Welcome to MilkyWay@home

Posts by Donald Qualls

21) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58314)
Posted 17 May 2013 by Donald Qualls
Post:

Of course, you and Milky Way have once more contributed to my feelings of computer inadequacy -- my computer took (it says here) several hours to complete that unit, worth 119.9 credits, but another computer assigned the same unit took only 54 seconds, presumably on a GPU of Extraordinary Magnitude.


Actually it was a:
Coprocessors AMD AMD Radeon HD 6900 series (Cayman) (2048MB) driver: 1.4.1741

If you look at your tasks you can filter the valid units, then click on 'workunit' and it will let you see the other pc's that also worked on your same workunit and their results. It is almost as though your pc used the cpu instead of the gpu to crunch the units, here it says:
6,959.80 6,956.10 119.90 MilkyWay@Home v1.01 The first number is total run time, then comes the time the cpu was used, then the credits, and lastly the app used. You can see most of the itme it was your cpu that was crunching not your gpu. the person who finished the same unit in 54 seconds numbers are: 54.73 2.54 119.90 MilkyWay@Home v1.02 (opencl_amd_ati).
While a 3rd persons numbers are:
479.47 35.88 119.90 MilkyWay@Home v0.82 (ati14).

You can see both of the others have much less cpu time by comparison then you do. It COULD be that your gpu is a pretty basic one that just can't keep up with the newer ones, but since it IS contributing it is NOT wasted!


That's where I got the info to feel CPU-inferior... My system is working with CPU only; I recently learned that the new video card I bought for Christmas 2011 is one model too "low" to have the double precision floating point capability required for Milky Way. And by current standards, my CPU is pretty basic, too -- Intel E5200, dual core with, as I recall, limited or no on-chip cache at 2.5 GHz with 2 GB RAM; it's a four year old machine, haven't had money to upgrade or wanted to have to reinstall everything again after upgrading the MB (as happened last time -- remember the good old days when you could install a new motherboard and only have to replace a few chipset drivers?).

Looks like one of the things I should look at if/when I get the promotion I'm up for (I'll know next week, but won't have a paycheck from it until, probably, mid-June) is upgrading my video; I can get a new card now that will crunch with GPU for about what I paid 18 months ago for one that won't (and was still a huge upgrade from the on-board video on my then 2 1/2 year old MB).
22) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58279)
Posted 15 May 2013 by Donald Qualls
Post:
Doesn't BOINC automatically download the latest version of a project's software by itself? All you need is the latest version of BOINC. I don't remember having to manually install software for any project I'm attached to.


Apparently, the version of BOINC I have (MEPIS packages have 6.10.58) will do this only if you don't have a Milky Way module installed. Based on your post, I uninstalled BOINC and Milky Way again, then resinstalled only BOINC Manager and client, not the Milky Way application from the MEPIS 11 repositories -- and bingo, when I attached to the project, it appears to have obtained Milky Way 1.01 32-bit (the one I've been trying to install for two days) on its own and happily started computing. I ran out of time and didn't have it correctly configured for ongoing work, but it completed that work unit and passed validation and I'm now crunching away again.

Of course, you and Milky Way have once more contributed to my feelings of computer inadequacy -- my computer took (it says here) several hours to complete that unit, worth 119.9 credits, but another computer assigned the same unit took only 54 seconds, presumably on a GPU of Extraordinary Magnitude. My dual core, 2.5 GHz CPU (which is a bit less than 10,000 times faster than the first computer I owned back in 1986, a Tandy Color Computer 2 with just under 0.5 MHz clock speed) just can't begin to compete with that kind of machine, and my budget won't stretch to upgrading to a top-line gaming machine (which would be nice to have just for general use performance, never mind BOINC).

Anyway, this appears to be solved -- the correct way to get the current version of Milky Way in MEPIS, or presumably in other stability-driven Linux distros, is to install only BOINC and attach BOINC to the project(s) of your choice, letting the manager obtain and install the project specific clients.
23) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58269)
Posted 15 May 2013 by Donald Qualls
Post:
Well, I found after my last post that all the work units I've crunched since switching to Linux on the 4th have have "validation errors" -- that's why I've gotten no credit.

If I can't get information on how to install a current version of Milky Way in MEPIS 11, I won't be able to generate any more data; I don't expect to return to Windows on a full time basis. Now, I realize that my one dual-core computer with unsupported GPU isn't any big deal against those who can crank out a million credits in three weeks (took me most of a year to get that far), but it seems to me there are a lot more computers in my class being converted to Linux because Windows requires too much money (direct to Microsoft and for hardware upgrades to run the next new version), and if the version of Milky Way included with their packages doesn't work (that's my problem), they might not even realize it.

I've tried the help link, and have yet to see a single English speaking helper available, either in the 7 AM or 6-11 PM EDT time frames. Can anyone tell me how to turn the 1.02 32-bit download linked above into a working project entry in BOINC?
24) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58246)
Posted 13 May 2013 by Donald Qualls
Post:
Okay, tried a little more to get the newer download to work -- tried disconnecting BOINC Manager while doing the copy/rename, and couldn't get it to reconnect, wound up completely uninstalling and reinstalling BOINC and the Milky Way client. I've got it running again, but now I can't find the icon file (minor issue, I can probably make one if I have to), so once again I'm producing work, two units at a time, at around 1:45 to 2 hours per unit -- i.e. back where I was before trying to install the updated client.

I checked for "live help" and found every single English speaking volunteer was either offline, away, or "unknown" status. I guess I'll wait (semi)patiently for more information on how to get the current version of the Linux client (more than a year old, it seems) working in MEPIS 11.
25) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58244)
Posted 13 May 2013 by Donald Qualls
Post:
0.18d was removed from production about 2 years ago.

You will want just the CPU version. This is the 64-bit version.

http://www.arkayn.us/forum/index.php?action=tpmod;dl=item17


Can't say I'm surprised it's that far out of date in the MEPIS repositories. So, I've got the 1.01 32-bit (I'm not running 64-bit Linux -- after trying 64-bit Mint, I went to 32-bit MEPIS for stability) downloaded, but simply replacing the binary I'm currently using doesn't work -- where would I find instructions on how to apply the updated client (preferably specific to MEPIS 11 in terms of folder locations etc.)?
26) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58235)
Posted 13 May 2013 by Donald Qualls
Post:
Okay, my package installer says I have Milky Way 0.18d-1 -- I don't see any correspondence to the downloads you pointed to. I tried downloading the linux_1.02_32 version (no special GPU versions; my GPU lacks double precision floating point capability anyway), but didn't see anything saying how to install it; I moved the binary to the same folder my old version was in, changed the symlink in /var/lib/boinc-client/projects/milkyway.cs.rpi.edu_milkyway to point to the new download, and when I resumed computation, got instant errors -- I've put the symlink back to its original value, so I'm back to generating work, but that clearly was either the wrong file, or I installed it wrong.

I should note that the package maintainers for MEPIS seem extremely conservative in terms of stability; most of the items I can install without learning a bunch more about Linux (like how to compile sources) are Debian old-stable versions packaged for MEPIS -- for instance, the Firefox I have is 9.0.1, while the current download is 20. I'll never have anything "up to date" with MEPIS unless I bypass the package system -- which is generally fine, since I'm more interested in being able to use my computer than in constantly fooling with it -- so, I need more detailed information on how to install the Linux Milky Way client and which one to use.
27) Message boards : Number crunching : Milky Way -- no credits in Linux? (Message 58231)
Posted 12 May 2013 by Donald Qualls
Post:
I've recently jumped from Windows XP to Linux (mostly -- still a couple things I have to go back to Windows to do). No problem, there are BOINC and Milky Way packages for MEPIS 11, and like all the packages on the primary repositories, they install smoother than Windows software. They run correctly, too -- except that the statistics tab in the BOINC Manager swears up and down I haven't gotten a credit in close to two weeks, despite processing two work units (one per core) every two and a half hours (or so) for at least eight days -- my average has dropped from 2000+ before my last Windows trojan incident to below 1000 without any credits recognized in more than a week.

Anyone know if there's a setting I can tweak to correct this? I'm not ego-bound on the credits, but watching the average was my most sensitive measure of overall system health (it was the first warning of all three of the malware incidents I've had in the past six months).
28) Message boards : Number crunching : Why such high GPU requirements? (Message 58148)
Posted 6 May 2013 by Donald Qualls
Post:
Yep, okay, that's the deal -- no double precision floating point capability (in fact, that appears to be the only significant difference between CC 1.2 and 1.3).

Essentially, then, if you want to add GPU computation, at least on MilkyWay@home, you need to spend around $120+ on each video card for PCX, or $100+ for plain PCI (mine was, as I recall, about $70 when I got it, and was still a huge step up from the on-board graphics I had previously). That's not going to happen here any time soon...

BTW, SETI@home still shows on my projects list because that's where I started @home computing, but I haven't run SETI in a while (I had it on my Linux install for a short time, but I've dropped it again, running MilkyWay exclusively now).
29) Message boards : Number crunching : Why such high GPU requirements? (Message 58135)
Posted 5 May 2013 by Donald Qualls
Post:
A year or so ago, I finally managed to upgrade my graphics hardware from the on-board to a PCX video card -- expecting, among other things, to see BOINC start using my GPU for computation.

That's never happened, I just have two work units at a time, one for each CPU core. I've recently discovered that this is because my GPU (on a video card I bought for Christmas 2012) is too old/slow (nVidia GT520, 128 MB DDR2). Despite up to date drivers and appropriate support in the drivers for GPU computing, apparently I don't have enough computing power factor in this card (even though it's many times better than the video hardware I used to have).

My question is, why the lower limit? My GPU should still be capable of several times the performance of each of my CPU cores; even with the overhead of running a Windows XP or MEPIS 11 Linux desktop, it ought to let me produce close to twice as many work units per day as the CPU alone.

Older/budget computers need not apply?
30) Message boards : Number crunching : Get Linux BOIC and Windows BOINC to use same work units? (Message 58097)
Posted 3 May 2013 by Donald Qualls
Post:
[Y]ou want to download workunit onto a Linux machine and then crunch those same units using a Windows machine, or vice versa, although they are in the same box they are two separate machines to Boinc.


I guess, effectively, that's what I'm after, yes -- I'm trying to get as much interoperation as possible so I don't have to pay attention to which OS I leave the machine in when I'm not doing something where it matters.

I'm not a Windows rebel as such, yelling obscenities toward Redmond and making my Linux as different from Windows as possible, I'm just after a secure, friendly system that doesn't require me to spend money on upgrades every couple years, and trying to learn my way around enough to rescue a couple machines (which I can't afford to replace) that are seriously obsolete from the Windows perspective. I can't yet completely abandon Windows, because there are things I want/need to do that don't have Linux versions (or at least not versions compatible with Mepis). All that to say, everything I currently do in Linux I can do in Windows (in fact, my original post was done from the Mepis version of SeaMonkey, while this one is from the Windows version), and I'd like to expand on that theme as far as possible.
31) Message boards : Number crunching : Get Linux BOIC and Windows BOINC to use same work units? (Message 58092)
Posted 3 May 2013 by Donald Qualls
Post:
I use Windows much of the time, but I'm trying to learn Linux and have MEPIS 11 installed under a dual boot. After considerable research, I've gotten my browser and e-mail (SeaMonkey) to share bookmarks, history, mail folders, and so forth between the two systems; I'd like to do the same with Milky Way @ Home. I think I can do it by making the folder where the data files are stored a symbolic link to the Windows data folder, but I don't see where Linux BOINC and Milky Way store their files.

Anyone have a pointer for this?
32) Message boards : News : updating server daemons (Message 54655)
Posted 3 Jun 2012 by Donald Qualls
Post:
I keep finding no tasks, and was just unable to get any work units when I forced an update. I do see this occasionally on weekends (I figure the server has failed in some manner and there's no one there to reset it), but it's sure taking a bite out of my average...
33) Message boards : News : updating server daemons (Message 54563)
Posted 1 Jun 2012 by Donald Qualls
Post:
I'm not getting any work units... :(


Previous 20

©2024 Astroinformatics Group