BOINC

Also, how did you attach the host? Did you use BAM! or did you attach manually?
 
I attached projects/hosts at the BOINCStats web page. BOINC Manager should be called BOINC Assistant. Seems like the web page is the real manager.
 
The boinc manager will controll how many WU's it will download and when if you tell it to run based on preferences which is found under the activity tab.
 
I got the server set up with BOINC now. No GPU in that machine so WCG and WUProp only. Not sure if it's worth running it. We'll see.
 
Worth is whether you can affordd to run it... lol Especially since it shouldn't be compared to GPU projects for points. That would be apples to oranges. I'm still running some pretty old equipment because I can. I'm limited mostly by my home's electric setup.
 
I asked Carle at QCN about posting the stats at their sister sites stats on the various stats sites. Here is his reply:

hi, well the "http://qcn.stanford.edu/sensor/" is probably the only worthwhile site as far as boincstats & credits etc, the other sites are for very specialized, local projects. also QCN will probably be winding down this year due to lack of funding so it probably wouldn't make sense to run stats for every sub-site just for a few months.....
Click to expand...
 
SIMAP posted this on their home page:

Many smartphones are a supercomputer! Samsung's Power Sleep App has been launched by Feb 14. This Android app is not only an alarm clock, but also supports SIMAP. During the night, when connected to WiFi and after fully loading the battery, the CPU of the smartphone or tablet crunches SIMAP work packages.

The scientific computing part of Power Sleep is just the BOINC client. The Power Sleep App has been developed by Samsung Austria and PocketScience in collaboration with the SIMAP team. It makes donating CPU time very simple for people who didn't know about BOINC before. Samsung has launched a media campaign about Power Sleep in Austria. The app, however, can be downloaded from all over the world. We hope this initiative will become a big success and will help to widely establish volunteer computing with BOINC on mobile devices.
Click to expand...

http://www.samsung.com/at/microsite/powersleep/

English version: http://www.youtube.com/watch?v=MLalcVLLbHU
 
I don't know if it's been mentioned, but the latest version of BOINC automatically detects and attaches work units to all co-processors. So no more adding the cc_config.xml for that specific addition.
 
What's the proper way to upgrade? Just install and overwrite the old version? Or does the old version need to be uninstalled?
 
Just overwrite the old version, no uninstallation required. My find was on a fresh install of Windows of course, so nothing left over to make it see the GPU's.
 
On Windows, you can just install over the old version. On Linux, it would be best to remove the old client first from what I hear.
 
If you are running WCG, you would find a point bump using 64bit Linux on the VINA apps. Even under a VM. People claim anywhere from 30-70% point increases. I however have pretty well stuck with Windows as well.
 
I have a pair of Gigabyte 750Ti cards arriving tomorrow to replace my 550Ti.
 
Last edited:
Gilthanis said:
I have a co-worker who just ordered one for gaming. Too bad I can't get him on board with DC'ing....:(
Click to expand...

I think someone doesn't deserve their Maxwell then.

It's installed and currently running Einstein BRP5 units. The estimated completion time is 4:30. But we'll see!

I just realized that my GPGPU box is CPU underpowered, so it's down until I order something else to put in it. It has a 25W Xeon 1230L V3 at the moment. In retrospect I realize what a mistake that was... Oops!
 
Kai said:
I think someone doesn't deserve their Maxwell then.

It's installed and currently running Einstein BRP5 units. The estimated completion time is 4:30. But we'll see!

I just realized that my GPGPU box is CPU underpowered, so it's down until I order something else to put in it. It has a 25W Xeon 1230L V3 at the moment. In retrospect I realize what a mistake that was... Oops!
Click to expand...

Max # of PCI Express Lanes 16

Ouchies for a GPU box.....
 
Oh it's not that. They all run at 8x, which isn't the issue. It's not feeding them fast enough. CPU time is drastically higher. I've got something faster and full powered coming though. Gen 3 8x is plenty of bandwidth for these cards.
 
Yeah...I haven't felt that hit yet since I'm still old tech on the GPU front. :) But hopefully I will get that straightened out here soon. That is stage 3 and I'm still on stage 1 for my current plans. Once I get this file server going, then I can start building the replacement for the i7 920. Then I can get the video cards upgraded. :D
 
Lol, you'll get there soon my friend. Very soon.

750 Ti Einstein BRP5 was completed in 2.38 hours as opposed to 3.07 for an R9 270 that was in the same system. And around 2 hours for a 7970. Very fast indeed!

This is with a 4570S at 2.9Ghz.
 
Last edited:
That's wonderful news! I'm glad the crunching has real results such as this. It really makes you proud and happy to crunch.
 
Harmonious Trees and Odd Weird Search are now considered complete at YoYo. It is a shame because I was hoping to get my badge in OWS before it finished. IMHO it was dumb to take on a very small sub project like that. OGR-27 is also complete, but they are now sending out OGR-28 work units.
 
Looks like Einstein has the Gamma Ray Pulsar Search 3 OpenCL app out again. It uses a lot of CPU time, but the nvidia version on the 750Ti takes 4630 seconds with 4618 seconds of CPU time. Also, very stable times for the 750Ti on the BRP5 app. Around 8500 seconds to complete with 1800 seconds of CPU time with a low power 4570S. In comparison, a 290x with a 4930k takes around 8600 seconds to complete the BRP5 with 3500 seconds of CPU time on average. That's pretty amazing what they've done with the CUDA app compared to the OpenCL BRP5 app.

Oh, an OEM 760 takes around 7800 seconds to complete the BRP5 with around 2000 seconds of CPU time with a 4770 non K variety.
 
Last edited:
Yeah...CUDA is a ton more efficient on nVidia cards compared to the "clumsy" OpenCL. I say clumsy because it isn't really made for any one specific piece of hardware and programmers find it much more difficult to learn and use than CUDA.
 
I see Grandpa added WUProp. Now he just needs to attach those other beasts. He should rac up run time on various apps rather quick. Let's see how long it takes him to mow us over on those badge levels. :p

Brilong, how many of the machines that you have there are physical machines vs. VM's? :eek:
 
Gilthanis said:
I see Grandpa added WUProp. Now he just needs to attach those other beasts. He should rac up run time on various apps rather quick. Let's see how long it takes him to mow us over on those badge levels. :p

Brilong, how many of the machines that you have there are physical machines vs. VM's? :eek:
Click to expand...

I can answer that, ALLOT I belive most if not all of them are physical machines. :eek:
 
And some of our newer BOINC users may not be aware, but at WUProp specifically, you can show what projects you are accumulating run time at under your WUProp@Home preferences.

Should WUProp@Home show app hours statistics on your account data page and export statistics? is set to No by default. If you change it to yes, when someone clicks on your name, they could also see what apps you have earned time for towards your badge rather than your computer list and such.
 
Swapped out my 550Ti for the two 750Ti cards yesterday. So far things aren't going as expected. I want to run GPUGRID on them but it insta fails. I got Einstein running on the cards ok except that it doesn't seem to be using the cards to their max, barely using them even 50%. The fans don't ramp up and temps are not much higher than at idle. Is this how Einstein rolls or am I doing something wrong?
 
Not all apps use the GPU's to 100%. This is where running multiple work units on a card comes into play unleashing it's full potential. Try using an app_config file to run more than one at a time. Einstein also has the option for running more than one app at a time on your cards at the project preferences from their website. Keep in mind that changing the settings at the project effects all of your hosts. Where an app_config file will only affect the one host.
 
The 750Ti runs the Einstein Gamma Ray Pulsar Search 3 faster than a Titan or a 780Ti... 5400 seconds vs 5800 seconds on average.
 
You must log in or register to reply here.
Back
Top