April DC Commandos

E

etothe3ric

Limp Gawd
Joined
Feb 4, 2009
Messages
422
Another month has come Commandos!

We have a collective it seems of a push in Einstein@home. All projects are still open to suggestion.

Currently we are at position 279 with 5,226,047 points. To register for the project it is similar to all BOINC projects were you can add a project through the program.

Website Link: Einstein@home
Description: This project uses both CPU and GPU WUs and is currently looking for more information about "pulsars". A pulsar is a gravitational wave from a spinning neutron stars.

This is the project that got me addicted to DC so it will always have a special place for me.

If you can get a following for another project we could have a dual project month.
 
I'm in.

In fact .... I got a running start last week start because I quessed all you big dogs were going to steer major hardware towards Einstein@home soon. Wave when pass me. :)
 
Has anyone had problems when they run certain projects it cause your computer to lock up? I'm having that issue on my i7 with E@H but I didn't when I ran Spinhenge. It will usually will lock up once a day when I am at class. Any ideas?

Currently I am running SoB with Razor on the i7 to keep up my output.
 
Yes, Einstein@home seems to be a very demanding project. My 980X was bluescreening regularly until I backed off on the overclock a bit, even though it had been stable running F@H bigadv and Primegrid. That's the only box where I had stability issues.

In addition to Einstein on four boxen (including some GPUs), I'm running GIMPS on two boxen and DPAD on one. I just realized my DPAD box lost its overclock at some point so I need to fix that.

I realized another difference with my "slow" GIMPS box is that it's 32-bit. Maybe I can install the 32-bit software on my 64-bit box and see if it slows down to match.
 
Does Einstein run well on ATI GPU's ? Does it work on GPU's in linux ?
 
There does not seem to be GPU units for ATI. For linux however, it has units for CPU and cuda GPU for linux.
 
etothe3ric said:
There does not seem to be GPU units for ATI. For linux however, it has units for CPU and cuda GPU for linux.
Click to expand...

cant do it then ;( i use the 1090T for F@H or WCG, and only have a ATI card in this system(only one with power to spare)
 
Just arrived yesterday from UPS.

Ram and MB arrive Tuesday for my new server. Hopefully can get something up and running by the end of next week. Einstein is no match for Xeon.

Should be a huge bump from a D525 Atom.
 
Since I sold the last of my nVidia cards I wont be participating in Einstein this time.:(
I am trying to move forward in Simap and keep us steady in Aqua.
Thanks for joining in on SOB Eric that will help move things along nicely.
I see I have some help in Aqua today from kenlo. This is great since I am falling behind.
 
I have just finished ordering the parts for my 2nd Windows 2008 R2 DataCenter that runs x4 Hyper-V Windows 7 clients for crawling on Majestic-12 since I will be upgrading my Internet connection from 50/5Mbit to 100/20Mbit late this summer when it is available. Team member McCoy will also be making this move to the 100/20Mbit line to help with the efforts.:)

Looks like we will move up from position 15th to 14th in about 9-10 days and with the new line upgrades later this year we are planning a push into the top 10 !:D
icon_mj12.gif
 
I have been very confused with MJ12 as of late. It seems there new plan is an SEO backlinks model and they gave up on the true search engine based model. Do you know more about it Razor or is this what info you have been getting. I tried to look into the DC-Vault forum about it before because they were debating in removing it.
 
We are still on track to having a true text search engine. We are making great strides toward this since the new bucket server installed Oct. 2010.
We should be pretty far along on the text search engine later this year. I will be receiving my first stock dividend payment here in a month or two (not much for now, but will be someday).
The back links data is great to rent/sale to customers now to help with hardware and development expenses but the true goal is still a full text search engine.

If you plan to participate again, be sure to ask me any questions you may have since the node setup for different networks (ie. routers or using ISP routers) is much more challenging and hard to configure for the average member compared to just a year ago.
 
Thamks Razor for the clarification. So it seems the SEO data and pricing is to be used to create money to support the full text search engine.
 
I upgraded my GIMPS boxen to the latest version, and the speed hasn't changed. It's either the slower RAM or 32-bit that's causing it to run slower. I'll try running the 32-bit code on the 64-bit box and see what speed I get.
 
Eric I will be off SOB until Wednesday due to robbing parts from that computer to build the new Windows Server 2008 R2 DataCenter for MJ-12 but should be back on it by Thursday.
 
CraftyOtter said:
I'm in.

In fact .... I got a running start last week start because I quessed all you big dogs were going to steer major hardware towards Einstein@home soon. Wave when pass me. :)
Click to expand...

Ok now I'm really in. There are two HardOCP teams for Einstein@Home. I joined this one by mistake. So much for the running start. :D
 
Hey guys. I'm switching my work stations over to Einstein@Home and I'm confused about something. I have several WUs downloaded but only one is being worked, and it seems as though it's only using one core on each computer too.

Any insights? I had multiple WU running for WCG over xmas, not sure why it's not working on this.

Thanks!
 
Razor, I have not been able to work on SoB recently because my i7 is becoming very unstable. It will not run with a WU for more than 20 mins before it reboot. I cannot find the root cause of this issue. If anyone has had this issue before or something similar please make a post because I am running out of ideas. :confused:
 
I'll watch the show from the sidelines this month - I haven't had the time to think about switching projects, so both my Quads are still happily chewing away on Spinhenge... Or for my excuse, I don't want us to lose any of the places we fought so hard for! See you in May... :)
 
It looks like we may be able to crack the top 200 in Einstein@home by the end of the month. That's a good goal. It seems like the competition gets tougher the higher we get. I've pulled most (all?) GPUs off Primegrid to focus on Einstein.

Briliu, check your BOINC configuration to make sure you have it set to use all cores, etc. Also, Einstein has both CPU and GPU (Cuda) work and will download both by default. You can check your project configuration on the site to make sure it's set to download what you want. I set up a couple of custom configuations so that my "school" box (Win home server) gets CPU WUs only, and my "work" box (GIMPS box) gets GPU WUs only, and the rest get both.

I noticed I currently have 49,999,493 points in Muon1. :D After the next overtake, I'll probably switch that box to a math project.
 
etothe3ric said:
Razor, I have not been able to work on SoB recently because my i7 is becoming very unstable. It will not run with a WU for more than 20 mins before it reboot. I cannot find the root cause of this issue. If anyone has had this issue before or something similar please make a post because I am running out of ideas. :confused:
Click to expand...
Be sure to manually set your "FSB Strap to North Bridge" to the speed that matched your processor's factory default. Many people will leave this on Auto and change there memory speed to a number close to there max memory speed but this changes the FSB strap and can throw reboots when running prime like programs.
After setting it manually then select the memory speed closest to your Ram's rated speed. Be sure to check the max safe over volt on memory and set that up by checking "Hardware Monitor" tab (never set voltage to a number without checking hardware monitor after a reboot to see what it actually ends up at, same with CPU.)
 
Razor could you please look over these setting because I am a bit confused with what you posted.

I have an evga x58
Memory Freq: 1600 2:12
timings: 8 8 8 24 88 1T

VDroop control: with VDroop
cpu vcore:1.225v (set to auto)
cpu vtt voltage +200mv (set to auto)
cpu pll voltage: 1.8v (set to auto)
Dim voltage 1.65 (rated for the memory)
QPI pll vcore: 1.1v (set to auto)
cpu clock ratio 20x (133 x 21 2800mhz)
MCH strap 1600mhz (originally I had it set to auto, I think this is what you meant)
cpu uncore freq: auto

I change that one from auto to 1600mhz but I am not sure if that is what you were explaining to understand what effect that will have on my system.
 
SazanEyes said:
...

Briliu, check your BOINC configuration to make sure you have it set to use all cores, etc. Also, Einstein has both CPU and GPU (Cuda) work and will download both by default. You can check your project configuration on the site to make sure it's set to download what you want. I set up a couple of custom configuations so that my "school" box (Win home server) gets CPU WUs only, and my "work" box (GIMPS box) gets GPU WUs only, and the rest get both.

...
Click to expand...

Thanks SazanEyes, I went in and changed that over.

Apparently what was happening (since I started the client just prior to posting that) was that I was downloading a whole bunch of WUs for GPU first. So it took probably 2-3 hours to get into the CPU units after the first one it picked up. Luck of the draw I guess. This computer is chugging away on 4 WUs right now :) I haven't checked the other one but I've no doubt its working fine on that too.
 
Briliu, I forgot about that. Either Einstein WUs are huge or the project has a really slow server, because they take forever to download when you first join. Glad it's working now.
 
Yes Eric the MCH strap is the one. Be sure to manually set it to whatever the CPU is rated at. Hope that helps.
Is you're memory rated at 1T at that speed?
 
I checked online and a lot of people have it at 2T. I know 1T is much tighter but my motherboard on auto decided that 1T was what it wanted. I adjusted that and the +200mv back to auto for now to see if that will be more stable. I do some burning in to make sure it doesn't reboot. Thanks Razor.
 
The Administrator Alex over at Majestic-12 is considering donating a lot of CPU power to Boinc based team efforts for 24 hours at a time. I have invited him to help give us a boost when he is ready to do so.
We will need to figure out what project we would like to concentrate on so I can let him know when the time comes.
http://www.majestic12.co.uk/forum/viewtopic.php?t=4485
 
Last edited:
Alex accepted our invitation !

"Sure Razor_FX_II - invitation accepted, feel free to send me some preferred dates starting from later this month.
72 GB in each of the servers in that cluster, and some are now up to 96 GB
icon_lol.gif

We don't want to create our own team as we can't do it all the time, the idea of the "boost" is to help other teams under their name for 24 hours
icon_cool.gif

We'll need to take part in those projects where single core can do at least one full work unit in 24 hours (ideally more!)."
 
My suggestion for the Majestic-12 servers to work on would be a project that doesn't have a quorum of more than 1 and work units need to be completed in the 24 hour donation period.
The project should have lots of work units available to.:)
Any suggestions Commandos?
 
Thanks for getting the ball rolling on this Razor. That is very cool.

What do you think about SIMAP? Its one of our lower projects in the Medical Science category and it uses BOINC. Looks like there is a lot of work there. I have never run it so I guess I'd have to ask for input on the run times of the SIMAP WUs.

A little extra processing power will be great! Im headed over to lend a hand in Aqua as soon as my queue empties. :)

Edit: Rosetta@Home might also be a good one. Its going to take a ton of points to even get into the top 100 in that project.
 
Last edited:
You read my mind. I was thinking the same things.
The issue with Simap and lots of other projects is they have more than a quorum of 1.
I will ask Alex about this and see if that has to be a stipulation.

I had to pull back off Aqua for now and get back on SOB but hope to return soon after a finish building a new box in the next few weeks.
I have been hoping to get on Rosetta myself but find I need to add more D to cover all the projects I want to run.
2 servers, 4 workstations and a PfSense router box just isn't cutting the mustard.:(
 
I did some pseudo-scientific number crunching on the DC Vault stats. I looked at three factors: (1) points gained for increasing one position in a project, (2) maximum possible score increase in a project, and (3) current project position. I then ranked the projects by each of the three factors. For example, Prime Sierpinski PRP is ranked #1 in both points gained and max increase, because we gain 357 points every time we move up one position, and it has the lowest project score. Climate Prediction is ranked #1 in current position because it's our lowest (team rank 576). Next I weighted the ranks and calculated an overall rank for each project.

Based on all that, here are the top 10 projects that should be prioritized to increase our overall DC Vault position:

1. GIMPS
2. Prime Sierpinski Problem - PRP
3. OGR-27
4. Muon1 DPAD
4. Seventeen Or Bust (tie)
6. Wieferich@Home
7. Majestic-12
8. Aqua@home
9. yoyo@home
10. NFS@Home

For the curious, here's the rest of the list:

11. Docking@Home
12. Cosmology@Home
13. Enigma@Home
14. GPUGRID
14. ABC@Home
16. POEM@Home
16. The Lattice Project
18. Collatz Conjecture
19. Climate Prediction
20. uFluids@Home
20. SIMAP
22. Malariacontrol.net
23. SZTAKI Desktop Grid
24. QMC@Home
25. Leiden Classic
26. Rosetta@Home
27. PrimeGrid
28. Einstein@Home
29. Dimes
30. MilkyWay@Home
31. RC5-72
32. Seti (BOINC)
32. Spinhenge@Home
34. World Community Grid
35. Folding@Home

I'm pretty happy with how this worked out, so I'm going to try to put something online to update the rankings based on the latest DC Vault scores. I may also tweak the rank weightings a bit. Obviously, you can't just blindly follow the formulas, since it's harder to advance in some projects than others. I basically gave up on Wieferich@Home for that reason (but I'd be up for a push sometime if anyone's interested). It does help us pick projects, though, like either Aqua, Yoyo, or NFS for next month's BOINC push.

Also, seven of the top eight projects have a DC Vault score less than 9000, so a goal could be to get to 9000 in each project. That's going to be almost impossible in some projects, like getting to team #3 in PSP-PRP. However, it shouldn't be too hard in Aqua (goal = 111), OGR-27 (goal = 75), or Seventeen or Bust (goal = 57).

BTW, I'm not picking on Einstein@home - its rank has dropped since the beginning of the month when the current push started.
 
Nice work Sazan. For some reason, PRP and OGR-27 have always been placed on a lower priority within the Commandos. Glad to see your numbers match mine.
 
Razor_FX_II said:
You read my mind. I was thinking the same things.
The issue with Simap and lots of other projects is they have more than a quorum of 1.
Click to expand...

Both NFS@home and Yoyo@home have a quorum of 1. NFS is quick but can use a lot of RAM. Some Yoyo WUs run for a long time, and could be longer than 24h depending on hardware.
 
Sazan that is some great info. Thanks for doing all of the work to post this. I always try to look through our Vault page and do something like this (although not as detailed) in my head, but its great to see a listing.

We are getting close to another team in Muon. Aqua is a really active project. At one time I had us up in the top 100 or maybe even top 90 but the minute you stop on that projects you tend to drop places.

Maybe I'll go back to some NFS after a squeeze out a little more Aqua and some Muon. :)
 
Nice work Sazan. Just loaded NFS on my X6. The project is using close to 4gb of RAM. You were not kidding. And people ask what good is 16 gigs of RAM for:)
 
You must log in or register to reply here.
Back
Top