Um, for some reason, i can't log in.........


I vote to delete Genome for our DC projects. There still is no mac client and if there is one that i think we would be spreading our team a bit too far.

Keep:
SETI
RC5
Folding
DC
Ubero.


5 is good....

[Shaktai]

Reader 50 & scottheking, while less than ideal testing was done, I thought you might be interested in my early DF benchmarks.

I ran the Distributed Folding text client (not to be confused with Folding @home client) on my 4 computers. The results were rather surprising. However the benchmark time period was rather short (2 hours) and there was quite a variance in memory as noted below. I may run more detailed longer term benchmarks later after I can equalize the memory situation.

The text client was used on all machines.

MAC OS-X
iBook 600 - 600mhz G3 with 100mhz bus and 256k L2 cache, 256 mb ram -- 2398 per hour or 3.99 units per hour per megahertz.

Power Mac 800 - 800mhz G4 with 133 mhz bus, 256k L2 cache, 512 mb ram -- 3889 per hour or 4.86 per hour per megahertz.

WINDOWS XP
Pentium Celeron 1 ghz with 100mhz FSB, 128 k L2 cache, 128 mb ram -- 3405 per hour or 3.4 per hour per megahertz.

AMD Athlon 1600+ (1400 mhz) with 200mhz FSB, 256 or 384k L2 cache (uncertain) 256 mb Ram -- 5394 per hour or 3.85 per megahertz per hour.

The Power Mac easily outperformed the Celeron on structures per hour, and outperformed the Athlon on a per Megahertz basis.

I have no doubt though that the Celeron was severly crippled by its small cache and smaller memory. Still, on a megahertz per megahertz basis, both Macs running OS-X outperformed their Windows XP counterparts. However the cost per megahertz was substantially less for the PC's. I was surprised that the Athlon didn't turn in a better performance, but it be that it is held back by XP or its cheap motherboard. (It is an E-machines) Dollar for dollar it was still a good deal for not building the box.

My early conclusion is that Mac's will be able to hold their own in the Distributed Folding project. Though for cheap crunching the windows machines are the more affordable choice.

Mileage will certainly vary with different operating systems, memory and motherboards.

[Edit] A fourth person has joined the DF team, but has yet to turn in any units.

[ 04-04-2002: Message edited by: Shaktai ]

[slipjack]

Dang, if we get too many more DC projects going... I'm going to have to go out and buy more machines.

I kind of like the ascii DNA strands... so 1990.

[reader50]

Most interesting. If we are really that competitive, or just on equal ground, I had better start setting up the stats.

Edit: A 5th person has joined.

[ 04-04-2002: Message edited by: reader50 ]

[Shaktai]

Originally posted by reader50:
<STRONG>Most interesting. If we are really that competitive, or just on equal ground, I had better start setting up the stats.</STRONG>

Holy Toledo, we now have 5 team members and just one hour of testing moved us up to the #185 team slot among 239 teams.

Check out this stats page although it still doesn't show all of our units. Interesting though.

My current rank is 2292 out of 6529 registered participants, and I wasn't even trying.

[Scotttheking]

That's much better then my testing.
I'll look into it a bit more now.

[Shaktai]

Originally posted by Scotttheking:
<STRONG>That's much better then my testing.
I'll look into it a bit more now.</STRONG>

I didn't look at it in detail, but it looks like there have been some updates to Linux clients. Maybe that will help.

[reader50]

It should be noted that we are passing mostly idle teams, much like the dregs at the bottom of the Ubero listings.

It would take a really serious fight to get into the Top 10, FreeDC (and others) seem to have picked dFold as the hot project to compete in. In fact, DF is where FreeDC went when they powered down on Ubero.

[malson]

I've been running the client for about 10 hours now. I'm still in the 5000 something rank because it doesn;'t show that I've uploaded any results. Is there something I missed? I know I've done a fair number of units. Maybe there just hasn't been a stats run yet.

[reader50]

Originally posted by malson:
<STRONG>I've been running the client for about 10 hours now. I'm still in the 5000 something rank because it doesn;'t show that I've uploaded any results. Is there something I missed? I know I've done a fair number of units. Maybe there just hasn't been a stats run yet.</STRONG>

As I recall, dFold wasn't very good at showing results right away sometimes. The units you have sent in are counted, they will eventually show up.

[Shaktai]

Originally posted by malson:
<STRONG>I've been running the client for about 10 hours now. I'm still in the 5000 something rank because it doesn;'t show that I've uploaded any results. Is there something I missed? I know I've done a fair number of units. Maybe there just hasn't been a stats run yet.</STRONG>

Dfold only uploads about every 5,000 to 10,000 structures completed. However you can force an upload if you are using the text client by typing Q to quit (DON'T command c or x out of the terminal) This will upload all completed units then quit the client. Once your units upload, it will take a bit for them to show. Anywhere from 10 minutes to an hour usually. How many structures have you completed? I don't know if you can force an upload with the screensaver or not, as I haven't tried it. For speed though, I reccomend the text client, no matter your platform.

BTW: Welcome to DF!!

Edited: It looks as though your numbers have posted to the team list at least.

[ 04-05-2002: Message edited by: Shaktai ]

[malson]

I guess you were right about the 10,000 mark. I'm listed on the team page but when I go to my personal stats it still shows 0 as my contribution. Oh well, maybe it will update soon. Anyway, as of right now, it says that I have 10915 structures complete. A quick calculation shows that I'm doing 100 structures every 5 minutes or a structure every 3 seconds. That's not too bad I guess especially since I have 2 instances of Ubero cranking as well.

[ 04-05-2002: Message edited by: malson ]

[slipjack]

You just came on board about 30 mins ago.

[GlobalNomad]

Well I went to the distributed folding project, and they state you need 128 mb o ram. Well it works fine with 32 .

and Team MacNN now has 7 members.

[Shaktai]

Originally posted by malson:
<STRONG>I guess you were right about the 10,000 mark. I'm listed on the team page but when I go to my personal stats it still shows 0 as my contribution. Oh well, maybe it will update soon. Anyway, as of right now, it says that I have 10915 structures complete. A quick calculation shows that I'm doing 100 structures every 5 minutes or a structure every 3 seconds. That's not too bad I guess especially since I have 2 instances of Ubero cranking as well.

[ 04-05-2002: Message edited by: malson ]</STRONG>

Yeah, my personal stats still don't show all of mine either. The team page stats are the most up-to-date right now. The personal stats don't seem to update very often.

And welcome Global Nomad. I have no idea how you got it to run in 32 mb Ram, but whatever you did congratulations.

[ 04-05-2002: Message edited by: Shaktai ]

[Scotttheking]

bumped back up because of the date screwup.

[reader50]

Originally posted by &lt;gumby5647&gt;:
<STRONG>... I vote to delete Genome for our DC projects. There still is no mac client and if there is one that i think we would be spreading our team a bit too far ...</STRONG>

Variations of that reasoning often come up on Ars and the boards of other peecee DC teams. They have more projects to pick among than we do, so the "spreading thin" arguement comes up sooner. Usually, what the arguement is really about is the SETI team members want everyone else to come back to the fold, dump those other "useless" projects.

Unfortunately, the reasoning is flawed. DC is about having fun and supporting projects that interest you. Restricting project choices for other team members would serve neither purpose -- the other team members would not find it fun being told where to crunch, and might not like any of the remaining projects. The end result? Perhaps those team members would have more fun on a different team, one that offers the project they find interesting.

I would like to support any and all projects with Mac clients that our members express interest in. If no one is interested in a project, there would be no problem, and no active site section. I would rather not be the one who decides which project is "worthy", that should be left to our members to decide individually.

Perhaps our team efforts in earlier projects will suffer a little, perhaps not. Currently, we have been increasing power on all projects at once. Each time we jump into a new project, more new people join up to enter the fray. I'd rather have them join on a new MacNN team, than join someone else's team.

One valid arguement that I know of is time limits. Specifically, mine. A number of site pages have not appeared yet, but most of that was because of stats code work that would have happened anyway, without the new projects. Things like graphs code, and assorted cosmetic bug fixes, or competitions code. That code work is essentially done, so it's back to resource pages. I'm game to tackle more projects if our members want it.

Note for the curious, a breakdown on project interest, based mostly on site page hits:
1. SETI@home
2. RC5
3. Folding@home
4. Ubero
5. Genome@home
6. Distributed Folding (emails & posted questions)

2,3, and 4 often switch places from week to week. Though RC5 has the #2 spot the most often. We get a fairly steady stream of Genome hits and emails requesting news. Stanford has decided to roll Genome into the Folding client, with a checkbox or menu choice to switch between the two projects.

[Bockie]

Originally posted by reader50:
[QB]

Variations of that reasoning often come up on Ars and the boards of other peecee DC teams. They have more projects to pick among than we do, so the "spreading thin" arguement comes up sooner. Usually, what the arguement is really about is the SETI team members want everyone else to come back to the fold, dump those other "useless" projects.

&lt;snip&gt;

I would like to support any and all projects with Mac clients that our members express interest in. If no one is interested in a project, there would be no problem, and no active site section. I would rather not be the one who decides which project is "worthy", that should be left to our members to decide individually.

&lt;snip&gt;[QB]

PC DC teams have more projects/clients to choose from because they're (about) 95% of the market. The presence of Mac teams as top producers shows that Macs aren't just cool, graphic artist machines. The presence of Mac clients show interest/belief in the Mac platform and I believe we should qualify the programmers' beliefs.

[Shaktai]

In DF, our current team rank is 145. As reader 50 mentioned, up until now we have been pretty much passing the dregs or non active teams, and we have a long ways to go to get near the top producers, but still we have had some very impressive production for our first 24 hours.

This is a project that looks like it is suitable for any computer that can run OS-X, without serious handicaps, regardless of CPU speed. We now have 9 team members and growing.

<font color = red>Slipjack, don't get too comfortable in that #1 team spot. I'm powering up and am going to be right on your tail. </font>

The project people have at least expressed an interest in looking at if it is worthwhile to add altivec support. Doesn't mean they will, or if it is worth it to them to do so, but if you have a G4 let them know you are interested. The more interest, the better the chance we might see it.

Post your interest at the forum here.

It would be nice to have another project where our G4's can shine.

[slipjack]

Originally posted by Shaktai:
<STRONG>

<font color = red>Slipjack, don't get too comfortable in that #1 team spot. I'm powering up and am going to be right on your tail. </font>

Post your interest at the forum here.

It would be nice to have another project where our G4's can shine.</STRONG>

Heh heh, them's fightin' words

I'm pretty much already using all the power that I want to for DF... but maybe I'll transition another comp or two just to keep you at arm's length.

[slipjack]

Famous last words.. you're already catchin' up big time... d'oh!

[Shaktai]

Originally posted by slipjack:
<STRONG>Famous last words.. you're already catchin' up big time... d'oh!</STRONG>

I promise not to throw everything at you all at one time. In fact right now I am taking my computers down, one at a time, for software updates. Of course, then I'll bring them back up.

Hey, this may be my only shot at number one. Besides we never know when one of the big guns will decide to hop in and blow us both out of the water. Figure I might as well take advantage of the situation while I can. My four little computers can't compete against the big farms.

[slipjack]

Exactly. Which is why I like it where I am.

[reader50]

[catcall mode]
I was there first. And for a longer time.
[/catcall mode]

[Shaktai]

Originally posted by reader50:
<STRONG>[catcall mode]
I was there first. And for a longer time.
[/catcall mode]</STRONG>

I will give you that!

Looks like I fell back a little while I was updating software on all mine. Was trying the 1 gig Celeron on Ubero, but so far it is not doing very well. Only slightly better than my iBook 600. Think I will pull it off and put it back on DF until I can boost the memory and figure out how to do the tweaks and run terminal mode on a windows machine. No sense wasting the cycles when I could be passing slipjack.

Ubero fans should read Scott Hodson's comments regarding what we have to look forward to with Ubero.

[slipjack]

Just to be fair Shaktai, here's what I'm crunching DF with:

1 Ghz G4 DP
350 Mhz G4
600 Mhz PIII

Which, should put me just behind you in pace - even worse, I can't run the big boy all day/night, my wife won't allow it.

(I did last night though. )

[ 04-06-2002: Message edited by: slipjack ]

[Shaktai]

Originally posted by slipjack:
<STRONG>Just to be fair Shaktai, here's what I'm crunching DF with:

1 Ghz G4 DP
350 Mhz G4
600 Mhz PIII

Which, should put me just behind you in pace - even worse, I can't run the big boy all day/night, my wife won't allow it.

(I did last night though. )

[ 04-06-2002: Message edited by: slipjack ]</STRONG>

Heh! I kind of figured the dual gig was in there. That goes against my 800 single processor Power Mac, 600 iBook, 1 gig celeron and Athlon 1.4 gig. Of course I don't neccessarily run them all at once, and the iBook and Power Mac are the machines I do all my real work on. The PC's are just for crunching.

Edit Update: Our team rank is now 118 and we are up to 12 members. Our total production to date is 642,089. This is very impressive for less than 36 hours, and most of this has taken place in the past 24 hours.

Also the benchmarks I started this thread with, seem to be holding up pretty well over a 12 hour period, give or take 5% variance per hour. It continues to look like this is a good project for Macs of any speed. I would be interested in seeing what kinds of results other folks are getting.

[ 04-06-2002: Message edited by: Shaktai ]

[wheeles]

I have had a quick play with this client. The thing I like about it is that you don't have to wait all day to get one WU like Seti or F@H. OK so this merely an illusion of greater progress, and it doesn't mean you are doing any more work than F@H but it is more likely to keep you interested.

Not sure whether you really need the rather silly 'Wiggly Worm' graphics on the command line version. All you really need is a progress indicator. Shame you can't just turn off Wiggly Worm and keep the progress indicator. If you make the client quiet you then have to dive into the progress.txt file to see whether you have finished all the work off.

Slipjack, don't get too comfortable in that #1 team spot. I'm powering up and am going to be right on your tail.

And there was me thinking I could slide my way to the top while you guys had your attentions elsewhere.

They say great minds think alike. But I guess devious ones do too.

[Shaktai]

Originally posted by wheeles:
<STRONG>And there was me thinking I could slide my way to the top while you guys had your attentions elsewhere.

They say great minds think alike. But I guess devious ones do too. </STRONG>

Hey, give it your best shot. The more the merrier. It keeps everyone on their toes.

We are now in the top 100 teams (#96) after only 48 hours and up to 14 members, 12 of which are active. It will be slower going from here on out, but not impossible. Let's keep crunching.

I hope to provide some updated longer term benchmarks tonight.

[ 04-06-2002: Message edited by: Shaktai ]

[Shaktai]

UPDATED BENCHMARKS

I started this thread with some preliminary benchmarks, but they were from very short term testing only. Below are updated benchmarks for a period of approx 18 hours (varies slightly from machine to machine) In the longer term benchmark the Athlon performed much better and so did the ibook. The Power Mac 800 and Celeron 1gz remained approx the same. Again, the Celeron is probably handicapped by a smaller L2 cache, and only 128 mb RAM. I plan to increase the RAM and then benchmark it again later.

iBook 600 mhz G3 - 2630 per hr, 4.38 per mhz per hr.
Power Mac 800 G4 - 3840 per hr, 4.8 per mhz per hr.
Celeron 1 ghz - 3431 per hr, 3.43 per mhz per hr.
Athlon 1600+ 1.4ghz - 6290 per hr, 4.49 pr mhz per hr.

It still appears that the Macs hold their own quite well in this project. I would like to see what kind of performance someone with a dual 1 gig powermac with its larger cache could get.

To test: start the client and let it run continously for several hours. At the end of your test period, display the terminal window and open up the process monitor (in the utilities folder), highlight the foldtrajlite process and then click the "more info" triangle at the bottom of the process monitor window. Then click the status tab. It will show how long the process has been running in minutes. Record the time, and then in the terminal window at the top left, record the # after "beginning structure. This is the total number of structures completed since you last started the process in the terminal window.

Divide the total minutes by 60 to convert it to hours. Divide the total structures by the number of hours and this will give you the number per hour. Divide that number by your CPU's mhz and that will give you the number per mhz per hour.

SPEED UP TIP: Do not minimize the terminal window to the dock. Because the dock image is animated, the dock will steal a lot of cpu cycles from your process. Either leave the window displayed or use the Hide command (Command H) in the menu.

[reader50]

Shaktai,

Have you tried renice-ing the crunch core, to see if you could push the speed up?

Note that this client runs fine in &gt;console mode, that would be another way to speed it up.

[ 04-07-2002: Message edited by: reader50 ]

[Shaktai]

Originally posted by reader50:
<STRONG>Shaktai,

Have you tried renice-ing the crunch core, to see if you could push the speed up?

Note that this client runs fine in &gt;console mode, that would be another way to speed it up.

[ 04-07-2002: Message edited by: reader50 ]</STRONG>

No I haven't tried anything special. I am new enough to the terminal, that I am not really certain how to do some of that yet. How would I run it in console mode, so that the terminal app wouldn't have to be left running. That should defininitely improve performance as the terminal program uses quite a few clock cycles at intermittent intervals. As to renicing it, the benchmarks were pretty much with the computers left mostly untouched. Is it possible to do this on the PC's as well? The operating systems are OS-X and Windows XP.

The PC's ran without interruption, but the Macs both had some usage for email and web during the time, but it was not extensive (maybe an hour each) so if anything, they were at a disadvantage during the testing period. For instance my iBook is crunching even while I type this. I am amazed at how well it does. With the terminal displayed I can see it crunching away even as I type, although a little slower.

(I am Mac savvy, but still a rookie when it comes to command lines.)

[SkiBikeSki]

Here are my benchmarks. After roughly 24 hours my 867 did 66,000 structures. That works out to 2750 per hour, and 3.2 per MHZ per hour. However I was doing other things with it during parts of the day, and I had it minimized in the dock, so then the dock at up 30% of the CPU. Someday soon I'll let it run in &gt;console mode for a good period of time and see what happens then.

[ 04-07-2002: Message edited by: SkiBikeSki ]

[reader50]

Originally posted by Shaktai:
<STRONG>... How would I run it in console mode, so that the terminal app wouldn't have to be left running ...</STRONG>

&gt;console mode is mostly suitable for unattended comps or overnight crunching, since the entire GUI is turned off. You can see the SETI &gt;console mode instructions here about 2/3 of the way down. dFold works the same way -- except for the Control-C to exit. Don't follow that part of the instructions.

If you su to root, you can renice a running process to a higher priority. Setting the process to a very high priority (say, negative 10 or negative 15) would give it a boost, cutting back CPU from non-critical system functions, and other user processess. The details are rather poorly explained under "man renice".

If you login to &gt;console as the root user (or su to root), it would be more convenient to "nice" a process on launch to higher priorities, since you can't open another Terminal window to fool around with the first process. Example: "./dfold -nice -18"

The example assumes the dFold launch file is named "dfold" and that you want it to have almost unlimited CPU. The nice scale goes from +20 (lowest priority) to 0 (normal) to -20 (max priority). A normal user can set below-normal priorities on processess that they own. Only the root user can set above-normal priorities.

In the case of dFold, the file you start is actually a startup script for the real client "foldtrajlite". It would be best to put the nice command inside the launch file, which was called "foldit" before I renamed it, I think. Example file contents, modified:

#!/bin/sh
./foldtrajlite -nice -18 -f protein -n native -df

[ 04-07-2002: Message edited by: reader50 ]

[Shaktai]

Console probably wouldn't work for me then. I am constantly back and forth on my Macs at all kinds of odd times, so it would probably be more hassle than it is worth. Re-nicing though works in the terminal. (thanks) I might play around with a chrontab for it too. I need to learn more about the Unix commands.

[ 04-07-2002: Message edited by: Shaktai ]

[wheeles]

I've found that when I set the -qt flag it didn't seem to respond to me shutting down with the 'Q' key. Maybe it was just being 'quiet' and trying to upload files, but it didn't do much for a while. Instead I had to delete the foldtrajlite.lock file instead to stop the process cleanly.

For a dual cpu box I have set up 2 directories - one for each cpu. Both have all the software and files in each of them. I renamed the foldit scripts to dfold1 and dfold2 so I could tell which was which more easily with ps or the ProcessViewer. This works well. Also I've noticed there is less impact on me using the box with these two running than say Ubero which brings everything to its knees if I run 2 of those. Think this might be more down to Java hogging memory.

As the command line gets tied up by dfold I haven't tried the &gt;console yet. I would have to background the processes as I wouldn't be able to easily run 2 of those. Then it would be a case of deleting the lock files to stop them. Also If on a dialup and not connected when shutting the processes down, the process take an age to realise their is no connection before they finally give up trying. Fortunately, they do keep the work they have done and try to upload it next time you start them up.

I'l leave this machine untouched for most of the day and see how my box performs.

One last thing to note, I keep my Dock hidden off the right side of my screen (I gave up using it much when I got LaunchBar and Quitling). If I do minimise a process then it's effectively hidden as well. I trust the Dock is aware it doesn't need to draw anything when hidden. Not sure if that's the case, though.

[Shaktai]

Originally posted by wheeles:
<STRONG> I trust the Dock is aware it doesn't need to draw anything when hidden. Not sure if that's the case, though.</STRONG>

I was observing it in process viewer and with the terminal minimized to the dock, the dock was drawing nearly 30% of the CPU cycles. However, with the terminal "hidden" using the hidden command, the dock used 0 cycles.

Was off-line here for over an hour, when a power outage freaked out my DSL router. After a half hour with tech support, finally got it straightened out. (Note to self: got to get the DSL router on the UPS). Unfortunately, I did lose some work on my PC's due to the outage. I need to get them on UPS as well. My macs were fine though. Lost a good hour or two of work though on both of the PC's.

I look forward to hearing what kind of performance you get as well.

Started foldit on my iBook with a nice of -5 to see if it makes any difference. (some of this unix stuff is starting to sink in.)

[Shaktai]

Team rank is now 75, personal rank is 670 and we should break 2,000,000 structures very shortly. Not bad for a team with only 15 active members and less than 2.5 days crunching.

We have moved ahead of "Apple Macintosh Crunchers" and RealMac, by quite a bit.

[Shaktai]

Originally posted by SkiBikeSki:
<STRONG>However I was doing other things with it during parts of the day, and I had it minimized in the dock, so then the dock at up 30% of the CPU.

[ 04-07-2002: Message edited by: SkiBikeSki ]</STRONG>

If you "hide it using the hide command in the menu instead of minimizing it, then the dock won't eat up the cpu.

[The Placid Casual]

Just a note to say that I have just joined the Foldit effort...

For the moment I'm running the client on my 800DP in conjunction with RC5 so won't be expecting too much in terms of stats for the moment until RC5 is finally cracked...

I've also just set up an iMac G3 450 to work on Seti...it has done a few units (about 29 hours each) and is left on 24/7. Is it worth keeping it on Seti or would it best used elsewhere? any info would be appreciated....

Marc

[reader50]

Originally posted by Marc2211:
<STRONG>... I've also just set up an iMac G3 450 to work on Seti...it has done a few units (about 29 hours each) and is left on 24/7. Is it worth keeping it on Seti or would it best used elsewhere? any info would be appreciated....

Marc</STRONG>

That is a personal decision, do what you feel is best. From a team standpoint, the SETI team is having the hardest time advancing. But no one is "required" to crunch for a particular project. Crunch whatever catches your interest, and have fun.

Note, your 29 hours per WU seem way too long on SETI. For your machine, I would expect times around 10-15 hours per unit. You might want to look over the optimization pages for SETI on the team site. They are linked from the SETI News page under "Resources", by each client type.

[The Placid Casual]

Originally posted by reader50:
<STRONG>

That is a personal decision, do what you feel is best. From a team standpoint, the SETI team is having the hardest time advancing. But no one is "required" to crunch for a particular project. Crunch whatever catches your interest, and have fun.

Note, your 29 hours per WU seem way too long on SETI. For your machine, I would expect times around 10-15 hours per unit. You might want to look over the optimization pages for SETI on the team site. They are linked from the SETI News page under "Resources", by each client type.</STRONG>

Cheers for the info I was just enquiring on a technical basis....I have a few G3s that I really don't mind what they crunch as long as it will cause some benefit (no matter how small!) to the team somewhere along the line, I figured SETI was the best bet....and the whole SETI ideal has appealed to me for ages! I have only the iMac set up for Seti at present but I will add some more Macs later. However, my DP is reserved for most of the fun crunching (RC5 + foldit!)!!

As for the unit time, thanks for the link....I'm going to have a look into it this afternoon and sort it all out. I guess I'm so used to the G4 DP being much faster than the G3 I thought that the unit time span was normal

I have also just found a source for some cheap components to build a few dedicated crunching PCs...If I can find a linux flavour I like, I may just get to it!I It looks as though I'm getting quite hooked to this whole thing

Cheers,

Marc

[ 04-07-2002: Message edited by: Marc2211 ]

[wheeles]

Originally posted by wheeles:
<STRONG>
I'l leave this machine untouched for most of the day and see how my box performs.
</STRONG>

This was the first time I have left this to run on its own for any significant period without me using my Mac. In an attempt to be environmentally friendly, I had set the machine to sleep the display and disk after about 20 minutes. When I came back about 10 hours later the display was blank and the disk quiet, as expected.

I managed to wake the disk up but not the screen. In the end I had to reset the machine.

I have now turned off the energy saver (I'll just turn the monitor off manually in future). One thing I have noted is that when I fired up both processes they had managed to keep 20k worth of work. This makes me think that every time they fail to upload the last 10k they then checkpoint the work to disk.

I'll give it another go tomorrow and see if I have more success.

[Shaktai]

Originally posted by Marc2211:
<STRONG>Just a note to say that I have just joined the Foldit effort...

For the moment I'm running the client on my 800DP in conjunction with RC5 so won't be expecting too much in terms of stats for the moment until RC5 is finally cracked...

I've also just set up an iMac G3 450 to work on Seti...it has done a few units (about 29 hours each) and is left on 24/7. Is it worth keeping it on Seti or would it best used elsewhere? any info would be appreciated....

Marc</STRONG>

My old iMac 400 used to do a Seti work unit in 19-20 hours, under OS-9, so you should be able to do better.

As far as projects and your production, don't worry about it. It looks like your DP800 is doing pretty well on DF even sharing with RC-5. Just do whatever project(s) you like the best. They all need help. Your efforts will be appreciated wherever you direct them.

[Grozni Majmun]

Yup, the quiet mode does speed it up quite a bit too. The terminal eats anything from 5-10% of my cpu just to draw the animations. Using quiet mode (adding -qt at the end of your ./foldit file) eliminates that.

For me the speedup is quite spectacular. See this screenshot from process viewer: scrnshot.jpg


100.8%!!
I even tried &gt;console mode, but I think I'll stick with this. As wheeles said, the problem with that is that you can't quit it with q. You just have to trash the file foldtrajlite.lock from the distribfold folder. That way it exits nicely and uploads the completed units first.

In my experience hiding the terminal with apple-H while it is drawing the ascii art does not cause it to use fewer CPU cycles than if it is shown.

Sorry if this was all old had, but I thought the 100.8% was so cool I had to post it.

[Shaktai]

Originally posted by Grozni Majmun:
<STRONG>Yup, the quiet mode does speed it up quite a bit too. The terminal eats anything from 5-10% of my cpu just to draw the animations. Using quiet mode (adding -qt at the end of your ./foldit file) eliminates that.

For me the speedup is quite spectacular.


100.8%!!
I even tried &gt;console mode, but I think I'll stick with this. As wheeles said, the problem with that is that you can't quit it with q. You just have to trash the file foldtrajlite.lock from the distribfold folder. That way it exits nicely and uploads the completed units first.

In my experience hiding the terminal with apple-H while it is drawing the ascii art does not cause it to use fewer CPU cycles than if it is shown.

Sorry if this was all old had, but I thought the 100.8% was so cool I had to post it. </STRONG>

Definitely cool. I am going to give it a try. I hardly ever need to quit now anyway, so if it gives a few more cpu cycles it is worth it. Hey, maybe even my iBook 600 could break 3000 per hour or at least get close to it.

[slipjack]

I tried the -qt thing, but it still draws the ascii art... is that how you write it? ./foldit -qt ?

It didn't work for me. Also, any tips on how to run &gt;console mode? I've got a G4 that's just crunching right now and I might as well save a few cycles.

[reader50]

You have to add the "-qt" inside the foldit script (the file you launch is actually a launch script).

To run in &gt;console mode, look over the instructions for &gt;console on SETI or RC5 (CLI optimizations). Just remember not to quit it with Control-C (as the SETI/RC5 instructions call for).

[Grozni Majmun]

As reader50 said, the foldit file is just a script. This is what mine looks like at the moment:

<BLOCKQUOTE><font size="1"face="Geneva, Verdana, Arial">code:</font><HR><pre><font size=1 face=courier>
#!/bin/sh
nohup ./foldtrajlite -f protein -n native -qt &
</font>[/code]

That way I can quit the terminal, logout, login, go into &gt;console mode, whatever, and the DF client just keeps running.

[The Placid Casual]

Originally posted by Shaktai:
<STRONG>

My old iMac 400 used to do a Seti work unit in 19-20 hours, under OS-9, so you should be able to do better.

As far as projects and your production, don't worry about it. It looks like your DP800 is doing pretty well on DF even sharing with RC-5. Just do whatever project(s) you like the best. They all need help. Your efforts will be appreciated wherever you direct them.</STRONG>

I have had a play with the SETI iMac and now have the time down to about 17-18 hours...I think I may be able to know an hour or 2 off that as well when I have abit more time....if only I didn't have to work!!

The folding project appeals to me more than the others at the moment because It may do some good for people somewhere along the line, and you can see discernable results...I like the 'target/aim' aspect of it all. I'm going to stick with RC5 until the end, but after that may dedicate all my efforts to folding...

I am getting a new Ti for my work when they are announced, and also *may* build a dedicated cruncher or two when I get some time. However it will be the summer before I'll be able to do much about it all....if only I didn't have to work for a living!!

Marc