|
Questions and Answers :
Preferences :
Terms used for CPU/Processor/Core/Thread or the likes ... (usage limits in preferences)in
Message board moderation
Author | Message |
---|---|
Send message Joined: 16 Mar 16 Posts: 6 Credit: 858,545 RAC: 0 |
We are very confused about the terminology used (not only) by CPDN and/or BOINC. To our knowledge we understand, IN GENERAL, the following: 1) A motherboard can have one or more CPUs (physical object/s). 2) A CPU is the so called "Central Processing Unit". 3) A CPU is also called "processor" (the physical thing you buy at the store). 4) A "processor" has at least one "core" (inside the obove mentioned physical object). 5) A "multi-processor" has more than one core (makes i.e. mutliprocessing possible). 6) A "core" can have one or more "threads". 7) A "thread" is a certain logical pathway through the "core". Hope you all, more or less, agree ... If so, then one could be very irritated by the option that can be set in "preferences": "use at most xxx % of the CPUs"? Or, do we all have workstations which mostly have more than one CPU, err sorry, physical processor! This setting can be found at numerous places. Have a nice day, even after reading this! "Not everything that can be counted counts, and not everything that counts can be counted." - Albert Einstein (1879-1955) |
![]() Send message Joined: 7 Aug 04 Posts: 2172 Credit: 64,684,962 RAC: 857 |
Boinc treats any possible thread as a CPU for the purposes of that percentage calculation. So, an i3 CPU on a desktop computer may have 2 cores, both of which have hyperthreading, so boinc sees this as 4 possible threads. Choosing 50% for the "use at most xxx % of the CPUs" will allow just 2 tasks being run at a time on that i3. There may be a better way for boinc to describe this option, but I don't want to write it and any other way of writing it may be just as confusing, although technically more correct. |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
Choosing 50% for the "use at most xxx % of the CPUs" will allow just 2 tasks being run at a time on that i3.Didn't work for me, as have 100% on 4/8 Cores/Threads and took only one. What WAS working is set the * Your account, * cpdnboinc preferences and here * Resource share from 100 to another number, for Example I used 400 and then I got 4 CPUs loaded of 8, so far I can tell that after just 3 Tasks successful, 2 failed - mean had 4 on 8 Threads running. so at the 8 Thread I use 400 to use 50% of all, and at the 4 Thread CPUs I use 200 now and will see if also 50% of is bussy next time. |
![]() Send message Joined: 7 Aug 04 Posts: 2172 Credit: 64,684,962 RAC: 857 |
Choosing 50% for the "use at most xxx % of the CPUs" will allow just 2 tasks being run at a time on that i3.Didn't work for me, as have 100% on 4/8 Cores/Threads and took only one. I'm not sure that things are as you infer. There has only been a couple new batches released since June 23rd, and those batches were misconfigured and failed on all PCs. So any tasks that your PC has picked up since the beginning of July have been tasks that failed, or timed out, on other PCs. These are few and far between and determined by the luck of the draw as to when your boinc requests vs. the infrequent availability of those previously failed tasks. If you have boinc configured to use 50% of your available CPUs, and you have a core i7 with hyperthreading on (4 cores/8 threads), you will get 4 tasks when there are lots of tasks available. |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
I HAD 4 Tasks same time on my 4/8 CPU. Two of them failed after short time, and 2 was still running and got ready. I have seen clear 4 times/different 100% used Threads on Htop. Everyone can decide by own ... and we see what will be with next bigger deliver. I have been also a good analytic for 20 years IT specialist. So I see a clear difference in % of CPU, what cant be more than 100% if not over-tuning, but I focus on the other explained setting (Resource share diff to 100). Maybe as in July and August abut the 1st September we get again, some of the failed Tasks like I had 2 times before. So I keep my computers on from 31. and will press Update after about 5 hours, as I have seen, then the last connect keep stand and got old. Never know what that might mean for get new Tasks also. Wiki describes this Backoff or referrer with a start of min 60 sec, getting longer up to one full day. Never tested that, but sure at my 24/7 server after 5 hours there is now new updadate in last connect anymore. |
Send message Joined: 5 Sep 04 Posts: 7629 Credit: 24,240,330 RAC: 0 |
"Resource share" is between projects, and is a long term matter. It takes several weeks to settle down after a change. It's for the proportion of time that your computer spends on each project. It has nothing to do with the number of processors that you're allowing your computer to use on any given project. That's done by the setting Use at most 100 % of the CPUs which is in in Computing preferences This setting will take effect the next time that your computer contacts the project's server. Large numbers of us are or were computer programmers. |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
Yes, most of us where or are programmers, as I also learned it. But as programs and all are made of human, there will be mistakes. So we have a look forward. And as long still Tasks are processed from so much computers, I am sure there are some basic mistakes in how to collect the computers should do a Task also (dont repeat that we have time - 2 Months of this part at so much power still in work, haha). I hope that mistakes will die out - as human should extinct it will come as I want. So back on topic easy fact: It is an 4/8 Core/Thread CPU i7-4770 (realized as 8 processors) setting was 200% CPU and share 200. So explain me and others here how I could have 4 Threads running, with this settings that both seam (as programmer and logic ???) the reason for my surprise on 2.8.19 with 4 Tasks. And then I see this: Tayloring BOINC settings for cpdn * Leave tasks in memory while suspended? : sure, as keep in 32 GB MEmory will avoid rereading, but that this need also "recalculation" is strange. Explain ? * Suspend work if : as I read only one Thread/Work was possible to get, sure no suspend - there are 7 other processors doing my normal work. * Use at most : as still thought I get 8 units/Tasks I set it to 90% / 90% and wow, my tasks didn't fail also. So here also please logical responce on what might be the problem ? (take much longer or what, check my 3 Tasks done and tell me that they are as fast as possible or not = then we get a point about this "use at most x%") (But anyway, as with 100% cpu and 400% share I can do it to 100% now also - so less work, no problem.) |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
Somewhere in the boinc wiki I found a hint about the communication, telling like slowly longer time to reconnect from low to maximum a day. Cant find it again to link ... * "Communication deferred ...: The client will wait the specified amount of time before attempting to contact the project server again. If no communication is required at that point, none will be attempted." = last point I got again to what I miss now ... So to come with facts I started 4 of my computer last evening: 1st screenshot from 21:53 ITC (14:53 UTC) shows Last contact * pc i7-4770-32 = 14:14 UTC * pc i7-4770-32b = 14:27 UTC * pc i5-3470-9 : 13:55 UTC * pc i3-2100-16 : 14:23 UTC then after the night with check "your computers" at 5:20 ITC or 22:20 UTC there are * pc i7-4770-32 = 14:14 UTC stopped 18:17 UTC with reconnect, or waiting now min 4 hours * pc i7-4770-32b = 14:27 UTC stopped 18:30 UTC * pc i5-3470-9 : 13:55 UTC stopped 18:15 UTC * pc i3-2100-16 : 14:23 UTC stopped 17:36 UTC after 3 or most 4 hours there is no any update anymore. And as a slow rising of reconnect I might believe that after 1 hour for some hours, take 2, 4, 8 up to the written max of 24 hours. (Never tested so long to see if 24 hours are even true, without touching the pcs) Result: When already this description is different from what I can see and get facts about backstop referrer, how I should believe in anything else (told or written here) ??? Now what ? Update by manual .... sometime as now they are shown all with Last contact around 22:27 UTC |
![]() Send message Joined: 15 May 09 Posts: 4444 Credit: 17,929,339 RAC: 29,974 |
I have seen a thread over on the BOINC boards which explains how the time between requests increases when a project has no work. |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
I have seen a thread over on the BOINC boards which explains how the time between requests increases when a project has no work. Nice, whatever I try I dont find it again, as increase should mean a change that can be followed. All I have seen yet is 4 hours wait another 1 hour, and after just nothing anymore. Confirmed 2 times today again. I have to update manual, to keep all 4 pc read to get Tasks - otherwise - so my meaning - they are set as offline, and wont get any Task. |
![]() Send message Joined: 15 May 09 Posts: 4444 Credit: 17,929,339 RAC: 29,974 |
I think this is only a problem with projects like CPDN which often go for periods without any work. Once the work supply gets going again, things should be fine. |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
since before my last post (30 Aug 2019, 15:12:34 UTC) I didn't change any configuration. The settings are: all 100% CPU AND depend on Work (400), home (200) or school (100) the share. The ID 1487938 is a Win7 Home computer setup with 100% CPU and share 200. The computer has 4 CPU and got new tasks on 1.9. ... there are 4 tasks running same time, mean 4 CPU run 100%. So what ... ? mean CPU % and share now ? Real CPU 100% is all CPU or one full CPU al last on my 8 CPU computer with CPU set 100% i got just 4 tasks run same time. |
![]() Send message Joined: 7 Aug 04 Posts: 2172 Credit: 64,684,962 RAC: 857 |
In each case, you were not the first to run tasks from those work units. They had failed on other people's computers. In fact, the 4 that your i7 (8 threads) picked up all failed at the same time on one computer, were marked available by the server, and it just happened to be your computer that requested new work from the server when they became available. Same thing with your i5 (4 threads). You only got 4 on your i7 because only 4 tasks were available at the time boinc requested work. |
Send message Joined: 4 Jul 19 Posts: 31 Credit: 252,192 RAC: 0 |
...You only got 4 on your i7 because only 4 tasks were available at the time boinc requested work. THAT is some new information, that is/was not available for me (or maybe others) to understand this case. hmm ... so wait for some new big one. and anyway, as my solar energy is build in now, all this work will be done during sunshine hours, mean just 1/3 of a day, also to prevent climate change by this tasks energty usage. Will then take 25 days instead of 8 but as said, we are not in hurry. Sure ? |
©2024 climateprediction.net