I stumbled upon mention of this project while in the AQUA@Home forums.
(http://aqua.dwavesys.com/forum_thread.php?id=433&nowrap=true#5804)
QuoteIf you are looking to boost credits, pop over to Freehal. It can be run in parrallel, in addition to any other Project with minimal to zero cpu hit, in effect something for nothing - not quite, but not far off, cpu use of about 10 seconds or less over a WU crunch time of 1hr 50mins.
Freehal Home Page: http://freehal.net/freehal_at_home/index.php
The Project is about Artificial Intelligence and interaction between an AI and human asking questions. The WUs provide the core material for the AI to munch on.
Practical realities:
- Each machine can run 25 WUs maximum and they run concurrently using between 900Kb and 65mb per WU. Crudely plan on 25 WUs taking up 1Gb of memory as not all peek to 65mb at the same time.
- If you have 4Gb and above RAM, you can also run one virtual machine in addition to each physical machine. You are allowed to run the vertuals where you like, as long as the all up vertual total does not exceed the physical machine total. EG: Got 4 machines, then you can startup 4 vertuals as well.
- Its RAM dependent, not CPU, so you can run it with any old machine, really old ..... and end up with 4,700 credits a day from it.
- Numbers of concurrent WUs per machine are Cruncher selectable from 1 to 25, so you can tailor each Machine to available RAM.
For Example my Phenom 2 (64bit 8Gb RAM) routinely runs Aqua on the four Cores, Collatz on the 9800GTX, 25WUs in main RAM, and three vertual machines of 23 WUs each using 1280Mb RAM per vertual. Aqua credits on a quad you know, Collatz gives me about 20,000+ on the card, and the Freehal 91 WUs gives me about 17,000.
I have two other machines running Projects ( a laptop and an FX60), and on those I also run 25 WUs each. So for three machines & three vertuals, I'll get about 27,000 credits a day for no effort, and miniscule hit on CPU ....
Its known as a "Non CPU Intensive Project" or NCIP, which as a genre uses little or no cpu to run. Its therefore only limited by the ammount of RAM you have spare, or put another way, is a great way to use up any spare RAM without hitting other ongoing Projects.
Slight downside is that it is an Alpha, and there has been a little stop start with some issues over the last 12 months. However, it continues to get better, and last night ran a new setup which is working extremely well, resulting in a reliable supply to the maximum allowed, and a stable server. Its working well, and now my three machines are tripping off a combined total of 141 WUs every 1hr 50mins for 15 credits per WU. Worth a look, if you fancy it, I would suggest you run one machine first for a couple of days to get into it, but after you are in there, assigning the rest is almost a no brainer as cpu hit is vertually zero.
I have no doubts more issues will appear, it is a developing Project, so dont expect everything to be sweetness and light, but the reconfig last night was a huge step forward, and its now at a stage where I would recommend anyone to give it a go. Those with Farms will add 4,700 RAC per machine, 9,400 if paired by a vertual. If anyone has a big farm of circa 10/20 machines ..... thats a chunky addition for almost zero effort and negligible effect on other Projects :)
EDIT:
One querky thing to be aware of, the WUs start up in succession, go to 100% on crunch time, then over a period of about 5 mins collate results ready for upload - during the final 5 min collation the counter is at 100% because its already done its crunching, hang on rofl, its still got collation to do. Once aware of the behaviour, its no problem, it'll just seem odd at first.
Regards
Zy
Well that was enough for me to try it out. I've been running 10 WUs on four different machines for most of today.
So far, I've had two issues. The x64 Ubuntu machine ran the first batch of WUs fine but locked up completely part way through the second. It did this once a while ago but that was fixed by removing and reconnecting the network cable (standard troubleshooting since I usually connect remotely to this box), but that didn't help today. The second issue was when I rebooted the Win 7 x64 box for a Windows Update, the progress on the WUs was lost and they restarted from scratch. That's not so bad.
Admittedly an Alpha project but it looks good to me. We are currently ranked 94th on RAC, 85th on credit, so I am to improve that a bit, and your help would be appreciated!! biggrin
http://boincstats.com/stats/team_graph.php?pr=freehal&id=92
Sounds interesting. I have attached and started my first task. Looks like it will take about 100 minutes or so to complete. I adjusted my preferences to do 25 at once but it won't download any more yet. Maybe it will do so when this one is finished.
Thanks for letting me know about this. :)
Yes, after it finishes and uploads that one WU, it will bring down the full batch.
I started slowly, and I'm up to 12 at once which looks like about the sweet point for free memory on my machines. That should work out at about 11,000 credit per day (5 boxes x 12 WUs every 90 mins x 12 credit per WU) if I'm lucky!! I'm seeing very little impact on existing CPU-intensive tasks like CPDN and AQUA.
Yes that's right, thanks.
Ah that's better, now I'm cooking with gas. 25 on the go. Bonzai, go go FreeHAL!
Hmm, that's strange, it says granted credit of 12 but it actually gave me 15. Must be a 25% bonus because I'm such such a fine fellow. :rofl:
Hmm - I'm the first one to take a free lunch, but this sounds too good to be true. Surely there must be a catch???
There does not appear to be one - yet. But you are right to be sceptical! So far today, 5k credit in exchange for something I don't yet realise I've lost :shock biggrin
.. and you were user of the day (until veebee usurped you biggrin )
Quote
Admittedly an Alpha project but it looks good to me. We are currently ranked 94th on RAC, 85th on credit, so I am to improve that a bit, and your help would be appreciated!!
Ok, I'll give it whirl...
You know I have been running this project for some time on and off but I never have I seen it run more than one WU at a time. It does run independant of the other work though so 2 cores can be running projects and freehal will run as well.
@ Dingo - you go into preferences at the project, and select (up to) 25 WU's in "maximal concurrent work units" setting, then click "change".
the of course, you must update freehal in your boinc manager.
Yes I see that now, either I didn't see that way back when I attached or it was added after, but I have it selected now. My laptop is running 25 FreeHal, 2 Ibercivis, and 1 Collatz. I have 4 GB of RAM and at the moment only using 1750Mb but it is going up as FreeHal runs. the 25 WU are at about 30% complete.
PS: My HD4770 came in the mail today, now all I need is for my furniture and PC's to arrive.
you got a 4770 ? cool !
great bargain if you can still find someone stocking them.
stick it on collatz for a while mate and get 50K a day out of it - THAT'LL get your RAC back up !
Quote from: veebee on January 28, 2010, 04:43:18 PM
you got a 4770 ? cool !
great bargain if you can still find someone stocking them.
stick it on collatz for a while mate and get 50K a day out of it - THAT'LL get your RAC back up !
.. or Milkyway - even more credits than Collatz!
that's incredible.Does 15 WU's at once .Just tried it on my cmp and got 1000 credits in an hour. UNBELIEVABLE
Dave
Dave, go into preferences at the project and you can set it to do 25 WU's at once !
edit: have just set up a virtual box (win 7 x64) inside one of my other win 7 boxes - used the real install disc without entering a key No - used the loader program I have which inserted a key and activated it for me.. perfect - unlimited installs off one disc !
so now running 50 WU's at once on that machine.
Okely Dokely buddy. Thanks
Dave
well, 0530 hrs and there's no wok left...
seems like everyboincer had a run on freehal ...
I'm curious as to how long these wu's are supposed to take ? do they use cpu or gpu or both ?
i just checked 1 box and it was still crunching 1 wu after 90mins ..if this is the case then 12 credits for 90 mins work is pathetic.
They take between 90 and 110 minutes on my machines. However they only use about 10 seconds of CPU time. No GPU. They are designed to be very light on CPU and just use RAM. They sit on 0% for about 8-10 mins, then when they reach 100%, they take another 5-10 mins before completing.
You get 15 credit, but the stats page at FreeHAL shows 12 due to old calculations that have not been updated. You can read more in their forum.
Quote from: veebee on January 30, 2010, 05:32:27 AM
well, 0530 hrs and there's no wok left...
seems like everyboincer had a run on freehal ...
Seems to be OK now, I am getting work :yahoo:
Now the sites crashed. Thought it was too good to be true :compbash:
Dave
bugger just when i had 10 wu's to return :faint:
Yeah, I've got a few due back too. The server has had pretty close attention by the admins in the past few days so hopefully it will be sorted in the next day or two. It is an Alpha project so these things can happen :cry2:
In the three days we've been giving extra attention, the team RAC has jumped from ranked in the 90s to the 40s and still climbing steeply.
http://boincstats.com/stats/team_graph.php?pr=freehal&id=92
There back up and the work units are flowing,,,
Yep crunching along nicely again. The outage was caused by a DDoS attack. The administrator needed to block 117 IP addresses.
I myself suffered a BDoC attack yesterday (BOINC Denial of Crunching), when the benchmarking that happens every 5 days caused all 25 tasks to restart after 90 minutes. :)
If it happened with a project that used any resources I would be tempted to try out the --skip_cpu_benchmarks command.
The team is roaring up the charts, that's always good to see.
looks like the site is down again - cannot upload or even get to the site.
Is there somewhere we can get the "alternate" address or "proxy" or whatever they are called ?
I remember similar happening with another project when I first started and we put an IP address in the HTTP proxy (or the SOCKS one, cannot remember exactly) to allow uploading/ downloading of work etc.
Hmm the title of this thread is a bit misleading now!
I haven't read anything about this site having a proxy or alternate server. I expect that only helps for projects with a working server and redundant ISP connections? I may be wrong, but I can't see how a proxy could collect or issue work if the project server is down. It would be great though - I have a bunch of WUs ready to go up.
Before the site went down, the home page mentioned they were hoping to more onto a new, dedicated server (i7) & array, independent of the previous YoYo hosting, this weekend.
Hope they get it fixed so it is more reliable. I like the idea of doing some extra work and only using minimal CPU resources.
If anyone wants to do some SIMAP, work units are currently available. At a team ranking of 31, this is another project that is rated urgent on Teemac's chart. Haven't done any SIMAP for a while, can't remember what the credit rate is like. It's probably not generous like most medical/biology projects but that won't stop me doing a bit there.
For anyone following along the FreeHAL story... In the last 2 days they've been dishing out new version WUs that throw computation errors on all my boxes, but in the last hour there is finally some valid work completing (win7 x64 anyway), and supply seems good.
Give it another go if you're game :thumbsup:
The team is certainly jumping up the chart so congrats everyone :congrats 39th by project RAC, 77th by total credit.
UPDATE: Success is a state of mind :) Some of the WU batches are still failing. It's unpredicatable. But the only impact is bandwidth for the failed downloads, as they use almost no CPU.
Version 1.30 of the application finally seems to have it working on XP 32 bit. Free Lunch time again biggrin
I notice the app has a memory check in it - I have an old laptop on which it concludes there is not enough memory for 25 tasks and downloads only 18.
I had a hell of a time getting shod of those pre-1.30 work units. The things kept running until I figured out the exe's weren't being deleted and were hard to get rid of even after I did. :thumbdown:
Reset, kill all freehal processes, reboot, reset again or detach and then watch the messages for any warnings about "Cannot delete XXXXXX".
Finally got rid of them from 3 machines. :thumbsup:
Plus one for the heads up.
The v1.30 work is crunching along nicely for me too now. :thumbsup: It's almost doubled my daily output; not as good as a GPU but cheaper biggrin
We're doing great overall, with ranking now 19th by RAC and 61st by credit. Thats up from 94th/RAC, 85th/credit as at two weeks ago.
Latest bunch of wus I downloaded for TheGnat seemed to have doubled in time taken, or thereabouts. Thought that TermiGnator was having chickens when I saw how long THEY were taking there.
No checkpointing remember. Possibly the tasks were interrupted by something and started again. When BOINC benchmarks every 5 days, it only takes a short time but it is long enough to cause FreeHAL tasks to start again from the beginning. See how your next batch goes.
You're right kashi, BOINC benchmarked on my main box yesterday. With FreeHAL tasks at around 90% completion.. :furious:
Correct again Kashi.... and with a PSU apparently also on its last legs..... :compbash: :hbang: