21)
Message boards :
Number crunching :
Problem with new W/Us
(Message 3573)
Posted 29 May 2008 by Webmaster Yoda Post: Sure is an ongoing plague. Did not have time to check my computers for an hour today. Sure enough, when I got back to them all 26 cores were hanging on those nasty things. Will have to suspend MW overnight so I don't waste another 150-200 hours of CPU time on useless work units. |
22)
Message boards :
Number crunching :
Problem with new W/Us
(Message 3540)
Posted 28 May 2008 by Webmaster Yoda Post: Well, here's the reason why it's not working (not an app issue) And I guess that means "Not a Number" and is an error. I've been aborting work units all day but can't be bothered baby-sitting my computers any further so have set MW to "nnw" (which is "No New Work") for now. |
23)
Message boards :
Number crunching :
Problem with new W/Us
(Message 3534)
Posted 28 May 2008 by Webmaster Yoda Post: Don't know if I have had any error out, but all my boxes that were crunching Milkyway overnight had results stuck at 0% after 2 to 5 hours of crunching. All of them had "gs_373082_" as the beginning of the result name. Others (gs_59#_) are working OK All are running 64 bit Ubuntu (Hardy) with 64bit BOINC client. They all worked fine yesterday and I have changed nothing at my end (no software updates installed overnight either). |
24)
Questions and Answers :
Wish list :
20 WU per core - not per host
(Message 3510)
Posted 24 May 2008 by Webmaster Yoda Post: Any further news on progress in this regard? Please? |
25)
Message boards :
Number crunching :
Best completion times
(Message 3505)
Posted 21 May 2008 by Webmaster Yoda Post: As for dual booting, that's an option I guess, but I think I'd be switching around quite a bit, considering I do a lot of DirectX programming as well. (not to mention some GLSL stuff that's very easy to test in windows) Run Windows in virtualbox under Linux :) |
26)
Message boards :
Number crunching :
Best completion times
(Message 3501)
Posted 20 May 2008 by Webmaster Yoda Post: Either there is a huge difference in the motherboards, or the Phenom @ 2.5 is kicking the c#*p out of the Q6600 @ 2.9. Methinks AMD has a few secrets beyond ghz. I say at the same speed, the Phenom is faster than a 65nm Intel Core 2 Quad. The same may not apply with the 45nm Intels which, at least on Milkyway, are much quicker again. Haven't quite decided what my next quad core will be yet, but the Phenom 9850 Black Edition is cheaper than an Intel Q9300 in my town. On the other hand, the Phenom will use more electricity (if we're staying at stock speeds) |
27)
Message boards :
Number crunching :
20 workunit limit
(Message 3488)
Posted 18 May 2008 by Webmaster Yoda Post: Travis, any further word on when we might go to 20 work units per core or at least a reduced timeout before we can get more work? Our quad core and better machines could do a lot more work for this project if at least the forced timeout between connections is reduced to something like 5 or 10 minutes. |
28)
Message boards :
Number crunching :
Best completion times
(Message 3485)
Posted 17 May 2008 by Webmaster Yoda Post: Wish I could push my X2 3800+ further but its mobo won't go beyond 230MHz FSB (no PCI lock) Swapped them around - the X2 is now on the NForce4 mobo. But I dare not push it beyond 2.4GHz - needs too much voltage to go higher and gets too hot (even with two 8cm fans blowing cool air in at full bore above the heatsink. Still, the swap has added a few hundred MHz so I can't complain. FWIW, the other mobo has a ULi chipset (not VIA) |
29)
Message boards :
Number crunching :
Best completion times
(Message 3472)
Posted 11 May 2008 by Webmaster Yoda Post: One of my budget backup boxes got a tune up today, running a new install of Hardy 64. It's an AMD 939 X2 3600 (2 X 256 L2) pushed to 2.7ghz, with 2 sticks of Geil pc 4000. That's very good and consistent with my single core S939 Athlon 64 3700+ at the same speed (also with PC4000 RAM, running command-line Hardy 64) Wish I could push my X2 3800+ further but its mobo won't go beyond 230MHz FSB (no PCI lock) |
30)
Message boards :
Number crunching :
How to get FreeBSD work?
(Message 3463)
Posted 7 May 2008 by Webmaster Yoda Post: Hmm, I'm a bit overwhelmed by all the response I guess there's not many people here who run FreeBSD (I don't either). Have you looked at this page? There's quite a few OpenBSD, NetBSD and FreeBSD BOINC clients listed there. |
31)
Message boards :
Number crunching :
Out of WU's?
(Message 3448)
Posted 4 May 2008 by Webmaster Yoda Post: My machines aren't getting any new WU's, which is confirmed by the server status page, where it says 0 WU's ready to send. All services seem to be running, so is there a problem somewhere? I believe so - at a guess, the assimilator needs a kickstart, given this line on the page you referred to: Workunits waiting for assimilation 230,895 |
32)
Message boards :
Number crunching :
20 workunit limit
(Message 3439)
Posted 1 May 2008 by Webmaster Yoda Post: by faking the number of cores/CPUs with a certain program I won't mention (and don't use) As I understand it, if you had a machine reporting in as 16 cores (of which many are faked) and asking for 0.1 day worth of work, you'd get up to 0.1 days work for each cors, including the fake ones. All I am saying is that if we go to a "per core" arrangement, some people might abuse the arrangement using such tools. I'd really like to get a bigger cache too but it's not what the project admins want (or there would not be a 20 WU limit). See the very first post in this thread... |
33)
Message boards :
Number crunching :
20 workunit limit
(Message 3414)
Posted 30 Apr 2008 by Webmaster Yoda Post: The problem with connecting that often is load on the server. There are something like 9,000 active hosts at the moment (according to BOINCStats). If every one of them connects to the server every 5 minutes, that's a lot of connections to the database, file system etc. BTW, 0.02 days is approx.half an hour, which is more than the 20 minute back-off. To only get 4 work units, a fast quad running 64 bit Linux would need to have their "connect every" set to something like 2 - 3 minutes (0.0014 - 0.002 days). That's a lot of connections for a single host. Now multiply that by the (unknown to me) number of quads (and better) attached to the project... So I don't think the key is to connect more often but to wait for "x" WU per core (or longer work units), which AFAIK Travis is working on. I'd say 10 per core and a backoff of 15 minutes would be the way to go, but only if it suits the project. Of course, this would leave the project open to clients asking for more work than they can handle, by faking the number of cores/CPUs with a certain program I won't mention (and don't use) |
34)
Message boards :
Number crunching :
Best completion times
(Message 3327)
Posted 24 Apr 2008 by Webmaster Yoda Post: Be thankful my very old laptop doesn't have enough memory to run a Milkyway WU - its benchmark is about one third the speed of the above host! Hmm, wonder if my 266MHz (Covington) Celeron with 64MB RAM would be able to complete a work unit that quickly :) All it needs is a network card! |
35)
Message boards :
Number crunching :
Best completion times
(Message 3271)
Posted 21 Apr 2008 by Webmaster Yoda Post: Q9550 overclocked 20% to 3.4GHz now. Completion times 155-158s. See http://milkyway.cs.rpi.edu/milkyway/results.php?hostid=14685 That's really amazing, given a Q6600 at higher speed takes longer (190s at 3.8GHz). Must be the extra L2 cache that makes all the difference? Or does it have very fast RAM (like DDR3/1800)? |
36)
Message boards :
Number crunching :
Milestones
(Message 3252)
Posted 19 Apr 2008 by Webmaster Yoda Post: The third in my team to make it to 1 Million, including over 250k on one computer. Woohoo! |
37)
Message boards :
Number crunching :
Best completion times
(Message 3162)
Posted 14 Apr 2008 by Webmaster Yoda Post: Q6600 at 3.6GHz (64 bit Ubuntu Gutsy) with 64 bit BOINC 5.10.8: Approx. 200 seconds per WU Have had it running faster - doing them in approx. 196 seconds at 3.7Ghz and 189 seconds at 3.8GHz |
38)
Message boards :
Number crunching :
Quads waiting on the Server
(Message 3117)
Posted 12 Apr 2008 by Webmaster Yoda Post: This question keeps rearing its head. But if you can't set the server up to issue 10, 20 or whatever number of work units per CORE, why not reduce (or eliminate) the "defering communication for 20 minutes" (as I mentioned in a post of 20 January) Set it to something like 5 minutes perhaps, rather than the current 20 minutes. It's a setting in the server's BOINC configuration. See also this post (quoted below) I'm no expert on the server-side options of BOINC, but a search of the BOINC site shows the following seemingly relevant config options (at http://boinc.berkeley.edu/trac/wiki/ProjectOptions): |
39)
Message boards :
Cafe MilkyWay :
Race on MilkyWay
(Message 3091)
Posted 11 Apr 2008 by Webmaster Yoda Post: numbers game ok, but it's rather a matter of CPU's than members. It's difficult to get the real figures for number of computers, since many people hide them but my guess is that it is a similar ratio. More of our members have now joined the project, many of them with only one or two computers too. We will not be able to rally enough resources to challenge the numbers your team is putting out but we'll see what we can do, especially after our Aussie Assault on ABC finishes (in about 3 days time) ps : I never received a response to sign in the guest part of your forum I have not seen any posts in the guest forum for a while (last post listed was on 1 March) so don't know what happened. If you tried to apply for forum membership, it was probably denied due to you not being a member of our team (per the registration agreement) |
40)
Message boards :
Cafe MilkyWay :
Race on MilkyWay
(Message 3040)
Posted 6 Apr 2008 by Webmaster Yoda Post: We're busy on another project so you probably will overtake us. Besides it's a numbers game and you've got 500 members vs our 76 on this project. No contest |
©2024 Astroinformatics Group