Wish list...
log in

Advanced search

Message boards : Wish list : Wish list...

Previous · 1 · 2 · 3 · Next
Author Message
Taynak
Send message
Joined: 29 Oct 10
Posts: 1
Credit: 79,610
RAC: 0
Message 10101 - Posted: 30 Oct 2010, 8:41:53 UTC

you know the only thing i would like to see added is a graphic representation of the work being done. seti has it and einstein has it... this one definately needs it.

S@NL - Nick Succorso
Send message
Joined: 31 Oct 10
Posts: 2
Credit: 11,177,816
RAC: 0
Message 10174 - Posted: 5 Nov 2010, 13:26:25 UTC

Hi, I'm running Collatz on a Intel Core 2 Quad 9550 under Windows XP (SP3) and have encountered a couple of problems which are most likely not due to Collatz, but to an annoying intermittent hardware problem of some bug in the system software. Whatever it is, it sometimes causes my system to a complete or almost complete halt.
The problem that comes with it is that it is not always possible to nicely stop BOINC manager and the Collatz client. When I bring my system back up again and everything starts running again, I noticed that in the case when it was not possible to 'gracefully' stop Boinc or Collatz, the units which were processed (SSE and Cuda) are set back significantly. Could it be that this is caused by the fact that during processing the status is not saved at certain intervals?

If this indeed is the case, I would wish for an improvement in the Collatz client that does this, let's say every minute or so. It is so frustrating to see many hours of calculation going down the drain.

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,694,776,710
RAC: 5,362,110
Message 10190 - Posted: 6 Nov 2010, 12:28:21 UTC - in response to Message 10174.

Hi, I'm running Collatz on a Intel Core 2 Quad 9550 under Windows XP (SP3) and have encountered a couple of problems which are most likely not due to Collatz, but to an annoying intermittent hardware problem of some bug in the system software. Whatever it is, it sometimes causes my system to a complete or almost complete halt.
The problem that comes with it is that it is not always possible to nicely stop BOINC manager and the Collatz client. When I bring my system back up again and everything starts running again, I noticed that in the case when it was not possible to 'gracefully' stop Boinc or Collatz, the units which were processed (SSE and Cuda) are set back significantly. Could it be that this is caused by the fact that during processing the status is not saved at certain intervals?

If this indeed is the case, I would wish for an improvement in the Collatz client that does this, let's say every minute or so. It is so frustrating to see many hours of calculation going down the drain.


Calculations freeze for many reasons but one thought is that they freeze whenever the benchmarks are run, which happens weekly. The new Project side software is supposed to fix this but I don't know if it has been implemented yet. Another thing to check is your gpu drivers, the latest and greatest is not always the best! Sometimes the latest IS better though, some experimentation on your part is needed if you think that might be the problem.

To make Boinc do a checkpoint, which is what you are asking for, more often you can change the settings in the Boinc Manager under Advanced, Preferences, the disk and memory use tab and then their is a line about check pointing. HOWEVER I do not believe you can make a workunit checkpoint more often than the project has it set up, so if you set it, like I do, at 15 minutes and the project says 30 minutes then it will NOT checkpoint at 15 minutes. But if the project has it set to 2 minutes you can extend that to 15 minutes for example. The other thing you can do is change the amount of time Boinc crunches one projects workunit before switching to another projects workunits. That is also in the Boinc Manager but under the processor tab this time and then near the bottom. If you set this so a unit finishes before switching to another project you will probably lose less when it crashes. Changing things in the Boinc Manager REQUIRES you to click ok at the bottom to save any changes and affects ONLY that pc, not all your pc's. If you want to affect all of your pc's it is better to do that thru each projects webpage.

The problem with doing a checkpoint every minute is that your hard drive will never slow down and on a laptop that will kill it very quickly. It will also make your pc very slow to do anything else because it is always writing checkpoints! Some checkpoints can be small and some can be HUGE! Each project does their own, so all are different, the ones at Malaria for instance can be 50meg in size!

davecons
Send message
Joined: 2 Nov 10
Posts: 1
Credit: 280,841
RAC: 0
Message 10227 - Posted: 9 Nov 2010, 1:49:41 UTC

I'd like to be able to specify maximum CPU usage by time.

Otherwise, love the software. Awesome!

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10230 - Posted: 9 Nov 2010, 6:39:59 UTC - in response to Message 10227.

I'd like to be able to specify maximum CPU usage by time.

Otherwise, love the software. Awesome!


That is set in the Advanced, Preferences, Processor Usage tab of BOINC Manager since you are nut crunching via a GPU (e.g. you do not have any coprocessors listed for your hosts). Just change the setting to something other than 100%.

Note: BOINC does NOT use processor affinity, so the percentage is a timed percentage (e.g. 75% of a quad core does not mean 3 CPUs, but rather 75% of the time it will use all 4 CPUs)

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,694,776,710
RAC: 5,362,110
Message 10240 - Posted: 9 Nov 2010, 12:47:31 UTC - in response to Message 10230.

I'd like to be able to specify maximum CPU usage by time.

Otherwise, love the software. Awesome!


That is set in the Advanced, Preferences, Processor Usage tab of BOINC Manager since you are nut crunching via a GPU (e.g. you do not have any coprocessors listed for your hosts). Just change the setting to something other than 100%.

Note: BOINC does NOT use processor affinity, so the percentage is a timed percentage (e.g. 75% of a quad core does not mean 3 CPUs, but rather 75% of the time it will use all 4 CPUs)


If you want it to only use 3 of the 4 cpu's for instance you can change that too, under the processor usage tab and then near the bottom. I have a dual core laptop and to keep the peace with my wife, who plays games on it, I have it set to 50% in that box and Boinc only uses 1 cpu to crunch.

S@NL - Nick Succorso
Send message
Joined: 31 Oct 10
Posts: 2
Credit: 11,177,816
RAC: 0
Message 10279 - Posted: 12 Nov 2010, 13:33:07 UTC - in response to Message 10190.

Calculations freeze for many reasons but one thought is that they freeze whenever the benchmarks are run, which happens weekly. The new Project side software is .... (clipped) .... the ones at Malaria for instance can be 50meg in size!


Thanks, Mikey for your extensive and clear reply. Although I am running Boinc already for some time now, I hadn't realised that intermediate saving was something you could actually influence in BOINC manager. I also understand that it something to use very carefully.

Nevertheless, from your reply I get the impression it is possible to do something a bout it and I shall experiment a little with it. The most important thing however is to find this nasty failure in my system, which causes strange behaviour not only in Collatz but also in Einstein!

Profile sir sant
Send message
Joined: 17 Nov 10
Posts: 40
Credit: 2,715,926,035
RAC: 6,164,092
Message 10703 - Posted: 22 Dec 2010, 20:38:29 UTC

Whish i had 10 teraflops of raw computing power at hand...
Ok, what i really wish for is the option to suspend primary gpu only when computer is in use, not all of the gpu's.

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,694,776,710
RAC: 5,362,110
Message 10710 - Posted: 23 Dec 2010, 12:02:42 UTC - in response to Message 10703.

Whish i had 10 teraflops of raw computing power at hand...
Ok, what i really wish for is the option to suspend primary gpu only when computer is in use, not all of the gpu's.


Send a pm to Claggy, he might have a command line way of doing that, although I am not sure it is that refined.

Jeanne
Send message
Joined: 23 Aug 10
Posts: 2
Credit: 11,276
RAC: 0
Message 10796 - Posted: 3 Jan 2011, 0:43:12 UTC

how can i stop collatz work downloading that takes 200hours in less than 10 days,
my dv7 gets too hot running 9 apps at the same time,..
and 186 hours of computing is not possible in only 10 days,..
unlike you geeks, im only in this for fun,..lol

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,694,776,710
RAC: 5,362,110
Message 10797 - Posted: 3 Jan 2011, 12:39:52 UTC - in response to Message 10796.

how can i stop collatz work downloading that takes 200hours in less than 10 days,
my dv7 gets too hot running 9 apps at the same time,..
and 186 hours of computing is not possible in only 10 days,..
unlike you geeks, im only in this for fun,..lol


There are lots of tweaks but if you want fewer units to run at one time just change the setting on the webpage under Your Account, Computing Preferences and this setting.."On multiprocessors, use at most 16 processors" to a lower number. Or the line just below that can be changed, you could change it to 50% and only half of your cpu would be used for crunching. If you want less work you must change the cache settings, they are down a few lines under Network usage. Especially change the line "Maintain enough work for an additional 0.1 days". The lower this number is the fewer days of cached work you will have on your pc. This number is NOT precise, so play around with it a little bit and see what works for you. This WILL change every machine you have, so if you have several machines you could also do this on each machine, so each could be different, but you cannot do it per PROJECT! If you set it here at Collatz it will affect all projects you crunch for, both gpu and cpu.

Oh and we all started crunching for the fun of it, and for most of us it still is fun! We just enjoy passing someone else in the stats too, so we get more and more pc's and have more and more fun!! Oh and 24 hours times 10 days equals 240 hours of crunching time, so no problem finishing 186 hours of work if your pc crunches 24/7. Be careful Boinc can be addictive and FUN!

rjbelans
Avatar
Send message
Joined: 31 Jan 12
Posts: 3
Credit: 67,573,350
RAC: 0
Message 13509 - Posted: 16 Feb 2012, 12:44:31 UTC

I would like to see better GPU utilization % on my cards. The cuda23 app uses about 80%, which isn't that bad, but the cuda31 app only uses about 50%. If there is anything that I can do to up it, please let me know. Thanks.


OS: Win7 Ult. 64
GPU: EVGA 590 Classified x2

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,694,776,710
RAC: 5,362,110
Message 13513 - Posted: 17 Feb 2012, 12:06:23 UTC - in response to Message 13509.

I would like to see better GPU utilization % on my cards. The cuda23 app uses about 80%, which isn't that bad, but the cuda31 app only uses about 50%. If there is anything that I can do to up it, please let me know. Thanks.


OS: Win7 Ult. 64
GPU: EVGA 590 Classified x2


An app_info.xml file might help! It is basically a set of commands tweaking Boinc to run as you want, not generically as most of us do.

Profile michael
Send message
Joined: 12 Dec 09
Posts: 1
Credit: 229,194,712
RAC: 96,132
Message 14755 - Posted: 23 Sep 2012, 6:32:50 UTC

am a PC computer with a Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [Family 6 Model 23 Stepping 10] and [2] CAL ATI Radeon HD 4700/4800 (RV740/RV770) (1024MB) driver: 1.4.1607 for GPU which say they are crossfired and a Microsoft Windows 7 Ultimate x64 Edition, Service Pack 1, (06.01.7601.00)operating system, I wish that... my [2]ATI FireProV8700 (CAL ATI Radeon HD 4700/4800 (RV740/RV770) (1024MB) driver: 1.4.1607 for GPU) GPU's would stop making computer errors.

Profile mikey
Avatar
Send message
Joined: 11 Aug 09
Posts: 3242
Credit: 1,694,776,710
RAC: 5,362,110
Message 14761 - Posted: 23 Sep 2012, 15:14:12 UTC - in response to Message 14755.

am a PC computer with a Intel(R) Core(TM)2 Quad CPU Q9650 @ 3.00GHz [Family 6 Model 23 Stepping 10] and [2] CAL ATI Radeon HD 4700/4800 (RV740/RV770) (1024MB) driver: 1.4.1607 for GPU which say they are crossfired and a Microsoft Windows 7 Ultimate x64 Edition, Service Pack 1, (06.01.7601.00)operating system, I wish that... my [2]ATI FireProV8700 (CAL ATI Radeon HD 4700/4800 (RV740/RV770) (1024MB) driver: 1.4.1607 for GPU) GPU's would stop making computer errors.


I did a search and found this:
"Exit code -1073741819 (0xC0000005): This is the famous "General Access Violation". There a numerous reasons for this error to occur, from hardware problems to graphics drivers and more. Ideally when this happens, the "Windows Runtime Debugger" should start up and write a stack dump to stderr out that helps to further diagnose the problem. If the "Access Violation" is listed in the "*** Dump of the Graphics thread ***", it was almost certainly a problem with the graphics driver. The most common cause for an Access Violation listed in the "*** Dump of the Worker thread ***" shows "houghmap.c:" near the end of the first line of the Callstack. This is a problem that we might be able to do something about, we are currently hunting this. Apparently it only happens on certain machines, it might be related to the data these hosts are processing, but may also be a property of the hardware or other software on the system."

When I look at your driver yours says this:
[2] CAL ATI Radeon HD 4700/4800 (RV740/RV770) (1024MB) driver: 1.4.1607

while mine says this:
AMD ATI Radeon HD 5800 series (Cypress) (1024MB) driver: 1.4.1741.

So try upgrading your drivers and see if it helps. Also using 'cross-fire' does NOT help with crunching, but it DOES help with gaming. So if you are JUST crunching you might consider taking the cable off, if you are gaming leave it on of course, but STOP the crunching so you can game faster.

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 14767 - Posted: 24 Sep 2012, 23:30:16 UTC - in response to Message 14761.

[quoteSo if you are JUST crunching you might consider taking the cable off, if you are gaming leave it on of course, but STOP the crunching so you can game faster.[/quote]

You should also be able to enable or disable it via the Catalyst Control Center if I'm not mistaken.

Grey
Send message
Joined: 6 May 12
Posts: 16
Credit: 1,059,243,544
RAC: 0
Message 15176 - Posted: 14 Oct 2012, 8:41:43 UTC

Where is source code of Collatz?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 15297 - Posted: 23 Oct 2012, 13:45:32 UTC - in response to Message 15176.

Where is source code of Collatz?


int GetSteps(int x) {
steps = 0;
while (x != 1) {
if (x % 2 == 0)
x = x / 2;
else
x = 3 * x + 1;
steps = steps + 1;
//check if we have found a number that may disprove the conjecture
if (steps == MAXINT)
break;}
return steps;


Expand the above to handle 256 bit integers and then add a lookup table of 2^20 elements as described in: http://en.wikipedia.org/wiki/Collatz_conjecture#Optimizations

Yes, I know that isn't what you meant. But, due to an attempt when the project just started up to just duplicate the output without actually doing any of the work (a.k.a. making up results or cheating), I made the decision to NOT make the source code avaiable to the public.

Profile Onur Ulusoy
Send message
Joined: 27 Aug 11
Posts: 2
Credit: 108,882,691
RAC: 328,457
Message 23815 - Posted: 2 Feb 2017, 5:26:24 UTC

(Englisch) Hello. Have been since seti@home. Since then with BOiNC and longer breaks. Once I destroyed a CPU in a beta application. Also times over 20 projects had. My concern is the project distribution with CPU's with several core cores. This should be BOiNC or the respective project, tasks lasting several days of several cores worked out. For older CPUs, e.g. my (AMD Phenom (2) 965 BE), this breaks most orders that last longer than one day with a calculation error. In the case of these long jobs, only one core is assigned. It is sensible to work with the several cores for this and such project tasks. Many had bought a computer with several cores. If you have understood my concern that in longer tasks several nuclei are in any case more intelligent than just one core. Please share this with other projects. Until here. I need to translate this here with Google translate from german to english. I can not share the sun with you. MS Edge, M Firefox does not translate automatically. So I also use the G Chrome. Sincerely, Onur Ulusoy.

(German) Hallo. Bin schon seit seti@home dabei. Seither mit BOiNC und längeren Pausen. Ein mal habe ich einen CPU bei einer Beta Anwendung zerstört. Auch mal über 20 Projekte gehabt. Mein anliegen liegt an der Projektverteilung bei CPU´s mit mehreren Rechenkernen. Dies sollte BOiNC einteilen oder das jeweilige Projekt, Aufgaben die mehrere Tage andauern von mehrerern Kernen abgearbeitet werden. Bei älteren CPU´s z.B. meine (AMD Phenom(2) 965 BE), bricht diese die meisten Aufträge die länger als ein Tag andauern mit einem Berechunungsfehler ab. Bei diesen langen Aufträgen wird auch nur meistens nur ein Kern zugeordnet. Es ist sinnvol bei solchen arbeiten das mehrere Kerne für diese und solche Projekt Aufgaben genutzt werden. Viele habe sich auch damals schon einen Rechner mit meherer Kernen gekauft. Wenn ihr mein anliegen verstanden habt das bei längeren Aufgaben mehrere kerne auf jeden Fall sinvoller sind als nur einen Kern einzuteilen. Bitte teilt diese auch mit den anderen Projekten mit. Bis hier her. Ich muss dieses schreiben hier mit Google translate vom deutschen in das englische übersetzen. Sonnst kann ich mich schlecht mit teilen. MS Edge, M Firefox diese übersetzen nicht gleich automatisch. Also benutze ich auch den G Chrome. Mit freundlichen Grüßen Onur Ulusoy.

Profile step2000
Avatar
Send message
Joined: 1 Aug 13
Posts: 96
Credit: 1,479,325,797
RAC: 1,952,515
Message 23820 - Posted: 3 Feb 2017, 12:41:08 UTC

Multi-thread is difficult on a calc process as the swap between on shared results would in fact add to the time. I think it would be a waste of programming time and also the net gain would be very small. What could happen is on GPU is to split the cores and share those but the net gain would be very very small so 1:1 is the best until multiple processors and multiple GPU are on a system ----say 4 processors (chips) and 2 GPU Cards not SLI'd doing 2 or more EXE processes at same time. Not sure I would like programming that nest of code and placing it into the GUI, oh my!
____________
Retired Business Owner/Developer
Working toward a real solution to why programming takes so many versions to get the end product that just keeps getting better with each version. Do Loop of Products!

Previous · 1 · 2 · 3 · Next
Post to thread

Message boards : Wish list : Wish list...


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.