mos. Until then I'll race ya!

Ty

[Scotttheking]

Originally posted by Arkham_c:
<STRONG>This actually sounds like a neat concept. Are they just passing the same data out to everyone, or do they have real datasets?</STRONG>

Real dataset.

Has anyone received the Protein A work unit. (Its considered a completed project...not even on the "all projects" list). It wont cruch at all. I had Folding@home on for 6 hours and the process bar never moved off 0.0%. I deleted the program, re-downloaded it and was stuck with the same protein, the same brick wall. Anyone else have this problem? Any suggesstions?

Jeremy

[Brendonshire]

Originally posted by &lt;Jeremy&gt;:
<STRONG>Has anyone received the Protein A work unit. (Its considered a completed project...not even on the "all projects" list). It wont cruch at all. I had Folding@home on for 6 hours and the process bar never moved off 0.0%. I deleted the program, re-downloaded it and was stuck with the same protein, the same brick wall. Anyone else have this problem? Any suggesstions?

Jeremy</STRONG>

Got two Protein Ag29a that are being crunched I think that they are 9 day projects. I think that these are valid proteins though.

Ty

[MrRick]

I am working a proteinA work unit. It has taken almost 14 hours and I am 75% complete. I'm not sure how long it takes on a Mac, though. This computer is a 1.4Ghz Athlon running XP.

Rick

[ 03-06-2002: Message edited by: MrRick ]