Message boards :
Number crunching :
Slowest Number Cruncher?
Message board moderation
Previous · 1 · 2 · 3
Author | Message |
---|---|
Crunch3r Send message Joined: 15 Apr 99 Posts: 1546 Credit: 3,438,823 RAC: 0 |
> thats why there are reporting deadlines. sheesh, you'll get your credit... > Yes your right that´s what deadlines are for but remember: ->Sparc2 had just finished it's one and only WU sent to it. It took: 568,564.31 CPU seconds, or 6.5 days!!! Guess what... after beeing reported without finnishing in deadline the wu is beeing send out to another host to crunch. With a new deadline that makes 5 host to crunch the same wu ONLY because of those host getting WU´s and not beeing able to crunch within deadline. Join BOINC United now! |
Paul D. Buck Send message Joined: 19 Jul 00 Posts: 3898 Credit: 1,158,042 RAC: 0 |
> Well i dont know what you all think about this but i think there should be > some restriction for clients wich take over 24 hours to crunch a WU. This is why the schedulers shoiuld be using the average turn around time to try to issue to computers with similar turn-around times. I don't believe that all projects are using this feature, or that it is even working as well as we would like. But, we are moving in that direction. Keep in mind that this is a brand new way of doing business and there are a lot of things going on. Right now we are seeing a lot of effort in making the system work well with multiple projects. Besides, the idea is to get the work done. Not to get the work done fast. |
Saenger Send message Joined: 3 Apr 99 Posts: 2452 Credit: 33,281 RAC: 0 |
> Well i dont know what you all think about this but i think there should be > some restriction for clients wich take over 24 hours to crunch a WU. > > I think it would be best if boinc, after finishing benchmarks and predicting > an average WU turn around time of max. 24 hours per CPU should stop on those > machines wich take longer than that. > > It realy sucks if 3 P4 or Athlons where crunching the same WU and validation > depends on a SGI / SPARC CPU wich takes 100 hours or even more to process a > WU. [rant mode] It really sucks if some geeks with more money than brain demand everyone to have the latest gimmick running just to boost their bragging competition for some wee days. [/rant mode] Deadline is 2 weeks, 6.5 days is far shorter than 2 weeks, so no problem whatsoever! Everybody who can crunch within the time restrictions be invited to do so. If the project team could do something to soothe the impatient crunchers, do so, but on a low priority level compared to real problems. Happy crunching everybody! Gruesse vom Saenger For questions about Boinc look in the BOINC-Wiki |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
The whole idea behind Seti is to gather together a whole bunch of peon PC computers together, to outcrunch the fastest CRAY computers out there in the world. The reason why I tried the Sparc 5, and Ultras, are cuz. If you know some university institutions, these can be had by the hundreds for free. I'm still looking for that magic price/performance ratio. And I think I found it, my next slowest cruncher -> XBOX! Too bad, these can't be had for free though. Well, if it doesn't work out, I'll have a nice game machine to boot :-). Jimmy |
Astlor.ca.SETI@home Send message Joined: 29 Jun 99 Posts: 12 Credit: 2,180,506 RAC: 0 |
Well, I am lucky: I have electricity included in my rent at the appartment I am in. So the price/performance comes strictly on the issue of hardware. Or does it? Most of the machines I have around here are either too old to be considered usefull by me, or supplied by work for one purpose or another. To be frank, the only machines I don't run boinc on anymore are a) Laptops, (they spend too much time off-line,) and b) my media centre PC, (Little Duron 1.3 that seems to freakout around 70% loading over an hour. Trust a very early ECS chip/mobo combo to have bad thermals.) Conversely, I can have a reasoned discussion on the following: as they are the machines I have chosen to keep running BOINC. Athlon 2000+ /512MB Athlon 2600+ /1024MB 2x Athlon 2500+ /512MB 2x P4 3.06Gh8z machines /1024MB-2048MB 3x Dual P-IIIs (approx 1Ghz/chip) /512-1024Mb P4 2.8Ghz /512MB Dual Opteron 244 /6144mb Well, for machines undergoing regular daily use, for 8hrs a day, (by regular daily use I mean office apps, and the interweb,) the suprising performance point is the Athlons. Athlon 2000-2600s are cheap like borsht, and seem to out-perform all intel boxes. I can pick up these boxes at my local PC store for 400$ Canadian, specced like this: Sempron 2800+ (roughly seems to be equivalent to a Athlon [Barton] 2200,) 80GB HDD, 256MB RAM, 52x CDRW, Keyboard/Mouse/Spkrs, and everything else integrated on board. We call 'em 400$ wonder turkeys, becuase they are damned fine office machines, and have yet to give up under stress. After careful analysis of all the above listed machines, I must state my disapointment with the P4s. The Dual P-IIIs are NICE price/performance boxes, spitting out a more than satisfactory amount of WUs, despite moderate loading during the day. The opteron has only been running a couple of days, but I am frankly, quite disapointed in it, when stacked up price/performace against my 400$ wonder-turkeys. I am consider building a farm out of them. I can get them for about 200$ a pop if I do away with harddrive, case, and downgrade the CDRW to a basic DVD-ROM or 52X CD-ROM. Run like noppix on 'em and start a SETI stack. What do you people think? For my first SETI stack, are stripped 400$ wonder-turkeys the way to go? |
jimmyhua Send message Joined: 16 Apr 05 Posts: 97 Credit: 369,588 RAC: 0 |
> This is why the schedulers shoiuld be using the average turn around time to > try to issue to computers with similar turn-around times. I don't believe that > all projects are using this feature, or that it is even working as well as we > would like. But, we are moving in that direction. Seems to work fine on Seti. My Pentium-M box seems to have the fastest turn around times and claims the least credit. This is not surprising. The surprising thing is, it also receives the least credit. I imagine this is cuz my Pentium-M is matched against other Pentium-M machines for verification. And we all claim very little credit, so the credit received is also less. Average Credits received per WU: Pentium-M: 20 Sempron 2500+: 30 Duron 800: 40 So I believe the faster machines (which claim less credit) are being matched up against my Pentium-M quite nicely. This probably screws up the whole validity of the credit rating system, but not by too much, so I won't worry about it. |
[DP]Ghent96 Send message Joined: 3 Apr 99 Posts: 18 Credit: 7,228 RAC: 0 |
<blockquote>> If I can find eight 16MB 30-pin 100ns SIMMs, I can out-slow you. > I might just have four of those from my (retired) 166MHz, which averaged about 96 hours per WU (± 8 hours depending on its disposition). By comparison its replacement, a 1.8GHz Duron, averages about four hours and small change (± 15 minutes).</blockquote> ...i thought the joke was needing 8 of them? %) ...since usually there's only 4 ram slots in a mobo. Btw, I was thinking of getting a modem in my old IBM PS2 model 25, 286, WITH... fully upgraded 640kb of ram, and 20mb hard drive. Then I simply must set it to crunching :) |
spacemeat Send message Joined: 4 Oct 99 Posts: 239 Credit: 8,425,288 RAC: 0 |
<blockquote>> If I can find eight 16MB 30-pin 100ns SIMMs, I can out-slow you. > ...i thought the joke was needing 8 of them? %) ...since usually there's only 4 ram slots in a mobo. </blockquote> Mac IIcx, discontinued 1991, 8x 30-pin SIMM slots |
[DP]Ghent96 Send message Joined: 3 Apr 99 Posts: 18 Credit: 7,228 RAC: 0 |
dang... now that's old... *owned* But seriously, I think I can get my old P100 or P120 setup and crunching, assuming I can find the 64mb of ram in one or both of them, and that should be pretty darn slow. I really can't guess how they might compare to these sparq or cyrix or etc until I can get a benchmark out of one. |
Alexander Lazar Send message Joined: 11 Apr 03 Posts: 16 Credit: 448,266 RAC: 0 |
<blockquote>What do you people think? For my first SETI stack, are stripped 400$ wonder-turkeys the way to go? </blockquote> Heeeey... Who ya callin' a turkey??? LOL One of my crunchers has been an Athlon XP 1800+ for about a year, at 4:20 or so per WU (now down to 3:30 with Maverick's new optimized 4.11 client). No better bang for the [extremely cheap] buck. Although I'm shooting for a 2400 if I can get one on eBay for $60. BTW, these overclock well, if you're into that sort of thing. |
MikeSW17 Send message Joined: 3 Apr 99 Posts: 1603 Credit: 2,700,523 RAC: 0 |
<blockquote> But seriously, I think I can get my old P100 or P120 setup and crunching, assuming I can find the 64mb of ram in one or both of them, and that should be pretty darn slow. I really can't guess how they might compare to these sparq or cyrix or etc until I can get a benchmark out of one.</blockquote> I'd suggest doing your bit to save the planet - don't switch it on :) You really will be looking at many, many hours per WU. It's unlikely to be worth the electricity, even for the credits. The only time I reckon a system of this spec comes into play is if it is on anyway for another purpose. For example I know a few who use old P100/133 systems with a basic linux Os as a firewall. While it's being a firewall it can also crunch. |
spacemeat Send message Joined: 4 Oct 99 Posts: 239 Credit: 8,425,288 RAC: 0 |
if you use 64MB memory and onboard video (even a 2MB controller), it won't work. BOINC will not see the shared RAM and will refuse to download WU's due to |
©2024 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.