Welcome to MilkyWay@home

Posts by bobgoblin

1) Message boards : Cafe MilkyWay : Word Link V (Message 38969)
Posted 21 Apr 2010 by bobgoblin
Post:

Still
;-p




Picture
2) Message boards : Cafe MilkyWay : Word Link V (Message 38585)
Posted 11 Apr 2010 by bobgoblin
Post:
Winter Sports



Olympic
3) Message boards : Cafe MilkyWay : Word Link V (Message 37613)
Posted 21 Mar 2010 by bobgoblin
Post:
jumpin'
4) Message boards : Cafe MilkyWay : Word Link V (Message 37578)
Posted 20 Mar 2010 by bobgoblin
Post:
cops
5) Message boards : Cafe MilkyWay : Word Link V (Message 37522)
Posted 18 Mar 2010 by bobgoblin
Post:
coping
6) Message boards : Number crunching : Server Crash November 10 (Message 33297)
Posted 15 Nov 2009 by bobgoblin
Post:
If you want to see totally poor management, go look around at Cosmology. There are issues which have not been fixed in over a year. Admins continually bungle SQL scripts. The server crash in February/March of this year still has not been completely "fixed", meaning there are newer issues than those that are more than a year old, so they're not even back to where they were before the crash.


I crunch for several projects and there are always ups and downs with all of them. Some just vanish with no word as to why. But Cosmo... that's the perfect example of how NOT to run a project.


7) Message boards : Number crunching : I see that there is no data to work (Message 32937)
Posted 1 Nov 2009 by bobgoblin
Post:
This project never promised to be 24/7. I expect that in the next 24 hours it will be sorted.



true that. it's also halloween weekend and a time change weekend. i wouldn't bother with work today either.
8) Message boards : Application Code Discussion : GPU app teaser (Message 32731)
Posted 25 Oct 2009 by bobgoblin
Post:
There are a lot of computation errors going around nowadays, about half of each batch of 48WUs on my 'general use' PC ends up that way. Almost none on my crunchers. Doesn't seem to be hurting the rac though, probably because they fail as soon as they start and don't waste much time crunching before failing.
3 of my 6 4870s I access via TightVNC, without problems.



I'm seeing several wu's with computation errors too but not in that great number. I noticed some running version .20, and my GPU would lock up every now and then, hd 4850 1g with catalyst 9.9. so I installed version .20b. but now the lock ups are happening more frequently, so I upgraded catalyst to version 9.10 last night. It only ran a few hours and locked up again.

It seems that if they fail right away, then the gpu doesn't lock up. But if they fail after they've been crunching for a few seconds or near the end, it locks up the gpu



I figured out my own problem. The issue of MW locking up on the gpu had been going on all along. I had been running multiples, 3 @ the same time. I would notice much later that only 2 or 1 were running, then close Boinc and restart. But I had switched over to running only 1 @ a time, and several weeks later I began having the intermittent problem of MW not processing at all as the 1 unit has locked up and cannot release the next to process.

I finally realized the two issues were the same issue and started running 2 units at the same time again.
9) Message boards : Application Code Discussion : GPU app teaser (Message 32703)
Posted 24 Oct 2009 by bobgoblin
Post:
There are a lot of computation errors going around nowadays, about half of each batch of 48WUs on my 'general use' PC ends up that way. Almost none on my crunchers. Doesn't seem to be hurting the rac though, probably because they fail as soon as they start and don't waste much time crunching before failing.
3 of my 6 4870s I access via TightVNC, without problems.



I'm seeing several wu's with computation errors too but not in that great number. I noticed some running version .20, and my GPU would lock up every now and then, hd 4850 1g with catalyst 9.9. so I installed version .20b. but now the lock ups are happening more frequently, so I upgraded catalyst to version 9.10 last night. It only ran a few hours and locked up again.

It seems that if they fail right away, then the gpu doesn't lock up. But if they fail after they've been crunching for a few seconds or near the end, it locks up the gpu
10) Message boards : Application Code Discussion : GPU app teaser (Message 32701)
Posted 24 Oct 2009 by bobgoblin
Post:
duplicate entry
11) Message boards : Number crunching : Results refused (Message 32571)
Posted 20 Oct 2009 by bobgoblin
Post:
Someone mentioned in another thread that it may be due to a communication problem during upload. There have been a lot of these in recent days.
Don't worry about it. If its already been reported as a success then they have the result and you have the credits.



right, boinc doesn't receive the reply back that the wu has been successfully reported.

oddly, i have a bunch of wu's stuck in Pending. that's unusual for this project.
12) Message boards : Number crunching : BOINC 6.10.14 (Message 32528)
Posted 18 Oct 2009 by bobgoblin
Post:
MGR: If aborting multiple tasks, ask "Are you sure?" only once.


I like this one as well, I have on occasion had to abort quite a few units and it was nuts clicking yes for all of them.


All you have to do is Highlight all the Wu's you want to Abort & then when the Pop Up Box appears asking you Yes/No just Press & hold the Enter Key until all the selected Wu's are Aborted. Suspend all work first & it only takes a few seconds that way.

A option to Abort All would be nice though but that's the work around I've used on many occasion to Abort multiple Wu's ...



but holding down the enter key like that, i've had issues with the ones already aborted report then boinc accidentally deletes wu's i didn't want to abort.

I guess a Reset Project would be Abort All
13) Message boards : Number crunching : GPU Projects (Message 31249)
Posted 20 Sep 2009 by bobgoblin
Post:
I'm running them in 14 minutes
14) Message boards : Number crunching : GPU Projects (Message 31245)
Posted 20 Sep 2009 by bobgoblin
Post:
ABC recently cut their credit by 75% when they released their new app. They doubled the credit a week or so ago, but it's still not back up to where it was. last time I looked they were trying to figure out how to get it the rest of the way up.


Collatz is good, and I used to run 3x+1. But their ATI units freeze up after an hour or so on my i7 w/ 4850. I think my card maybe overheating. plus it's only a 512m, which could be too small for their size wus. But possibly it's because I'm still running 6.6.36?
15) Message boards : Number crunching : Credits (Message 24520)
Posted 7 Jun 2009 by bobgoblin
Post:
I'm seeing the occasional 0.00 credit for ps_sgr_208... work units. It is often enough that I can see a change in my RAC but my change is not as significant as you report.

I'm running CPU only.



i've had a handful do the same:

Task ID 74982228
Name ps_sgr_208_2s_2_1284325_1244411529_0
Workunit 73863811
Created 7 Jun 2009 21:52:14 UTC
Sent 7 Jun 2009 21:54:19 UTC
Received 7 Jun 2009 21:57:07 UTC
Server state Over
Outcome Success
Client state Done
Exit status 0 (0x0)
Computer ID 41037
Report deadline 10 Jun 2009 21:54:19 UTC
CPU time 29.11299
stderr out

<core_client_version>6.6.20</core_client_version>
<![CDATA[
<stderr_txt>
Running Milkyway@home ATI GPU application version 0.19f by Gipsel
CPU: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz (8 cores/threads) 2.65999 GHz (925ms)

CAL Runtime: 1.3.158
Found 1 CAL device

Device 0: ATI Radeon HD 4800 (RV770) 512 MB local RAM (remote 2047 MB cached + 2047 MB uncached)
GPU core clock: 690 MHz, memory clock: 250 MHz
800 shader units organized in 10 SIMDs with 16 VLIW units (5-issue), wavefront size 64 threads
supporting double precision

2 WUs already running on GPU 0
Starting WU on GPU 0

main integral, 160 iterations
predicted runtime per iteration is 158 ms (33.3333 ms are allowed), dividing each iteration in 5 parts
borders of the domains at 0 320 640 960 1280 1600
Calculated about 3.70012e+012 floatingpoint ops on GPU, 6.34181e+007 on FPU. Approximate GPU time 29.113 seconds.

probability calculation (stars)
Calculated about 7.64289e+008 floatingpoint ops on FPU.

WU completed.
CPU time: 4.38363 seconds, GPU time: 29.113 seconds, wall clock time: 90.632 seconds, CPU frequency: 2.66 GHz

</stderr_txt>
]]>

Validate state Invalid
Claimed credit 0.177618976661038
Granted credit 0
application version 0.19
16) Message boards : Number crunching : Confused..about what GPU support is coming 1st (Message 24490)
Posted 7 Jun 2009 by bobgoblin
Post:

But why then is a ATI app out already, when Cuda is the one that is supported by BOINC?



The ATI app only processes CPU WUs on an ATI GPU. Once the Cuda code is released and MW GPU, which is a separate project, begins sending work an ATI app will be made available to process Cuda WUs on an ATI GPU.

Afterward, the ATI app that processes CPU WUs will no longer be allowed.
17) Message boards : Number crunching : WUs distribution (Message 21725)
Posted 9 May 2009 by bobgoblin
Post:
bob>where can i get one of these scripts?
Mine [XP] says:

@echo off
REM Boinc Automatic Project Update
:TOP
"c:\Program Files\BOINC\boinccmd.exe" --project http://milkyway.cs.rpi.edu/milkyway/ update
ECHO %TIME%
SLEEP 5m
GOTO TOP

You'll have to get SLEEP from the net, or fake it using PING as mentioned elsewhere on this board



Oh, you mean the thread that was locked for administrative purposes??? Yeah, I finally figured it out using windows task manager. and a few people have PM'd me since. Now my CPU machines have a small but constant flow of work, and my GPU gets units in small batches but allows me to continue running 4 ABC and 4 CPDN wus concurrently with 3 Milkyways, when I get them.

my average has jumped up from about 1000 to 25000 a day. still a far cry from the 80k/day I was getting just 1 1/2 months ago.

thanks, v
18) Message boards : Number crunching : WUs distribution (Message 21723)
Posted 9 May 2009 by bobgoblin
Post:
There is another way to reduce the hammering of the server -- eliminate support for ATI GPU <smile>. That is where the server stress started. I'm not advocating that (even though I have no ATI 38xx or 48xx cards deployed -- not a gamer or photo processor or CAD/CAM guy), but as long as folks are seeking to solve a problem, add that as grist for a potentially flaming discussion.



Of course, the easy way to stop the hammering of the server would be to reduce the credits to that of *cough*seti*cough*, or even below that. Then watch how quickly people will stop crunching and hammering here - after some very unpleasant postings in the forum, no doubt.

Excuse me while I put on my fire-resistant suit!




There was some server stress when gpu was opened up, but more so when they increased the min number of wu's on the server that would trigger new wu generation. then the server started sending out "Low" responses all the time and everyone's boinc clients started scheduling way off in the future.


what may have been a better solution would be to reset the trigger amount back where it used to be but generate more work at that level. Before travis monkeyed around with those trigger levels, there was something on the order of 80k wu's running at any time. now there are only 30-ish k.
19) Message boards : Number crunching : WUs distribution (Message 21374)
Posted 5 May 2009 by bobgoblin
Post:
For a start they are using GPUs in their machines, which get through the wu's very quickly so their machines then ask for more work more often.

Secondly they are either using a flag in older BOINC clients (-return_results_immediately) or they are using some form of script that overrides the 'normal' BOINC exponential back off system and contacts the server to get work more often. The script utilises a standard feature that comes with BOINC.

Thirdly, they are potentially just lucky....



where can i get one of these scripts? all my machines are dry as a bone and have been for a month.
20) Message boards : Number crunching : MilkyWay_GPU (Message 20253)
Posted 23 Apr 2009 by bobgoblin
Post:

But it is not always a simple answer as to what is the most efficient way to get from here to there.


My point was, and is, that there was a lot of negativitiy from people who have GPUs. Those of us not fortunate enough to have such hardware are being deemed as "second class citizens". There is a ton of greed that is going on from the GPU side. If the project itself is happy with the rate of progress, then I suggest to the "GPU guys" that they simply deal... It's either that or they lobby the project PRIVATELY to eliminate CPU participation.



greed for work units isn't restricted to gpu machines. and there seems to be plenty of complaints from both sides.

my GPU machine is just as dry as my CPU machines. in fact, i've switched it (an i7) over the cpdn and abc pretty much exclusively. and my core 2 duo is doing artificial intellegence when it cannot get any optimized mw. my old 3GHz p4 is relying heavily on poem and enigma when there are no mw units. And my even older 1.8 celeron & 1GHz p3 I just let run dry rather than trying to micromanage them.

once the gpu-genie was released, i don't think anyone expected the onslaught of new (gpu) users that knocked mw on it's cyber-butt.


and I've deleted the rest of my post as I've just realized I was degenerating into a rant of sorts.


Next 20

©2024 Astroinformatics Group