Yesterday (27 Mar) we had 3 Ars users in the top 10 for Moo! 3 in the top 20 in dnetc stats too. Still 211 years out from completion though.
[TBR] Viva La Roast! (Perpetual Thread V)
- Thread starter TheJet
- Start date
[url=http://arstechnica.com/civis/viewtopic.php?p=28753693#p28753693:ne0m3la6 said:
Really sad to see you leave the team, but good luck with your new team
Happy Crunchin.
I guess the Czech National Team has ramped up a bit. They don't want to let us get too far ahead of them.
I believe they are actively pushing for more production.
Any one up for a demonstration of that futility?
I believe they are actively pushing for more production.
Any one up for a demonstration of that futility?
Well I just re added Moo to try to add for a little while. running on a R9 270x at about 26-27 min/wu.
though I have noticed that for the work units that I have tried to process the dnetc_r72_1433334486_12_768_0 complete without issue but the dnetc_r72_1433334486_14_768_0 will start and just hang there and don't seem to be using any GPU. I am currently cruncing a wu with a 13 in the name and so far it looks good at 20% into the wu. the 14's would stop using the GPU very early if at all at around the 1.4% mark
though I have noticed that for the work units that I have tried to process the dnetc_r72_1433334486_12_768_0 complete without issue but the dnetc_r72_1433334486_14_768_0 will start and just hang there and don't seem to be using any GPU. I am currently cruncing a wu with a 13 in the name and so far it looks good at 20% into the wu. the 14's would stop using the GPU very early if at all at around the 1.4% mark
well it looks like the issue is not just the 14 wu's. the sample size was small to begin with but out of the first 8 wu's the only ones that complted were the 12's and the fails were both 12's. But later this afternoon I found that a 12 got stuck also where it was not using the gpu. this time instead of aborting the wu I decided to see if exiting BOINC and restarting would make a difference ... It did .. the stuck wu restarted and began to use the GPU.... bad thing is that out of 14 wu's so far I had to abort or restart 3 different wu's. So unless I figure something out where this dows not happen anymore I may have to just complete the wu's in my queue and move on for now.
Have you checked the Moo website? anything there on the problem?
I've had WU's lock up on me before I just restart and they usually finish, if they have gone for too long I delete them. But mine is only occasionally like one in a thousand WU's.
I was considering popping out the 780's doing GPUGrid and popping in a couple more 7970's for a bit just to pump a few more WU's into the total and see if they quit chasin.
They can't catch us but they are recrutin trying to.
I've had WU's lock up on me before I just restart and they usually finish, if they have gone for too long I delete them. But mine is only occasionally like one in a thousand WU's.
I was considering popping out the 780's doing GPUGrid and popping in a couple more 7970's for a bit just to pump a few more WU's into the total and see if they quit chasin.
They can't catch us but they are recrutin trying to.
I thing its my GPU running too hot.....I see it at 85-88c and I think it may be shutting down at 90 or so.... I have not checked if I have that see in my config or not but that is pretty high anyways and I may not want to go higher. My gpu crunching will be tapering off somewhat now that it is getting hot. will be in the 90's this week and my upstairs ac just cant keep up ... especially with my downstairs ac totally not working. I would replace it but I have not been working for a while now and just don't have the cash.
I may try and swap out the 270x with a 750ti. I saw on the site that they have optimized their code for Nvidia cards ... so I might see how that one performs.
I may try and swap out the 270x with a 750ti. I saw on the site that they have optimized their code for Nvidia cards ... so I might see how that one performs.
[url=http://arstechnica.com/civis/viewtopic.php?p=29152017#p29152017:1pvmmlpm said:
Yeah, Thermal throttling will give the appearance of a "stuck" WU also. What I did to resolve the heat issues with dual Maltas is devolt and declock the card in Afterburner. Normally they run at 1.2 volt and 1K clock speed. Running Moo at full bore they shoot right to 99 degrees and throttle topping out at 102 degrees. You can run them as low as 500 hertz and .95 volt. My malta's and tahiti's refuse to run under .95 I suspect the Hawaii's are the same being based upon the same chip.
I've managed to keep them all running now for three months with little to no interference at maximum devolt and medium declock settings. It doesn't seem to hurt the output any, they still crank out WU's at a 20 to 24 minute clip 95+ % usage.
If your running at 85-88 degrees you might want to try slight devolting from stock. It helps a lot. Stock operating temp is 85 degrees, keep them under 84 and your good to go.
Well, it has been five months to the day I first started Rc5-72, Seven days later I started Moo.
Tomorrow, I crack a billion credits, and jump three spots into fourth place on the overall leader board.
A round of drinks for my compadres!
Go Ars!
Tomorrow, I crack a billion credits, and jump three spots into fourth place on the overall leader board.
A round of drinks for my compadres!
Go Ars!
Congrats EG.....
On my end I will have to postpone any Moo work. I have dropped the voltage down 8% and the the clock speed down to 950 and I still see my card hit 95C. And of course still getting the stalled workunits. I will have to wait till it gets cooler outside....
On my end I will have to postpone any Moo work. I have dropped the voltage down 8% and the the clock speed down to 950 and I still see my card hit 95C. And of course still getting the stalled workunits. I will have to wait till it gets cooler outside....
[url=http://arstechnica.com/civis/viewtopic.php?p=29180973#p29180973:4nodasch said:
I can completely understand, If it wasn't for the 14k BTU in the window I would be down to three boxes right about now. Back when I was doing Collatz I didn't know any better, and suffered thru 105 deg days in the room where they were at, not anymore.
Haven't reached full summer yet here in the PNW but when late July/August comes around I expect I'll have to take a few down myself. Not worth burning anything out.
Take care brother...
I just stopped by to say I had dinner with one of the original distributed.net developers the other day. I had not idea his name was still on the staff list.
He and everyone he knows has been gone from there for years, I believe. He went into web hosting and then finance.
Well I guess hardware issues means that they are not producing WU's until they get it fixed.
I already have one box down. still have a few more day work on the others but what happens when they all run out?
Any projects worth doing that could use my herd?
Anyone notice that the dnet stats server hasn't changed/updated in 6 days?
I already have one box down. still have a few more day work on the others but what happens when they all run out?
Any projects worth doing that could use my herd?
Anyone notice that the dnet stats server hasn't changed/updated in 6 days?
[url=http://arstechnica.com/civis/viewtopic.php?p=29376883#p29376883:3dggk79x said:
Yeah, staff blogs say that stats will update when Keymaster works again.
My box ran dry 2 days ago.
No word at all from anywhere I can find.
Crunchin my last four WU's now. In the time being I've switched over to Einstein, if there isn't any movement on the Moo front in a week, I'll optimize them for production. Currently they are doing one per GPU.
I wonder if Moo will ever come back.
Crunchin my last four WU's now. In the time being I've switched over to Einstein, if there isn't any movement on the Moo front in a week, I'll optimize them for production. Currently they are doing one per GPU.
I wonder if Moo will ever come back.
Haven't heard or seen anything, I went and took the privilege of asking for an update on the sites message boards. It's been 8 days since he posted the out of work message.
Sorry, meant to post sooner, last I heard, they are still working on getting the replacement keymaster installed at the co-location facility and then the work begins to recover any data from the old system and get the keymaster back up and running.
I haven't seen anything official yet but my proxy servers are full of fresh work.
[2015-07-22 21:56:14] ogrng r=2000/2000, d=0/10, 20058.5 Mnodes/sec, tot=3590459/3579424 stubs :devious:
Yep, My work buffers are full also and I'm back on Moo 100%. Now, to get Blackhawk 1 back up and running! (although the numbers in Einstein weren't looking all that bad either, without optimizing)
(although the numbers in Einstein weren't looking all that bad either, without optimizing)Looks like resellers are dumping Xeon Phis. You can get the higher-end 5110p for just a bit more than that. I'm kinda tempted to give it a try.
(Edit: I should probably clarify that it's the older 22nm ones getting dumped, Intel will be releasing a newer 14nm one soon which is supposed to be much more powerful.)
Guru, did you ever try running OGR-NG on a Xeon Phi?
(Edit: I should probably clarify that it's the older 22nm ones getting dumped, Intel will be releasing a newer 14nm one soon which is supposed to be much more powerful.)
Guru, did you ever try running OGR-NG on a Xeon Phi?
I wish but there is little software support. The other issue is they require specific Intel server boards for running.[url=http://arstechnica.com/civis/viewtopic.php?p=29481225#p29481225:s7e6crat said:
Still cranking along, but weather issues have cut into my production the last few days. (intermittent power outages) but they will be back to full strength in a few hours.
Although my average will take a hit I'm afraid.
Although my average will take a hit I'm afraid.
[Looks at all these guys cranking out huge numbers of units. Shakes head. Goes back to old laptop running d.net.]
Well Guys, It's been a nice run, but alas I have need for my boxen to generate some financial returns. I've picked up an opportunity to make some income off the power these things consume.
My participation will drop for a while. I will keep what ever I can running as long as I can.
It's a contract job with a termination at some point, how long I don't know. But gotta go where they do the most good.
It's been a fun run....
EG
My participation will drop for a while. I will keep what ever I can running as long as I can.
It's a contract job with a termination at some point, how long I don't know. But gotta go where they do the most good.
It's been a fun run....
EG
Well, I rescued a toy from getting scrapped at work the other day. Four 12-core Opterons and 320GB of RAM, and they were literally going to scrap it! Unbelievable.
Not sure how long I can keep it running though. I can hear it anywhere in the house even through two closed doors, and it's not a steady white noise, either -- the fans are constantly cycling up an down. Also it noticeably heats my entire basement.
Not sure how long I can keep it running though. I can hear it anywhere in the house even through two closed doors, and it's not a steady white noise, either -- the fans are constantly cycling up an down. Also it noticeably heats my entire basement.
It doesn't seem to like Ubuntu, I might have to experiment with different distros. It doesn't seem to be getting the performance it ought, and /var/log/syslog filled up the filesystem the other day with USB errors.
Still, seeing "48 crunchers ('a'-'V') have been started" on starting up dnetc is pretty neat.
Still, seeing "48 crunchers ('a'-'V') have been started" on starting up dnetc is pretty neat.
I'm almost tempted to replace the 6176s with 6276s just to see what dnetc does when it runs out of lowercase and capital letters.
Heh finally found out that linux 64-bit is 20% faster than 32-bit in OGR-28, how many years have I been running this?
finally found out that linux 64-bit is 20% faster than 32-bit in OGR-28, how many years have I been running this? 
[url=http://arstechnica.com/civis/viewtopic.php?p=30941679#p30941679:233hxweu said:cwhyl[/url]":233hxweu]Heh
Well, at least you found out...