Posts by Shaktai

21) Message boards : Number crunching : Mac OS - status on 10.3.9 app? (Message 1367) Posted 15 Oct 2005 by Shaktai Post: Just curious if any progress has been made on compiling for Mac OS 10.3.9? I know a few people who are interested but don't want to invest in 10.4.x just to run Rosetta.
22) Questions and Answers : Macintosh : WU stuck at 83.33% (Message 1362) Posted 14 Oct 2005 by Shaktai Post: Can you provide some information about your computer. Speed, processor, etc. 24 hours is a long time, but we are seeing some larger proteins. Is it dedicated to Rosetta, or sharing time with other projects? The calculations become more complex at 83.33%, and do take longer. If you are sharing between projects, you need to either "keep in memory" or increase the time between project switches. This ensures that each step can complete. If the time is too short and it can't reach the next step which is 91.66%, then it will keep restarting at 83.33% when it switches back from another project. On my iBook 600 (G3) I keeping switching interval set to 180 minutes. My longest work unit was 16 hours, and that was unusual. Oh, and once in a while a work unit will get "stuck". A simple quit and restart will usually fix that.
23) Message boards : Number crunching : can it really be so slow ? (Message 1325) Posted 14 Oct 2005 by Shaktai Post: I still get them periodically, and the last couple of days mostly on my P4 with HT, instead of the dual core machines. Seems to go in spurts. I wonder if it is a particular type of protein?
24) Message boards : Number crunching : Way too much credit? (Message 1245) Posted 11 Oct 2005 by Shaktai Post: Their P4 3.4ghz are more a bit more then double what my P4 3.4ghz benchmarks with HT on And your P4 3.4GHz benchmarks are 1.5 times mine with HT on. I'm curious how your benchmarks can be so much faster than mine (what's wrong with MY system) I am running the 4.72 beta client, because it seems to have fewer errors and less frequent 1% freezes on my dual core processors. Mine Measured floating point speed 2188.18 million ops/sec Measured integer speed 2641.87 million ops/sec Yours Measured floating point speed 1767.55 million ops/sec Measured integer speed 2384.02 million ops/sec Actually it looks like I am about 16% faster, not 50%, unless my math is all messed up. Could you have had some kind of background process running when the benchmarks ran? (I know, probably a silly question, but the only one I can think of at the moment.) My box is a fully dedicated cruncher with a Pentium 4 550, 1meg L2 cache, no overclocking. Just a stock Gateway box. Mobo is Intel, 512 mb PC 3200 RAM. Windows XP Home Edition. Only other things running are Trend Micro PC-cillan Internet Security and Spyware doctor. CPU-Z shows it running at 3400.1. Pretty much right on.
25) Message boards : Number crunching : Way too much credit? (Message 1242) Posted 11 Oct 2005 by Shaktai Post: Their P4 3.4ghz are more a bit more then double what my P4 3.4ghz benchmarks with HT on And your P4 3.4GHz benchmarks are 1.5 times mine with HT on. I'm curious how your benchmarks can be so much faster than mine (what's wrong with MY system) Personally, for a project that doesn't require validating work units (to equalize the scoring) I think Rosetta should just reward a set number of points per work unit as climate prediction does. The only thing is... How many credits? Unless the project could allocate different amounts of credit for different types of work units (as CPDN do with Sulphur runs), I don't think this would be a solution. But even if it can be done, I expect it to be a lot of work for the project team - time better spent on more important issues. Hmmm! Good point. Gonna have to give it a lot more thought. Guess nothing comes easy. Probably isn't a perfectly fair system because of all the variances in system, and operating systems, but should be some type of a reasonable compromise. Don't know what it would be. Every idea so far appears to be more flawed then even the original BOINC system.
26) Message boards : Number crunching : Way too much credit? (Message 1236) Posted 11 Oct 2005 by Shaktai Post: Well, just looking at it, I would say that the person benchmarked with HT off, and then turned HT back on but didn't re-benchmark. Their P4 3.4ghz are more a bit more then double what my P4 3.4ghz benchmarks with HT on, that is why I am guessing that. Might be intentional or might be accidental. They are also probably using an optimized BOINC client. Looking at their highs and lows, it looks like they may be running more then one project. Some of their high days are the result of not turning anything in the day before. If you look at their RAC though, they are below you. Still, it is a bit odd, and the benchmarks are inflated by quite a bit. Your 3800 seems to be running about where it should compared to my 4200, so you are not under benchmarking. BOINC benchmarking is not ideal at present. Personally, for a project that doesn't require validating work units (to equalize the scoring) I think Rosetta should just reward a set number of points per work unit as climate prediction does. That would put everyone on equal ground, and the BOINC benchmarks couldn't be used to mess with scoring. On projects with multiple validating units, it tends to even things out, because the heavily optimized clients never really get their claimed credit. Of course, that means I might get fewer points too, but at least we would all be on equal ground.
27) Message boards : Number crunching : can it really be so slow ? (Message 1110) Posted 8 Oct 2005 by Shaktai Post: I've been watching my Windows PC's closer and the only machines that seem to get stuck at 1% are the dual core boxes, both the AMD and the Intel. My P4 3.4 ghz with HT hasn't had the problem yet that I've caught it, and none of the single CPU boxes have had it recently. The dual core boxes are experiencing it almost daily now, where one of the two units will get stuck at 1%. Each time, a simple restart of BOINC fixes it. No, i've just got one HT machine, and it get stuck this morning at 83.33% There are two different problems. One where it gets stuck at 1%, and the one where it gets stuck at 83.33%. For many folks the solution to the 83.33% issue was to "leave application in memory" and if running more then one project, to extend the time between "switches" from the default of 60 minutes, to 90-180 minutes (depending on the speed of the machine). With the 83.33%, the calculations become much more complex and take longer. If you are switching between projects every 60 minutes, then it may not reach the next step (91.66%) before the switch and then will restart at the 83.33% when it switches back. David Kim is looking at both of these issues for fixes.
28) Message boards : Number crunching : can it really be so slow ? (Message 1081) Posted 7 Oct 2005 by Shaktai Post: I've been watching my Windows PC's closer and the only machines that seem to get stuck at 1% are the dual core boxes, both the AMD and the Intel. My P4 3.4 ghz with HT hasn't had the problem yet that I've caught it, and none of the single CPU boxes have had it recently. The dual core boxes are experiencing it almost daily now, where one of the two units will get stuck at 1%. Each time, a simple restart of BOINC fixes it.
29) Message boards : Number crunching : Rosetta crashes every time BOINC menubar trys to start it. (Message 1060) Posted 7 Oct 2005 by Shaktai Post: Hi there. I'm on MacOS 10.3.9 running BOINC_4.44_Menubar. Every single time Rosetta gets it's turn from the amongst the other projects, it crashes. BOINC_4.44_Menubar is unaffected, and just moves on to the next project. Still, it's a bit disheartening to wake up and find a dozen crash notifications on my display. Any help getting Rosetta running smoothly on my system would be greatly appreciated. I can provide crash logs if necessary. Regards, Jamie Kahn Genet Currently the Rosetta app only supports 10.4.x OS. The project team is looking into making it compatible with 10.3.9.
30) Message boards : Number crunching : Stats (Message 1046) Posted 7 Oct 2005 by Shaktai Post: Nicely done. I like the charts.
31) Message boards : Number crunching : Client Errors (Message 1045) Posted 7 Oct 2005 by Shaktai Post: There may be a performance trade-off. I haven't yet had the time to look into this. Typically we haven't seen a noticable performance trade off on other projects. The 10.3.x compiles have generally been very successful. However, there has been no success with compiling BOINC for 10.2.x or earlier. The cross development SDK has usually worked well. If you create a cross development compile, I think our team can scrounge up some testers. Just curious, have you tried Xcode2's gcc 4 auto vectorization function? It has helped sometimes with G4 and G5 processors that have altivec capabilities. Of course, I am not a coder, just a user. A 10.3.9 compatible version will draw a lot of new mac users though. Upgrades from earlier 10.3.x versions to 10.3.9 are free, so most users have or will upgrade.
32) Message boards : Cafe Rosetta : Teammember recruiting thread (No. 1) (Message 1044) Posted 7 Oct 2005 by Shaktai Post: And for all Mac Users, want to be Mac users, or folks that just like to belong to a really great and friendly team, there is.....(drum roll)...... Team MacNN
33) Message boards : Number crunching : BOINC 5+ upgrade (Message 1001) Posted 6 Oct 2005 by Shaktai Post: Now David, You do realize that you have just upset the status quo for all BOINC projects. A major upgrade in less time then you set the expectation for, and no problems to be found? That is unheard of!!! Here I was expecting you to be down for at least 24 hours. I don't know if this project can survive with everything running so smoothly. ;-) (amazing job, can't find a problem anywhere.)
34) Message boards : Number crunching : can it really be so slow ? (Message 966) Posted 5 Oct 2005 by Shaktai Post: I'm looking into it. Can everyone on this thread (no one else as I don't want too many emails) who is having the problem send me the stdout.txt files in slot0 and slot1 (if you are running dual cpu) in the BOINC installation? dekim at u.washington.edu. Thanks, David K You've got mail. 2 computers. One a Pentium D 840 dual core and the other an AMD 64 X2 4200 dual core. Both just happened to have 1 each work unit frozen at 1%. One for 6 hours and 1 for 18 hours. Hope that helps.
35) Message boards : Number crunching : can it really be so slow ? (Message 934) Posted 4 Oct 2005 by Shaktai Post: I have seen them stuck at 1% on both Windows and Mac. Seems to occur more often on the dual core boxes, but that could just be because they crunch more units. Had one stuck today for about 12 hours on my AMD 64 X2 4200. A simple quit and restart of BOINC has fixed the problem every time. maybe 4-5 a week for me.
36) Message boards : Number crunching : can it really be so slow ? (Message 872) Posted 1 Oct 2005 by Shaktai Post: I had a WU stuck at 83.66% for 5 hours (2.8GHz P4). It suddenly jumped to 91.66, and then back to 83.66% per 5 or 6 times. I had to kill it. Are you running more then one project on that computer, so that it is switching projects. If so, under general preferences try increasing the "time" between project switches to 120 minutes instead of the default 60. From 83.33% on, the calculations are more complex and take longer. Another option is to "leave in memory". I had this same problem, and increasing the time between switches solved it.
37) Questions and Answers : Macintosh : 50 hours? (Message 799) Posted 30 Sep 2005 by Shaktai Post: I'm running on a iMac G5 1,8 GHz, and my first WU for this project is telling me 50 hours to completion. Is that normal or is there something wrong? Edit: Something strange happend, the numbers of % completed changed while rosetta was paused and something else was running. And than the boinc client crashed, which didn't happen before I added Rosetta. Rosetta tends to jump in increments of 8.3%. The estimated time of completion is not very accurate. One of the things I found that I had to do on my iMac G5 is extend the time for switching between projects from 60 minutes to around 120-180 minutes. This will allow the Rosetta app to get to the next checkpoints, even during the final phase of calculations which can be a bit long. My actual times on my iMac G5 1.6 ghz are 3.5 to 5 hours, depending on the work unit. I have successfully completed 34 work units since 23-9-05. Another thing to be aware of is that the Rosetta process is actually 2 step. Usually the first 75-83.3% goes quickly, but the last part after 83.8% takes longer. This is because the calculations in the second step are more complex. There is actually a problem with some projects such as Einstein and Rosetta not switching properly with SETI. The problem is apparently with the libraries. The Mac SETI app is using some older libraries. Suposedly the next version of the Mac SETI app will fix this, but that may not occur until v5.x
38) Message boards : Number crunching : A few things about RAC (Message 738) Posted 29 Sep 2005 by Shaktai Post: OH.... and the #1 & #3 Top computers (based on RAC) are still bogus. And #2 is the real deal. :p Looking at the processing times, I suspect that the #3 computer is overclocked. So it may not be bogus. It is also using a highly optimized BOINC client which is evident from the extremely high benchmarks, creating a higher claimed credit. Because many crunchers work on projects where they run highly optimized apps that tend create lower then average claimed credit, they compensate by running an optimized BOINC client. However, on a project that does not have the same highly optimized apps, they will overclaim credit. The #1 computer however has been active since before I joined, but it has a very "recent" node number. In fact it has had several recent new node numbers. There is a glitch in the standings where if a new node is created, and then merged with a prior node that has accrued credit. The prior credit will appear as though it is new credit, and for a short period of time it will experience an inflated ranking. The current #1 client has had multiple new nodes that were merged. This might be intentional, but then due to the high error rate of work units it may also be accidental, because of a hardware problem. It's RAC standing is inflated though due to the frequent new nodes created, that are then merged with the older nodes. If it is intentional, then it is really a silly waste of time, because the total credit will eventually be far less then other crunchers and the "game" will be obvious. Interestingly, this caused me to look at my own clients, one of which was used for testing highly optimized SETI clients. I discovered that I still had the optimized BOINC client and was greatly over-claiming credit for that machine on Rosetta. I have now scaled it back to a more reasonable client. Of course that only went on for a couple of days, and it is actually one of my slower machines so no real harm was done. Because of on and off use due to testing, its RAC is still well below its normal potential. But this question did remind me that I had been swapping BOINC clients for testing purposes on different projects. Going forward it will now be claiming half what it did for the last day or two, which should put it in the middle of the pack. I have stopped testing and dedicated it solely to Rosetta. For a project like Rosetta, which is more similar to Climate in its credit needs (redundancy is not needed), I think a set amount for each work unit completed would be more fair. That way optimized BOINC clients could not influence the outcome. Credit would be earned based upon the work actually completed. Faster apps or computers will naturally increase total credit more, because more science work is done. Can't see where there is any perfect solution though.
39) Message boards : Rosetta@home Science : How Is Rosetta Different Than Folding@home? (Message 680) Posted 28 Sep 2005 by Shaktai Post: Although the growth of the project over the last week has been amazing, our efforts to predict protein structures are still limited by total computing power. Does anybody have suggestions on how to recruit more participants? thanks, David I think you are already on the right track. - Good information about the project. - Responsiveness from the project team. - Consistent feedback to participants - Updates on the progress and the science. - Links to resources and articles on the project. - Optimized apps for the various platforms. - When you officially get out of beta status, more will come. Looks like the number of processors has doubled in the past week. - If problems occur, regular updates on progress, even if it is just periodic messages to say "we are working on it." - Reliability and consistent work will attract many long term crunchers. The word was slow getting out at first, and the first links I saw posted were incorrect. Now the word is getting out on multiple sites and in multiple forums. I think you can expect a moderate growth surge during beta, and a substantial surge after the project is out of beta. Make sure the servers are ready for it. Getting the bugs out of the Linux and Macintosh apps will also help. (remember that while these users are a minority with their platform of choice, that many of them bring substantial numbers of windows machines along with them, including yours truly) Keep up the good work you are doing and many more folks will come. I'll speculate that the next month will see near exponential growth, just doing what you are doing now. Oh, and once the word is out, your simplified credit system will attract many "casual" crunchers who are confused by, or don't like the multi-computer validation system required for other projects. Add a simple screensaver when you have time, and many of them will become long-term addicts.
40) Questions and Answers : Macintosh : WU Errors on Mac Platform (Message 678) Posted 28 Sep 2005 by Shaktai Post: There are some problems with OS 10.3.x. I think that David Kim will be working on these when he gets back from vacation. For now, 10.4.x is required. This was answered in the other thread in this group as well. Not the most desirable answer, but know that the plan is to make the project 10.3.9 compatible. It will just take a little while longer to do so.

Previous 20 · Next 20