BOINC

Thanks Gil and brilong for the suggestions. I'll check out the 750ti. That seems like the best way to go.

If you go with a different 750Ti, look closely as there were a few models that actually took external power like other cards.
 
If you go with a different 750Ti, look closely as there were a few models that actually took external power like other cards.

I'll second this - I got one last year without reading the fine print, and it has a single 6-pin connector. On the upside, a simple double 4 pin to single 6 pin adaptor took care of the issue, and it still sips power! Doesn't really compare to my GTX 680 in raw performance, though... That's why I picked up a used 680 on the forums to replace my 9800GTX+ instead of another 750TI :D
 
On a related note, and one that you will love, Gilthanis - just managed to sideload NativeBOINC on my Kindle FireTV stick. 2 more 1GHz ARM cores for fun... will update how they go after a few days of results...
 
On a related note, and one that you will love, Gilthanis - just managed to sideload NativeBOINC on my Kindle FireTV stick. 2 more 1GHz ARM cores for fun... will update how they go after a few days of results...

This is good news as we have discussed in IRC if the Kindle FireTV stick would work. Any idea if the Berkeley version will work too? I picked up an additional 10 dual core cell phones over the X-mas sales to run a money generating app alongside BOINC. Some of the other IRC members have jumped on board with it too. They easily pay for themselves within a few weeks.... I think my phARM is up to like 17 ARM devices with ~28 total ARM cores. If you are interested on the details of the money generating app, PM me.
 
This is good news as we have discussed in IRC if the Kindle FireTV stick would work. Any idea if the Berkeley version will work too? I picked up an additional 10 dual core cell phones over the X-mas sales to run a money generating app alongside BOINC. Some of the other IRC members have jumped on board with it too. They easily pay for themselves within a few weeks.... I think my phARM is up to like 17 ARM devices with ~28 total ARM cores. If you are interested on the details of the money generating app, PM me.

I just might do that... Just as an FYI, running only one core, the past two days the FireTV stick has done 628 and 434 points (although today might have been limited a bit watching 2 movies. In comparison, my first Gen Galaxt Tab single core 1Ghz OCed to 1.2 has done 522, 568 and 605 over the past three days and has a 7 day RAC of 553. So right about on par with each other. Now that I know it's running smooth, I'm going to try both cores starting tomorrow...

But first, I have a Xeon X5670 to swap in for my i7-950, and a GTX680 to go in place of my 9800GTX - priorities, right? (I actually feel a bit guilty - the 680 should go in my daughter's rig, as it can run GPU 24/7, and her 750TI should go in mine that can only GPU ~10 hours/day. But for once, I want to game in full quality!!!)
 
Jathanis, just let me know via PM as the link has a referral link (not permitted to be posted here) and I would need an email to send you the old app file. (the new one requires you to click OK every hour to indicate you are still watching where the old app does not) I would also give you a brief run down on how to set it up and tweak it. I also have a site linked that has some refurbed phones that should work and are cheaper than eBay.
 
Is it worth running BOINC on a phone? I have an S5 and 2 S4's. Running it on a Kindle HDX but it doesn't give very many points. What are the best projects to run?
 
Running BOINC on phones/tablets/other ARM, is a tricky answer when looking at "worth". Many people start running them expecting results like a PC. If you are looking for high points, you will be discouraged. What you should go in looking at is points per watt. ARM devices are MUCH more efficient in that way. However, it really does come down to the project as they all have different needs. ARM devices in general don't have floating point and it needs to be emulated. This emulation creates a loss of productivity. But even with that loss, it tends to be better efficiency. Then you get into some projects that just can't cut their work units down to a short enough time frame like PrimeGrid. If you ran one of their LLR work units, it would take months. Some ARM devices are light on RAM too.

As far as suggestions, I find Enigma to work really well. WCG has work but they have had lots of trouble getting it stable. Their work runs best (and more successful) if the device is fully dedicated and thus doesn't pause.

One other thing to keep in mind is that Google really made some big changes with Lollipop. Currently I am not aware of any projects that are Lollipop ready. I know that WCG is working on it but it requires a recompile of the apps.

For a full reference of projects with Android/ARM work, please reference my all inclusive dc list here: http://hardforum.com/showthread.php?t=1801908

And if you would like numbers from actual BOINC results, you can always check the results from the WUProp project (which also has ARM work and is certainly a project to add to all BOINC clients) http://wuprop.boinc-af.org/results/arm.py
 
I guess I will just stick WCG on all the phones I just pulled out of storage. Now to run these not at my desk. Probably not the smartest thing running 4 phones off the motherboard either.
 
Not all phones will fully charge while running full load off of a usb powered connection. I have found that some require to hooked to a wall charger. Some phones are more sensitive to battery swelling than others too. If your phone will run without the battery attached, some recommend that. I've only had two phones get a swollen battery in 3 years. My Bionic got one after 3 years of BOINC heavy use and 4 years of actual total use and the other was an old Optimus V that was a crap phone to begin with. Batteries are cheap on eBay though...so I don't worry at all.
 
So try to get the x5670 up and running, no go. The guys in the 1366/x58 thread say I need to try the newest BIOS, but I'm not flashing a BIOS at 1am... Then I decide to try out my "new" GTX680 from the forums, and it refuses to give me any video from either the DVI-I to VGA connection OR the HDMI connection. So I pop it in the second slot, just to make sure it's not bad, and BOINC recognizes it right away and starts tearing through Primegrid WUs. So I guess I'm stuck with the 9800GTX+ for gaming and the 680 gets to be a dedicated DC card after all...
 
I am pretty proud of myself and wanted to share - I hit 1 billion points overall in BOINC today. :D :cool:

My next goal is 1 billion points in Collatz (I'm at ~665mil now).
 
I am pretty proud of myself and wanted to share - I hit 1 billion points overall in BOINC today. :D :cool:

My next goal is 1 billion points in Collatz (I'm at ~665mil now).

That's an incredible achievement! Congrats!!!!!!!! Now to celebrate, right?
 
I just might do that... Just as an FYI, running only one core, the past two days the FireTV stick has done 628 and 434 points (although today might have been limited a bit watching 2 movies. In comparison, my first Gen Galaxt Tab single core 1Ghz OCed to 1.2 has done 522, 568 and 605 over the past three days and has a 7 day RAC of 553. So right about on par with each other. Now that I know it's running smooth, I'm going to try both cores starting tomorrow...

But first, I have a Xeon X5670 to swap in for my i7-950, and a GTX680 to go in place of my 9800GTX - priorities, right? (I actually feel a bit guilty - the 680 should go in my daughter's rig, as it can run GPU 24/7, and her 750TI should go in mine that can only GPU ~10 hours/day. But for once, I want to game in full quality!!!)

I'm actually excited about Intel's compute stick, which should provide a decent measure of performance for watts consumed. I already have Zotac's equivalent running and the GPU is surprisingly competent. Bay Trail ain't so bad!
 
Well, I have a good news/bad news report. The bad news is that either my X5670 is DOA, or my board won't work with it - so no hex core goodness for now. The (mostly) good news is that my second 680 is up and running strong, so I'm now able to put up over 1M PPD on Primegrid and GPU grid! Oh, well... You win some, you lose some...

But I think I'm done with the upgrade bug for a while unless I see a socket 1366/920/6GB combo on the forums for under $200!
 
Well, I have a good news/bad news report. The bad news is that either my X5670 is DOA, or my board won't work with it - so no hex core goodness for now. The (mostly) good news is that my second 680 is up and running strong, so I'm now able to put up over 1M PPD on Primegrid and GPU grid! Oh, well... You win some, you lose some...

But I think I'm done with the upgrade bug for a while unless I see a socket 1366/920/6GB combo on the forums for under $200!

I actually sold my combo for just under $200 on eBay a few weeks back...
 
I hate to say it, but I have to go to Dallas for work next week and will have to shut down all of my crunchers from 2/8 - 2/13. I hate having to shut them down, but my house-sitter can barely turn on a computer, let alone handle any potential issues, so I have no choice. Plus, I have to admit: the prospect of a ~25% lower electric bill this month does sound like a nice little break from the mint I normally send my local electric provider every month.

I promise to be back at it [H] as soon as I walk back through my door on 2/13! Gotta recover that RAC quickly! :cool:
 
Okay, I lied - my ebay seller confirmed the chip was DOA when they tested it, so I now have another confirmed working Xeon - will install tomorrow. Then I need to get more memory! My 6GB for 8 threads is already getting overwhelmed by both Rosetta & NFS by the 600-750 MB WUs. Time to go from 6 to 12. Or 18, just to be safe...
 
Back from Dallas and back to crunching. All rigs fired up and churning out work.

Slingshot engaged... :cool:
 
Slingshot engaged... :cool:

Love that movie...
talladega-nights.jpg
 
Since BOINCStats declared it dead, I have removed Spatiotemporal Quality Of Service from the All Inclusive DC Project List
 
I wish I had. It would make me sound cooler than I am. :cool: However, it is a shame as it was a non-CPU intensive app. BOINC needs more of those...
 
new project added to the All Inclusive DC Project List

Van Der Waerden Numbers
 
New BOINC Client available for testing:

BOINC 7.4.42 available for testing for Windows and Macintosh.

Preliminary Change Log 7.4.36 -> 7.4.38


MGR: Immediately save changes to the registry when sort column or direction is changed.
LOCALE: Update PO Compiler settings.
LOCALE: Update BOINC Manager template.
LOCALE: Tweak source location.
New files added from Pootle Demo based on templates.
Mac installer: Fix URL to wiki for information on GPU computing.
Mac installer: name install folders with "x86_64+ instead of "i686" to show that they require a 64-bit capable Mac.
LIB: Fix potential parser problem with applications compiled against really old BOINC API code (pre-2012).
MGR: Work around a probable bug in wxWidgets on Linux which prevented some wxListCtrl rows from updating.
MGR: Add "More info ..." hyperlink to Diagnostic Log Flags dialog; clicking it takes you to the Log Flags wiki page.
MGR: To allow WCG and other "branded" BOINC Managers to customize the "More info ..." hyperlink in the Diagnostic Log Flags dialog, build the URL using the value from the skin's <organization_help_url> field.
MGR: Build break fix when building with wxWidgets 3.0.2 on Linux.
client: add <dont_suspend_nci> config option.
If set, non-CPU-intensive jobs are exempted from mass suspend (e.g. suspend CPU, time of day, CPU throttling). You can still suspend them individually.
LIB: We do not need to log trace statements on release builds (MSVC only).
LIB: Reset the file buffer size to 0 after cycling the stderr log file.
MGR: Fix asserts under Linux: wxBG_STYLE_COLOUR does nothing under wxWidgets 3.0, is no longer supported and triggers an assert under wxGTK 3.0.2.
MGR: Work around a wxWidgets 3.0 bug in wxGenericListCtrl (Linux only) which causes headers to be misaligned after horizontal scrolling.
client: preserve min RPC time across client restart.
David's 18 Oct 2013 commit caused the client to clear PROJECT::min_rpc_time on startup.
This causes project-requested RPC delays to be ignored if the client is restarted. This is probably undesirable, so he undid that change.
&#8226; Note: min_rpc_time is used for both
a) project-requested RPC delay.
b) exponential backoff due to server down, not work, etc.

To him it makes sense to clear b) on restart, but not a). Maybe at some point we should separate these.
client: don't show error msg if fail to kill a process that already exited.
MGR: Reduce confusion in the attach wizard by making CWizardAttach the authoritative source for project and account information.
client: always show unparsed tags in config files.
client: parse <dont_suspend_nci> in cc_config.xml

Preliminary Change Log 7.4.38 -> 7.4.42


client: Change Windows 10's kernel detection version to 10.0 instead of 6.4.
client: Check for the correct SKU as well.
client: report error if no start tag in app_config.xml
client: fix bug where project erroneously marked as attached via acct mgr.
client: debug earlier account manager fix.
lib: fix incorrect parsing of <max_ncpus_pct> (0 means no limit)
client: run coprocessor tasks at higher priority even if they use >= 1 CPU.
MGR: Bump copyright statement in About box to 2015.
client: if get 417 HTTP status, switch to HTTP 1.0. 417 (Expectation Failed) probably means we're talking to a 1.0 proxy
SCR: Update to allow building with Xcode 6.1.1 (using Automatic Reference Counting) as well as Xcode 5.0.2.
&#8226; Because Xcode 6.1.1 requires ARC, it builds only the 64-bit screensaver, so the screensaver can't run on OS 10.5.
&#8226; Xcode 5.0.2 builds the screensaver without ARC, and builds a combined 32-bit and 64-bit "universal" binary which runs on OS 10.5 through the latest OS X (OS 10.10.)
SCR: Improve validity testing of settings on Mac Screensaver control panel.
SCR: Remove unused code from Mac Screensaver.
SCR: Fix a bug which prevented default screensaver (clientscr) from running when BOINC client was suspended.
WINBUILD: Change where the installer looks for the MSVC runtime the BOINC binaries are linked against.

http://boinc.berkeley.edu/dev/forum_thread.php?id=8378&postid=60805
 
pfft, you should always compile from source to get the latest nightly dev build.

;)
 
My Q9450 cruncher (Orthanc) has been acting up lately. I had to take it back to stock clocks a few months ago to keep it stable and now it isn't stable even at stock clocks, so I think it is time to retire it. I may throw Linux (probably Mint) on it and just run WEP2, since it might be fine just running that. Not sure about that yet.

Anyway, since I don't want to lose a cruncher, I just ordered parts to build a new i7-4790K based machine to use as my new main desktop and I will move my current main desktop (Barad-Dur) to full time cruncher duty. That gives me 4 more CPU threads and an Intel GPU I can crunch on over the Q9450, once I get everything set up, so that will help my WU throughput.

Once I have the new machine running, I will have the following resources dedicated to BOINC 24x7:

50 Intel CPU threads (all Haswell generation, except one Ivy Bridge)
4 ARM threads
6 Intel GPU's
3 AMD 7970/280X GPU's
2 AMD 7950 GPU's
1 AMD 7870 GPU
3 Nvidia GTX770 GPU's
1 Nvidia GTX780Ti GPU

I also ordered enough RAM to take all of my dedicated crunchers to 16GB each. I kept getting low memory errors on the 8 thread machines, so this should eliminate that issue.
 
Last edited:
DENIS@Home has been added to the All inclusive DC Project List
 
Just hit 1 million points at Numberfields, which makes 18 projects for me with >1 million points. I am currently working towards 1 million at NFS. After that is 1 million at Yoyo, then SRBase and WEP M2. That would give me 22 projects with >1 million points.

I'll never catch Gilthanis (he's currently at 32 projects with >1 million), but dammit I am going to try! :D
 
My next one will be Yafu more than likely...

and have fun when you get to Atlas, it is a data whore. 123numbers is currently RAM starving ..errm... hungry. There was a quad core borg I had with 4GB of RAM that had 2 work units going that brought it pretty much to unusable.
 
Last edited:
123numbers has the highest memory usage I have seen from almost every project so far, except maybe BURP. 123numbers averages about 1.5GB per WU. Since all my machines also have GPU's crunching, I average 6 running CPU threads at any given time. With 6 123numbers WU's running at one time, that means ~9GB of memory in use. I have seen BURP WU's using as much as 8GB. Good thing I now have 16GB of RAM in all my 8 thread machines and 8GB in my 4 thread machines!
 
The virtualbox apps will typically use 1.5-3GB per work unit too. But that is because of the underlying OS requirements alongside the actual apps.

Though I haven't seen it happen in a while, The Lattice project can also get huge work units.

However, I'm more concerned about data limits these days. Projects don't seem as concerned about the amount of data transferred back and forth...
 
Just hit 1 million points at Numberfields, which makes 18 projects for me with >1 million points. I am currently working towards 1 million at NFS. After that is 1 million at Yoyo, then SRBase and WEP M2. That would give me 22 projects with >1 million points.

I'll never catch Gilthanis (he's currently at 32 projects with >1 million), but dammit I am going to try! :D

Make that 33 projects.... :cool:
 
Cheer up... there are only so many... and the virtualbox ones are killing me on data usage...so you still have time.
 
Back
Top