Longer workunits
log in

Advanced search

Message boards : News : Longer workunits

1 · 2 · 3 · Next
Author Message
Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 9 Jun 12
Posts: 541
Credit: 51,845,904
RAC: 7,415
Message 2159 - Posted: 8 Dec 2013, 10:50:25 UTC
Last modified: 18 Dec 2013, 11:49:10 UTC

There are twice longer workunits in a queue. Credit is doubled and so the deadline.

Radim Vančo (Kyong)

MarkJ
Avatar
Send message
Joined: 27 Jun 12
Posts: 105
Credit: 42,033,920
RAC: 61,408
Message 2161 - Posted: 8 Dec 2013, 11:02:56 UTC

How many WU to work out an asteroid now that we've got these larger work units?
____________
BOINC blog

T%/!ta
Send message
Joined: 29 Nov 13
Posts: 1
Credit: 681,120
RAC: 613
Message 2162 - Posted: 8 Dec 2013, 15:22:45 UTC

Je někde české forum o BOINC, vysvětlení funkcí, český team, hodnocení... Prostě jak to vše funguje?

Kmonicek Tomas
Send message
Joined: 27 Jun 12
Posts: 1
Credit: 14,073,292
RAC: 2,953
Message 2163 - Posted: 8 Dec 2013, 15:40:46 UTC - in response to Message 2162.
Last modified: 8 Dec 2013, 15:43:46 UTC

http://forum.czechnationalteam.cz/viewforum.php?f=284
http://www.czechnationalteam.cz/

[TA]Assimilator1
Avatar
Send message
Joined: 24 Aug 13
Posts: 102
Credit: 17,881,680
RAC: 1,860
Message 2164 - Posted: 8 Dec 2013, 23:21:04 UTC - in response to Message 2161.

How many WU to work out an asteroid now that we've got these larger work units?

Yea I'd be interested to know too :)
____________
Team AnandTech - SETI@H, Muon1 DPAD, Folding@H, MilkyWay@H, Asteroids@H, LHC@H, POGS, Rosetta@H, Einstein@H.

Main rig - i7 4930k @4.1 GHz, 16GB DDR3 1866, HD 7970
2nd rig - Q9550 @3.6 GHz, 8GB DDR2 1066, HD 7870 XT 3GB(DS), Win 7

Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 9 Jun 12
Posts: 541
Credit: 51,845,904
RAC: 7,415
Message 2165 - Posted: 9 Dec 2013, 10:41:00 UTC - in response to Message 2161.

Current status is about 80 WUs per an asteroid. But it depends on how many data is available.

Rise Zenrise
Send message
Joined: 23 Nov 13
Posts: 2
Credit: 427,680
RAC: 0
Message 2166 - Posted: 9 Dec 2013, 15:39:18 UTC

? new to asteriod@home, how large is a WU, I guess I mean as in Flops. ex. I have an average of 4,689. Is that a day or is that all the days since i have joined. Ok i won't erase that, I can figure out an average but i thought it was funny i typed it without thinking first. any ways, just would like to know a little info about what a WU is, how big, nice to know 80 to process one, thanks for that tid bit. Oh yeah, and what are the badges that show up under badges for? thanks in advance. Happy Holidays
Rise Zenrise @unti11ambsb3c0m31i0ns

MarkJ
Avatar
Send message
Joined: 27 Jun 12
Posts: 105
Credit: 42,033,920
RAC: 61,408
Message 2169 - Posted: 12 Dec 2013, 8:33:22 UTC
Last modified: 12 Dec 2013, 8:36:36 UTC

I think they have gotten too big for the Raspberry Pi now. They were around 315,000 seconds per wu. This change pushes them right out. I realize you increased the deadline but they will be taking about a week per wu.

Is it possible to have smaller wu for the ARM and Android machines?
____________
BOINC blog

Janos of TSBT
Send message
Joined: 14 Jan 13
Posts: 2
Credit: 19,812,000
RAC: 0
Message 2170 - Posted: 12 Dec 2013, 9:36:04 UTC - in response to Message 2169.

I think they have gotten too big for the Raspberry Pi now. They were around 315,000 seconds per wu. This change pushes them right out. I realize you increased the deadline but they will be taking about a week per wu.

Is it possible to have smaller wu for the ARM and Android machines?


Agreed. I have a few Odroid U2's and the long WU's are a bit too long.

Profile (retired account)
Send message
Joined: 3 Jan 13
Posts: 30
Credit: 1,705,200
RAC: 0
Message 2173 - Posted: 12 Dec 2013, 16:54:52 UTC - in response to Message 2159.
Last modified: 12 Dec 2013, 16:57:52 UTC

There are twice longer workunits in a que. Credit is doubled and so the deadline.


Compared to the last workunits I got beginning of October, the current ones seem to be four times as long, yielding 480 instead of 120 cr., right?
____________

Profile HA-SOFT, s.r.o.
Project developer
Project tester
Send message
Joined: 21 Dec 12
Posts: 173
Credit: 50,218,560
RAC: 17,112
Message 2174 - Posted: 12 Dec 2013, 17:13:52 UTC - in response to Message 2173.

There are twice longer workunits in a que. Credit is doubled and so the deadline.


Compared to the last workunits I got beginning of October, the current ones seem to be four times as long, yielding 480 instead of 120 cr., right?


Yes.

[TA]Assimilator1
Avatar
Send message
Joined: 24 Aug 13
Posts: 102
Credit: 17,881,680
RAC: 1,860
Message 2175 - Posted: 12 Dec 2013, 18:09:32 UTC - in response to Message 2166.
Last modified: 12 Dec 2013, 18:43:39 UTC

? new to asteriod@home, how large is a WU, I guess I mean as in Flops. ex. I have an average of 4,689. Is that a day or is that all the days since i have joined. Ok i won't erase that........................ Oh yeah, and what are the badges that show up under badges for? thanks in advance. Happy Holidays
Rise Zenrise @unti11ambsb3c0m31i0ns

You should of erased that :P

You get certain badges when you've hit certain credit milestones, the only ones I know offhand (from my own stats) are :-
C 10k
Si 100k
Fe 500k
Ni 1M

And yea they're chemical symbols, in English they are C = Carbon, Si - Silicon, Fe = Iron, Ni = Nickel
Their are more milestones I just haven't bothered googling them ;).
____________
Team AnandTech - SETI@H, Muon1 DPAD, Folding@H, MilkyWay@H, Asteroids@H, LHC@H, POGS, Rosetta@H, Einstein@H.

Main rig - i7 4930k @4.1 GHz, 16GB DDR3 1866, HD 7970
2nd rig - Q9550 @3.6 GHz, 8GB DDR2 1066, HD 7870 XT 3GB(DS), Win 7

Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 9 Jun 12
Posts: 541
Credit: 51,845,904
RAC: 7,415
Message 2177 - Posted: 12 Dec 2013, 19:58:48 UTC

Well, adding the shorter workunits for Raspberry and Odroid isn't easy. Problem is that I would have to create a new application and users would have to set what they want to compute. Users who don't check the settings too much would have both in their BOINC manager and for us maintaining more versions of data with different sizes is also more complicated and more resources consuming on server side.
So I think that the current size is I hope a good compromise among the length. A new problem will be with GPU app where the times should be shorter. I hope the current length is well balanced/compromised for every version of app because we don't have to change the data structure for GPU so creating new applications (splitting) is not helping.

Profile (retired account)
Send message
Joined: 3 Jan 13
Posts: 30
Credit: 1,705,200
RAC: 0
Message 2178 - Posted: 12 Dec 2013, 20:52:20 UTC - in response to Message 2177.
Last modified: 12 Dec 2013, 21:12:13 UTC

Problem is that I would have to create a new application and users would have to set what they want to compute. Users who don't check the settings too much would have both in their BOINC manager (...)


For this new application you could do it the same way as Einstein did recently with the BRP4 non-GPU work (links, see last section). They simply don't ship this work to x86 CPUs anymore, only intel GPUs and ARM devices. Maybe you could also include the vanilla non-SSE 32bit x86 Asteroids app here?

By the way, does it make sense to provide a 32bit SSE3 and AVX app? I mean, I don't know, are there many people running a 32bit OS on such a modern CPU still?


A new problem will be with GPU app where the times should be shorter.


So I assume above problem has to be adressed at least when the GPU apps are available cause you can't satisfy both a GTX Titan and a Raspberry anymore?

Regards

Profile Kyong
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 9 Jun 12
Posts: 541
Credit: 51,845,904
RAC: 7,415
Message 2179 - Posted: 13 Dec 2013, 9:02:37 UTC

This is what I wanted to avoid. We have to prepare two version of workunits and additional server components so it also increase the load. I think it would be better to do this when there is a new server which will be running only a database. It is planning after upgrading the current server.

Profile BilBg
Avatar
Send message
Joined: 19 Jun 12
Posts: 219
Credit: 614,040
RAC: 0
Message 2190 - Posted: 14 Dec 2013, 8:05:13 UTC - in response to Message 2178.

By the way, does it make sense to provide a 32bit SSE3 and AVX app? I mean, I don't know, are there many people running a 32bit OS on such a modern CPU still?

For 32bit SSE3 - Yes, of course makes sense.
Many run 32bit Windows XP for BOINC:
http://boincstats.com/en/stats/-1/host/breakdown/os/0/4/0

This link probably also includes non-active (retired) hosts but many like/prefer 32bit Windows XP even on 64bit CPU (including me)


I may ask the opposite - is there need for 64bit applications if 32bit run on 64bit OS?
Why do you think that 64bit applications are faster?
In fact sometimes 64bit is slower than 32bit (on 64bit CPU/OS) because of wider address space (8-byte pointers)

SSE3 is no different in bitness in 32bit vs 64bit OS
'64bit' is about "64-bit integer arithmetic and logical operations, and 64-bit virtual addresses" - http://en.wikipedia.org/wiki/X86-64#Architectural_features
SSE2/SSE3 "instructions implement the vector floating-point operations" and use 128bit registers both on 32bit and 64bit CPU/mode - http://en.wikipedia.org/wiki/SSE2

And there are 32bit CPUs with SSE3 (e.g. Pentium 4 - http://en.wikipedia.org/wiki/SSE3)


____________



- ALF - "Find out what you don't do well ..... then don't do it!" :)

Rooooobert
Send message
Joined: 4 Dec 13
Posts: 3
Credit: 793,200
RAC: 1,674
Message 2191 - Posted: 16 Dec 2013, 19:44:04 UTC

My raspberry is now at 40% and 120 hours. The Work Unit will be finished in appoximate 180 hours, but it should be finished in 110 hour.
Such a misery - we will be too late.

Rooooobert

Profile HA-SOFT, s.r.o.
Project developer
Project tester
Send message
Joined: 21 Dec 12
Posts: 173
Credit: 50,218,560
RAC: 17,112
Message 2192 - Posted: 16 Dec 2013, 21:55:07 UTC - in response to Message 2191.

My raspberry is now at 40% and 120 hours. The Work Unit will be finished in appoximate 180 hours, but it should be finished in 110 hour.
Such a misery - we will be too late.

Rooooobert


My raspi did longer task in 447,546.86 s. It's overclocked to 1GHz.

bryan
Send message
Joined: 17 Dec 13
Posts: 1
Credit: 274,560
RAC: 0
Message 2193 - Posted: 17 Dec 2013, 14:19:15 UTC - in response to Message 2159.

I am new what is que and how do I get there please

Dagorath
Send message
Joined: 16 Aug 12
Posts: 293
Credit: 1,116,280
RAC: 0
Message 2194 - Posted: 17 Dec 2013, 20:38:15 UTC - in response to Message 2193.
Last modified: 17 Dec 2013, 20:38:45 UTC

A que is a stick used in billiards to strike the que ball. Kyong meant the queue. A queue is a line or series of things or people. The people waiting to get into a store to be one of the first to get a new Xbox (for example) form a queue. snatch up movie form up in a queue outside the door. In BOINC the queue usually refers to a series of tasks or the result reports obtained from tasks. Kyong was referring to the queue of tasks waiting on the server to be sent to hosts such as yours. Your host has already received a number of tasks from that queue and is crunching them. As your host crunches the tasks it will upload the result from each task immediately upon completing the task. It will queue the result reports (distinct from the results they refer to) and when there are a few in the queue your host will contact the server and upload those reports.

You can see how many tasks are in the server queue by clicking the Project Status link on the project's home page. On that page you will see "Tasks ready to send" which is the number of tasks remaining in the server's queue. You will also see "Tasks in progress" which is the number currently crunched spread out across the pool of volunteer hosts.

On your host you can't see your queue in BOINC manager if it is in Simple View mode. To switch to Advanced View where you can see your queue click View -> Advanced View.

You can also see your queue from the server's perspective if in your web browser click Your Account -> Tasks. There are many useful links to explore on that page. It's good idea to check your tasks and results by clicking Your Account -> Tasks frequently. If your host is computing good results they will be shown on that page as Success or Verified. If your host is computing bad results (due to faulty RAM or excessive overclocking, for example) they will show as Compute Error or some similar term.
____________
BOINC FAQ Service
Official BOINC wiki
Installing BOINC on Linux

1 · 2 · 3 · Next
Post to thread

Message boards : News : Longer workunits


Main page · Your account · Message boards


Copyright © 2017 Asteroids@home