My Apologies To The Einstein Crunchers

log in

Advanced search

Message boards : Cruncher's Corner : My Apologies To The Einstein Crunchers

1 · 2 · 3 · 4 . . . 9 · Next
Author Message
Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107625 - Posted: 8 Nov 2010, 12:48:15 UTC

Greetings Einstein Crunchers,

Last month, October, I decided to re-attache to Einstein@home since SETI@Home is going through, what I would like to call, a spurt of growth. Many of you know that SETI is in the process of receiving 2 new servers to better serve the project and the thousands of volunteers faithfully crunching SETI WUs.

Not long after I started crunching Einstein on my i7 PC, I noticed some strange anomalous behavior. I would see WUs that were "Waiting for memory". I would also notice that after some time, I would see that text was missing from things like web pages and even things on the i7 such as the Windows start button would no longer say "Start" and tabs in my browser and HTML editor would also be blank. When shutting down a program, it would take a couple minutes for the window to go away. Programs would no longer start. BOINC would no longer make a connection to localhost after restarting it. The i7 re-booted all on its own once. I immediately suspected Einstein.

Now, lets go back in time to May (I believe it was) when I built this i7 PC. I noticed strange things happening with my video display. My mouse cursor would suddenly jump from one point to another across the screen. Programs, like my HTML editor, would auto scroll while I would be doing my coding. When on a website, I would be reading a page and suddenly be transported back a page or 2 as if I hit the back button, which I hadn't. I suspected that there was something flakey about my new PCI-E video card. I lived with it for several months.

When I saw the problems I started having, when starting Einstein again, I figured, even though I still suspected that Einstein was at the root of the current problems, that my video card was getting worse. So, a couple days ago, I bought a new one and installed it. The problems I was seeing, before Einstein, went away. Good call on that one, the "old" card was indeed flakey. But, I was still observing the problems that manifested after starting Einstein again. My suspicion of Einstein was getting greater.

I mentioned these problems in another thread, here, and a few members gave me some suggestions to try to alleviate my problems. A few of them had no affect, but, I did get a definitive explanation for my problem. It was indeed Einstein that was the cause.

But, my i7 was also at fault, sort of. I'm not running a 64 bit OS, so WinXP Pro is not seeing all of my 4GB of RAM. Einstein uses a great deal of RAM, due to the size of the WUs, and perhaps other factors as well. Since my 32 bit OS is only seeing 3GB of RAM and 8 Einstein WUs are taking up 2 of those GBs, the i7 was left with 1GB for everything else including BOINC.

And now for the apology: About 45 minutes ago, give or take, I checked on the i7 and BOINC. You see, after about 8 to 10 hours of continuous running, BOINC would start creating havoc with the i7 and I would have to re-boot the PC. When I checked on the progress of BOINC, I saw that 175 WUs were reporting "Computational error"(s). I had 5 completed WUs waiting to report. Suffice to say, there probably are many of you out there waiting for the results of those 175 WUs, and for that I do apologize.

For the time being, I have resigned myself not to run Einstein. As a matter of fact, BOINC is sitting idle right now with nothing to do. I do plan on upgrading to a 64 bit OS (WinXP Pro, not Win7). I may even decide to double my RAM to 8GB. Maybe then I will start crunching Einstein again. But, right now, I do not want what happened this morning to happen again. It is NOT fair to the rest of you waiting for results so you can get the credit for the work you have already done.

My thanks go out to Mikey, Gundolf and DanNeely for the help they gave me. I really appreciate it! :)

And now, it's time to do my normal routine(s), now that the i7 is free to do my bidding, once again. ;)

Have a great day, everyone! :)

Keep on BOINCing...! :)

____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]

Profile Gundolf Jahn
Send message
Joined: 1 Mar 05
Posts: 1079
Credit: 338,587
RAC: 6
Message 107626 - Posted: 8 Nov 2010, 14:44:04 UTC - in response to Message 107625.

...Since my 32 bit OS is only seeing 3GB of RAM and 8 Einstein WUs are taking up 2 of those GBs, the i7 was left with 1GB for everything else including BOINC...

What about reducing the number of concurrently running BOINC tasks, instead of stopping altogether?

Set "On multiprocessors, use at most __% of the processors" to 50% (25%, 13%) to use only 4 (2, 1) out of your 8 cores. That should reduce the memory burden enough so that you can continue crunching without problems.

And don't forget that running CUDA tasks can also affect the graphic performance up to (seemingly) freezing the screen.

Gruß,
Gundolf
____________
Computer sind nicht alles im Leben. (Kleiner Scherz)
Michael Karlinsky
Avatar
Send message
Joined: 22 Jan 05
Posts: 896
Credit: 18,971,707
RAC: 14,839
Message 107627 - Posted: 8 Nov 2010, 14:53:03 UTC

Hi Siran,

error message is always:

<error_message>signature verification failed</error_message>


Seems sth. is messing with your downloads, firewall, antiv virus. Try to disable/exclude BOINC and try again!?

Michael
____________
Team Linux Users Everywhere
Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107628 - Posted: 8 Nov 2010, 16:40:34 UTC - in response to Message 107626.

-[ snip ]-

What about reducing the number of concurrently running BOINC tasks, instead of stopping altogether?

Set "On multiprocessors, use at most __% of the processors" to 50% (25%, 13%) to use only 4 (2, 1) out of your 8 cores. That should reduce the memory burden enough so that you can continue crunching without problems.

And don't forget that running CUDA tasks can also affect the graphic performance up to (seemingly) freezing the screen.

Gruß,
Gundolf

Greetings Gundolf,

Sorry, but that would defeat the purpose that I built my i7 for to begin with.

My new video card will probably handle CUDA WUs better than the defective one did, but that was not why I got the card. I don't do CUDA, yet.

Another thing that kinda disturbs me is the amount of storage Einstein requires and is currently using. When I had a full compliment of WUs, about 215 or so, Einstein took up almost 2 GB of space. Granted, I have a 500 GB drive, but it's the fact that with all the WUs gone, Einstein is still taking over .5 GB. SETI is taking only 31 MB, Orbit 0(zero) and Virtual Prairie, which I just attached to, 484K. Seems to me, unless there is some underlying logical reason for the over sized WUs, Einstein could reduce the size of them and not put so much stress on users PCs. In my opinion, it's arrogant to assume that EVERYONE has a PC that can handle the work they bulldoze out to us. Or that users will (can) just go out and get what's needed to upgrade their PC to handle the load from Einstein.

When I ran Einstein, in the past, I do not remember having this problem with their WUs. And back then I was running a single core with HT and I doubt even 1 GB of memory.

Einstein will remain on hold, but attached, until I decide on whether I will "upgrade" my i7 to better handle the load or not.

Thanks for the suggestion anyway, Gundolf! :)

Keep on BOINCing...! :)

____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]
Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107629 - Posted: 8 Nov 2010, 17:00:06 UTC - in response to Message 107627.

Hi Siran,

error message is always:

<error_message>signature verification failed</error_message>


Seems sth. is messing with your downloads, firewall, antiv virus. Try to disable/exclude BOINC and try again!?

Michael

Greetings Michael,

Well, at this point in time, I believe that error message to be a moot point. That message was on one of the 175 WUs that whacked on me this morning. Up until that point, I had no problem getting, crunching, uploading and reporting WUs as long as I kept a close eye on the i7 and managed to "refresh" it before something really drastic happen like that which happened this morning. I have never had a conflict problem between BOINC and my firewalls or anti-virus software. And, I have both a hardware and software firewall.

Thanks, though, for bringing this to my attention anyway. I truly doubt I need to deal with that error.

Have a great day, Michael. :)

Keep on BOINCing...! :)

____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]
Profile tullio
Send message
Joined: 22 Jan 05
Posts: 1883
Credit: 738,269
RAC: 796
Message 107630 - Posted: 8 Nov 2010, 17:04:23 UTC

Einstein is taking me 512 MB of disk space. I have two 160 GB disks inside my workstation and an external disk of 1.4 TB used as a backup disk. Storage is rather cheap today. My RAM is 5 GB out of a possible 8 GB. I am using a 32-bit Linux which is pae and could use up to 64 GB RAM. I am running 3 BOINC projects on my Opteron 1210 with two cores (AQUA, Einstein and QMC). Other 4 projects (SETI, QuantumFIRE, CPDN and LHC are not giving me work at this moment, for various reasons). My last CPDN job took 4600 hours before erroring but, due to the trickle method, I did not lose all my credits.
Tullio
____________

Highlander
Send message
Joined: 1 Jul 05
Posts: 21
Credit: 36,453,900
RAC: 15,973
Message 107631 - Posted: 8 Nov 2010, 17:36:37 UTC

Hi Siran!

I can only tell something from my own experience:
- The 860 is known to be running a bit hot, hope u have a sufficient cooling solution on the chip.
- i looked on the asus webpage about your P7P55D-E PRO, i see only a 4 pin power connector for the cpu. For normal use that is no problem, but for crunchers, its nearly on the edge. i take 4 pin boards only for i5 chips, not for i7. i get the i7 only stable with a 8 pin power supply connector.

With the other problems, dont know, can be all or nothing, for my own, i think, top modern hardware needs an OS wich can handle the new technology. On nearly 10 year old OS like XP, its nearly obvious to see some unwanted effects.

Hope, my brainstorming helps you a bit further.

Greetings

Chris
____________

Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107632 - Posted: 8 Nov 2010, 18:39:35 UTC - in response to Message 107631.
Last modified: 8 Nov 2010, 18:45:34 UTC

Hi Siran!

I can only tell something from my own experience:
- The 860 is known to be running a bit hot, hope u have a sufficient cooling solution on the chip.
- i looked on the asus webpage about your P7P55D-E PRO, i see only a 4 pin power connector for the cpu. For normal use that is no problem, but for crunchers, its nearly on the edge. i take 4 pin boards only for i5 chips, not for i7. i get the i7 only stable with a 8 pin power supply connector.

With the other problems, dont know, can be all or nothing, for my own, i think, top modern hardware needs an OS wich can handle the new technology. On nearly 10 year old OS like XP, its nearly obvious to see some unwanted effects.

Hope, my brainstorming helps you a bit further.

Greetings

Chris

Greetings Chris,

Here are my i7 SPECIFICATIONS:
Intel i7-860 2.80 GHz CPU
Corsair H50 CPU cooling system
Asus P7P55D-E Pro MoBo
Galaxy nVidia GeForce GT 430 PCI-E 1MB GDDR3 RAM (CUDA Enabled)
Kingston 4GB PC3 10600 DDR3 RAM (Dual Channel)
Western Digital 500GB 3G SATA HDD x 2 = 1TB Storage
Micro$oft WinXP Pro 32 bit
BOINC v6.10.58

Yes, I know about the 860 running hot. :( That is why I installed the Corsair cooler for it.

According to my motherboard manual, and what I saw physically on the board, if the 8 pin connector you speak of is for ATX power, not to be confused with the 24 pin ATX power connector, then yes it is 8 pin. But, I only have a 4 lead cable/4 pin connector running to it. 4 of the connector's pins, on the board, have a block-off plate covering them. The manual does not specify removing it, just that if the 8 pin is not connected, then the PC will not boot. Well, it boots, no problem. :)

I have a modular PSU. I'm not sure if it has another port for a 4 pin ATX connection or not. If it does, I may have to use the Google machine and see what I can find on the Internet about my motherboard. Maybe someone else asked about the 8 pin connector on the board, and the cover over the 4 pins.

Another thought, I have a newer modular PSU and looked at my unused cables. I do have an 8 pin to 6 pin cable and an 8 pin to 6/8 pin cable. Perhaps the 8 pin connection came out after the PSU was built that is currently powering my i7. Perhaps it is time to get a new PSU for the i7 as well. Scratch those cables! Upon further investigation (looking at the specs on the PSU box), those cables are for PCI Express. It does show that an 8 pin to 8 pin cable is part of the inventory, but it is not in my spares. Perhaps it is being used in my Linux PC that the PSU went into.

So, what you're saying, about my OS, is that it's time I upgrade to Win7 Pro 64 bit? ;) Well, since there are things I do that I cannot do with the Linux PC, I guess the upgrade is inevitable, huh? :)

Thanks for your "brainstorming" Chris! Gave me more to think about. :)

Keep on BOINCing...! :)

[edit]
edited for clarity...
[/edit]
____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]
archae86
Send message
Joined: 6 Dec 05
Posts: 1499
Credit: 275,326,449
RAC: 480,604
Message 107634 - Posted: 8 Nov 2010, 20:17:41 UTC - in response to Message 107632.

then yes it is 8 pin. But, I only have a 4 lead cable/4 pin connector running to it. 4 of the connector's pins, on the board, have a block-off plate covering them. The manual does not specify removing it, just that if the 8 pin is not connected, then the PC will not boot. Well, it boots, no problem. :)

That 4+4 pin power connector has parallel connections in the two halves. So populating both halves reduces resistance in the power input to the board, which lowers heat (power dissipated) at the connection and improves onboard supply voltage and reduces variation. As with the "optional" portion of the wide ATX power connector, which also has extra power connections in parallel with the main part, it is not surprising to have the system run with it not populated, but you are better off hooking it up if you can.

My i7 motherboard had stickers over both of these optional sections--I just took them off as I inserted the proper inputs from my power supply.

____________
Highlander
Send message
Joined: 1 Jul 05
Posts: 21
Credit: 36,453,900
RAC: 15,973
Message 107635 - Posted: 8 Nov 2010, 21:22:57 UTC - in response to Message 107632.


Greetings Chris,

Here are my i7 SPECIFICATIONS:
Intel i7-860 2.80 GHz CPU
Corsair H50 CPU cooling system
Asus P7P55D-E Pro MoBo
Galaxy nVidia GeForce GT 430 PCI-E 1MB GDDR3 RAM (CUDA Enabled)
Kingston 4GB PC3 10600 DDR3 RAM (Dual Channel)
Western Digital 500GB 3G SATA HDD x 2 = 1TB Storage
Micro$oft WinXP Pro 32 bit
BOINC v6.10.58

Yes, I know about the 860 running hot. :( That is why I installed the Corsair cooler for it.

According to my motherboard manual, and what I saw physically on the board, if the 8 pin connector you speak of is for ATX power, not to be confused with the 24 pin ATX power connector, then yes it is 8 pin. But, I only have a 4 lead cable/4 pin connector running to it. 4 of the connector's pins, on the board, have a block-off plate covering them. The manual does not specify removing it, just that if the 8 pin is not connected, then the PC will not boot. Well, it boots, no problem. :)


Yes, sry, was my fault, was time to clean up my TFT, but for being sure, i DLed also the manual :-)


I have a modular PSU. I'm not sure if it has another port for a 4 pin ATX connection or not. If it does, I may have to use the Google machine and see what I can find on the Internet about my motherboard. Maybe someone else asked about the 8 pin connector on the board, and the cover over the 4 pins.


like archae86 said before, try to populate them, its better.


Another thought, I have a newer modular PSU and looked at my unused cables. I do have an 8 pin to 6 pin cable and an 8 pin to 6/8 pin cable. Perhaps the 8 pin connection came out after the PSU was built that is currently powering my i7. Perhaps it is time to get a new PSU for the i7 as well. Scratch those cables! Upon further investigation (looking at the specs on the PSU box), those cables are for PCI Express. It does show that an 8 pin to 8 pin cable is part of the inventory, but it is not in my spares. Perhaps it is being used in my Linux PC that the PSU went into.

So, what you're saying, about my OS, is that it's time I upgrade to Win7 Pro 64 bit? ;) Well, since there are things I do that I cannot do with the Linux PC, I guess the upgrade is inevitable, huh? :)

Thanks for your "brainstorming" Chris! Gave me more to think about. :)

Keep on BOINCing...! :)

[edit]
edited for clarity...
[/edit]


I dont see your PSU on your Component-list; so following only guessing:
If you take your equipment and let the power consumption calculate on http://support.asus.com/PowerSupplyCalculator/PSCalculator.aspx, a minimum of 550 W is required. Imho some safety adding, you need to have something over 600 W.

For the OS: I can only say, W7 64 bit is really nice to have, but main point are the 64 bit. Even with XP. The memory border with 4 GB is not really up to date with todays performance monsters.

Hope my english bubbling make some sense :-)

Chris

____________
ExtraTerrestrial Apes
Avatar
Send message
Joined: 10 Nov 04
Posts: 678
Credit: 38,778,944
RAC: 4,488
Message 107636 - Posted: 8 Nov 2010, 22:19:17 UTC

You're definitely seeing paging / swapping to disk. It's normal that this brings any HDD-equipped system to a crawl.

Sorry, but that would defeat the purpose that I built my i7 for to begin with.


There's a far easier solution than restricting the amount of concurrent threads: restrict the amount of main memory used by BOINC to ~33% (=1GB) or whatever number you like (is probably 50% now) and attach to at least one other project which uses less memory. Personally I like QMC: computationally intensive, stable work supply, good credits, interesting science and German ;)

You may need to adjust the ressource share, so that BOINC doesn't insist on trying to run 8 Einsteins and keeping 4 of them "waiting for memory" forever. Not sure, though.

MrS
____________
Scanning for our furry friends since Jan 2002
Profile Mike Hewson
Volunteer moderator
Avatar
Send message
Joined: 1 Dec 05
Posts: 4638
Credit: 39,785,552
RAC: 12,349
Message 107637 - Posted: 8 Nov 2010, 22:19:49 UTC - in response to Message 107632.
Last modified: 8 Nov 2010, 23:27:32 UTC

Here are my i7 SPECIFICATIONS:
Intel i7-860 2.80 GHz CPU
Corsair H50 CPU cooling system
Asus P7P55D-E Pro MoBo
Galaxy nVidia GeForce GT 430 PCI-E 1MB GDDR3 RAM (CUDA Enabled)
Kingston 4GB PC3 10600 DDR3 RAM (Dual Channel)
Western Digital 500GB 3G SATA HDD x 2 = 1TB Storage
Micro$oft WinXP Pro 32 bit
BOINC v6.10.58

On the face of it with this spec, I'd have to say you ought have no trouble with E@H! It would seem that something about our WU's has caused a belly up though. The unit listing for your rig does not indicate a preference for any particular type of WU ( GW/ABP/CUDA) when it comes to errors. Now if one looks ( bottom of page ) at the stderr output for these units in error you'll see something like this ( some variations thereof ) :

<core_client_version>6.10.58</core_client_version> <![CDATA[ <message> app_version download error: couldn't get input files: <file_xfer_error> <file_name>einstein_S5GC1_3.02_windows_intelx86__S5GCESSE2.exe</file_name> <error_code>-120</error_code> <error_message>signature verification failed</error_message> </file_xfer_error> </message> ]]>

These go back right to the oldest error in you listing. This message pretty well hints that, whatever else is going on, BOINC can't find what it thinks ought be there come time to fire up a WU. Now also most of those WU's in error don't run for long ( or at all ) before that mode of exit occurs. However what I really don't understand is why a few of these units ( say this one ) go for 9K, 21K etc seconds before falling over if indeed the executable is missing - as stderr suggests.

So now I suggest we think of why can't BOINC find what it wants, and when it wants it. First thing I'd do is look in the E@H part of the BOINC directories and see if einstein_S5GC1_3.02_windows_intelx86__S5GCESSE2.exe or einsteinbinary_ABP2_3.11_windows_intelx86__ABP2cuda23.exe are there at all - heck, search the whole machine for that matter. So if it is present, and where it ought to be ( BOINC has per project repositories on your drive ) then we need to think why can't it access? That is, what does signature verification failed mean etc .... :-)

Cheers, Mike.

( edit ) Another thought. Are any of your disks in a RAID configuration ( even RAID 0 )? For WD, it's either EADS or EARS ( I can't remember which ), that don't 'hold' very long in RAID arrays ( spin-up lag? or somesuch ) and so drop out of concurrency forcing a rebuild or the RAID becomes invalid ....

( edit ) .... and since 'RAID' 0 doesn't actually have any redundancy then you can be stuffed in such circumstances ie. you were better off not linking the drives in an array fashion at all .
____________
"I have made this letter longer than usual, because I lack the time to make it short." - Blaise Pascal
Profile Ageless
Avatar
Send message
Joined: 26 Jan 05
Posts: 2974
Credit: 5,374,792
RAC: 0
Message 107638 - Posted: 8 Nov 2010, 22:34:53 UTC - in response to Message 107628.
Last modified: 8 Nov 2010, 22:36:35 UTC

Sorry, but that would defeat the purpose that I built my i7 for to begin with.

Using an i7 with a 32bit OS is blasphemy. You expect to run with 8 threads and then you physically hamper those 8 threads by telling them they're only able to use 3 GB of memory total. You defeated the purpose of the i7 by adding XP, Rick. Go 7, go 64bit Ultimate. It's cool. :P
____________
Jord
Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107643 - Posted: 8 Nov 2010, 23:34:15 UTC

Greetings everyone,

Wow! So much information and suggestions. Not sure where to start so...

@ archae86: I went to the basement and dug out the box that my modular PSU and its cables came in and rifled through the cables. I found the 8 pin ATX cable and replaced the 4 pin with it. I'm not sure if it was my imagination or what, but the PC seemed to boot faster. Probably imagination... ;) Thanks! :)

@ Chris (Highlander): I went to the Asus website and entered my information. I even added one or 2 extra items just for GP (general principle). The results were a recommendation 100W less than my PSU is rated at. Recommendation: 450W PSU: 550W So, good to go there... Thanks! :)

@ MrS (ETA): I have BOINC set to 50% when in use and 90% when idle. I lowered the 'when in use' down from 75% when this all started. I will wait until I can upgrade to full 64 bit (OS and RAM) before resuming with Einstein. Thanks! :)

@ Mike Hewson: At this point the errors are moot. I'm not going to worry about them. I'm just sorry that the 175 error-ed WUs will cause other users to be delayed in getting credit for doing their work on them. My drives are not set in a RAID configuration. Winblows XP just would not see RAID even with the RAID driver(s) pre-loaded. Thanks! :)

@ Jord (Ageless): You never cease to make me laugh my friend. :D Blasphemy, huh? Yeah, I suppose so. I will go 64 bit when I get my tax return next year. It's only a few months, I believe I can handle 32 bit for the time being. ;) Thanks! :)

Well, it's almost time for dinner and TV, so I better get going. I'll check back in the morning.

Thanks again, everyone! :)

Keep on BOINCing...! :)

____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]

Profile Rob Windgassen
Avatar
Send message
Joined: 16 Aug 10
Posts: 5
Credit: 25,368,155
RAC: 31,053
Message 107644 - Posted: 9 Nov 2010, 0:36:05 UTC - in response to Message 107643.
Last modified: 9 Nov 2010, 0:40:28 UTC

At this point the errors are moot. I'm not going to worry about them.


In general it is good to understand what the root cause of the problem is. Wouldn't it be a pity if you upgrade your system and then it turns out that the problem isn't related to the upgraded hardware?

I'm just sorry that the 175 error-ed WUs will cause other users to be delayed in getting credit for doing their work on them.


Don't worry, as your WUs failed early, most or all of them are reassigned and others are happy crunching on them so your wingmen are not really affected badly. [edit] After all we're all just volunteers.

Rob
mikey
Avatar
Send message
Joined: 22 Jan 05
Posts: 2377
Credit: 148,837,930
RAC: 137,255
Message 107661 - Posted: 9 Nov 2010, 12:22:43 UTC - in response to Message 107628.

Another thing that kinda disturbs me is the amount of storage Einstein requires and is currently using. When I had a full compliment of WUs, about 215 or so, Einstein took up almost 2 GB of space. Granted, I have a 500 GB drive, but it's the fact that with all the WUs gone, Einstein is still taking over .5 GB. SETI is taking only 31 MB, Orbit 0(zero) and Virtual Prairie, which I just attached to, 484K. Seems to me, unless there is some underlying logical reason for the over sized WUs, Einstein could reduce the size of them and not put so much stress on users PCs. In my opinion, it's arrogant to assume that EVERYONE has a PC that can handle the work they bulldoze out to us. Or that users will (can) just go out and get what's needed to upgrade their PC to handle the load from Einstein.
Keep on BOINCing...! :)


I think Einstein downloads a huge file full of stuff and then references it for alot for the workunits, so instead of you and I downloading 30 files over 30 connections to the Einstein Server, it sends us one big file that we reference and then we send it back and get another big file full of stuff. I used words like 'stuff' because I do not technically know what is in the big files, but it was discussed elsewhere.
induktio
Send message
Joined: 1 Oct 10
Posts: 15
Credit: 10,144,774
RAC: 0
Message 107663 - Posted: 9 Nov 2010, 12:36:34 UTC - in response to Message 107628.


Another thing that kinda disturbs me is the amount of storage Einstein requires and is currently using. When I had a full compliment of WUs, about 215 or so, Einstein took up almost 2 GB of space. Granted, I have a 500 GB drive, but it's the fact that with all the WUs gone, Einstein is still taking over .5 GB. SETI is taking only 31 MB, Orbit 0(zero) and Virtual Prairie, which I just attached to, 484K. Seems to me, unless there is some underlying logical reason for the over sized WUs, Einstein could reduce the size of them and not put so much stress on users PCs. In my opinion, it's arrogant to assume that EVERYONE has a PC that can handle the work they bulldoze out to us. Or that users will (can) just go out and get what's needed to upgrade their PC to handle the load from Einstein.


I really don't understand why people can complain of the disk usage since hard drive space is almost free nowadays. Usually computations should be optimized to minimize runtime, not memory usage. Spent computation time cannot be later 'recovered' but disk space can be regained afterwards.

Maybe it's the fact I'm running Boinc on dedicated machines, but it seems Boinc projects just will not take advantage of the all the available memory. Currently the machines are running Linux and have 4-8 GB ram. It seems E@H will only use at most 256 MB per core, no matter what. If there was an option to double memory/disk usage and reduce runtime by 10%, I would definitely use it.
Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107664 - Posted: 9 Nov 2010, 14:08:31 UTC

Greetings all,

I have now seen something new concerning my i7, something I hadn't seen before or during the current problem. But before I get into that, let me just say...

Somewhere in all this I stated that I have 2 500 GB HDDs on my i7. I really don't care about the size of the WUs as it would pertain to my drives and have no real adverse affect on them. The point I was attempting to make, and what disturbed me, was the assumption that EVERYONE has a PC that can handle the load from Einstein.

The WUs are roughly 250 MB in size. I was told that with my system setup, running 4 cores with HT and 3 recognized GB of RAM, 8 Einstein WUs are taking up 2 of the 3 GB of RAM, leaving 1 GB for everything else. Perhaps my swap file is not large enough to handle the task, which is what I am about to get into that I discovered this morning...

I am currently trying to run Virtual Prairie. I finally got a boat load of WUs last night while I was getting my beauty sleep.;) I noticed a few symptoms had reappeared. I cleared everything up, including shutting down BOINC. No applications were running and the OS seemed to be functioning without any problems.

I ran a-squared, an anti-malware scanner, and all it found were 9 bad cookies. I deleted them and shut down a-squared. Everything appeared to be running ok. I tried to start BOINC and once again, it could not make a connection to localhost. All I had was the tiny window that stated that BOINC was communicating with the client, or whatever it says. I shut down BOINC, which took a bit of time, less than a minute, when a balloon popped up on my task bar saying something about my virtual memory was not sufficient for the operation or some such thing and that it was being re-sized.

My conclusion, in all this, is that, once again, my Windoze installation has become corrupted. I had to do at least 1 repair install several months ago. And, the funny thing about this is the fact that there were no problems with the i7 or Windoze until after I started running Einstein again. Is it any wonder that I have been suspecting Einstein ever since this problem started?

I believe that what I need to do, now, is set VP to NNT and let the cache run dry. Keep a close, watchful eye on the i7 and BOINC and since I have noticed that the crap starts happening after about 8 to 10 hours of operation, shut down BOINC and re-boot about every 6 hours. Then, do a repair install, AGAIN, and hope for the best. In the wisdom of those reading this thread, would that be a wise course of action?

BTW, I am posting this using my Linux box again, attempting to alleviate any stress on Windoze on the i7. :)

I await any recommendations... :)

Keep on BOINCing...! :) (I am certainly trying to!)

____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]

Richard Haselgrove
Send message
Joined: 10 Dec 05
Posts: 1660
Credit: 54,971,504
RAC: 43,514
Message 107665 - Posted: 9 Nov 2010, 14:19:13 UTC - in response to Message 107664.

...
My conclusion, in all this, is that, once again, my Windoze installation has become corrupted. I had to do at least 1 repair install several months ago. And, the funny thing about this is the fact that there were no problems with the i7 or Windoze until after I started running Einstein again. Is it any wonder that I have been suspecting Einstein ever since this problem started?
...

You might like to look at this comment, and Bernd's response. My feeling is that most of the symptoms you're seeing, including the problem the BOINC Manager has communicating with the client, are due to this start-up disk access for Einsten tasks. And the disk access demands of your 8 i7 cores at startup will be even more severe than the demands of my 8 E5320 cores.

If you can bear to walk away from your host during the start-up phase (and find some way of not starting Einstein on all 8 cores at once), I think you'll find that Einstein's steady-state resource usage is more reasonable when you come back.
Profile Siran d'Vel'nahr
Avatar
Send message
Joined: 15 Feb 05
Posts: 105
Credit: 310,937
RAC: 0
Message 107667 - Posted: 9 Nov 2010, 15:12:41 UTC - in response to Message 107665.

-[ snip ]-

You might like to look at this comment, and Bernd's response. My feeling is that most of the symptoms you're seeing, including the problem the BOINC Manager has communicating with the client, are due to this start-up disk access for Einsten tasks. And the disk access demands of your 8 i7 cores at startup will be even more severe than the demands of my 8 E5320 cores.

If you can bear to walk away from your host during the start-up phase (and find some way of not starting Einstein on all 8 cores at once), I think you'll find that Einstein's steady-state resource usage is more reasonable when you come back.

Greetings Richard,

Ok, that makes sense. I have Einstein on hold right now and am running VP. I just re-booted to get BOINC communicating with the client again. I knocked down the number or cores used by 4, in BOINC preferences.

After all that, I fired up task manager to observe the performance graphs. The page file usage was over 5 GB and my available physical memory was at about 4 MB, this with only 4 VP WUs running! After a few minutes everything mellowed out to just over 2 GB page file usage and 1.6 GB available physical memory.

Ok, now I'm really confused! I just looked at my VP WUs and they're barely over 100 bytes each! And still, a great deal of physical memory is being used. And not only that, I'm still getting "Waiting for memory" messages!

Now I'm really baffled! I have never had this much problem with BOINC before, not even on the past, less powerful PCs I've built. At least not that I can remember. I have 2 WUs running and 2 "Waiting for memory". I still show 1.6 GB available physical memory. Why would a 100 byte WU need to be waiting for memory!?

Keep on BOINCing...! :) (I'm trying already!!!)

____________
CAPT Siran d'Vel'nahr XO
USS Vre'kasht NCC-33187

Siran's website: [ ONLINE! ]
1 · 2 · 3 · 4 . . . 9 · Next

Message boards : Cruncher's Corner : My Apologies To The Einstein Crunchers


Home · Your account · Message boards

This material is based upon work supported by the National Science Foundation (NSF) under Grants PHY-1104902, PHY-1104617 and PHY-1105572 and by the Max Planck Gesellschaft (MPG). Any opinions, findings, and conclusions or recommendations expressed in this material are those of the investigators and do not necessarily reflect the views of the NSF or the MPG.

Copyright © 2016 Bruce Allen