PrimeGrid
Please visit donation page to help the project cover running costs for this month

Toggle Menu

Join PrimeGrid

Returning Participants

Community

Leader Boards

Results

Other

drummers-lowrise
1) Message boards : Project Staging Area : 2017 PRPNet May-June Challenge (Message 108230)
Posted 874 days ago by Profile kashi
Thank you for the explanation.

By "recently" for replacement new memory I meant a few months ago. Same computer with new memory installed completed oodles of BOINC multithreaded PSP LLR tasks without any errors running non stop for 15 days in the recent Mathematics Awareness Month Challenge. It did have H/T disabled then however and was running single PSP LLR tasks multithreaded on 3 cores.

For this PRPNet Challenge, H/T was enabled and ran only 2 PRPNet clients without multithreading, another 2 cores were running POGS in a VMware MacOS virtual machine and another core was allocated to supporting Einstein@Home GPU processing in Windows BOINC. Perhaps the combination was too much for PRPNet to handle, although only 5 cores out of 8 were loaded and the computer was not running hot at all. Or maybe PRPNet just don't take kindly to hyperthreading.

Can't continue running PRPNet on this computer now that it has proved to give errors. With no stop errors shown and no wingman check like in BOINC there is no way for me to to tell if all PRPNet tasks on this computer are processing incorrectly.

Plus my other computer has started having repeated delays today connecting to server to download new tasks "connect to socket failed with error 10060" so looks like something may be awry with PRPNet on that computer too. So I've now stopped PRPNet on that one also.

Have enjoyed participating in a few PRPNet Challenges and thank everyone concerned with running them.
2) Message boards : Project Staging Area : 2017 PRPNet May-June Challenge (Message 108222)
Posted 874 days ago by Profile kashi
Ah shucks, would have been my first prime as finder, found a list which indicated my other 2 from 2010 were as checker. Well I better keep looking for one then, but perhaps not with PRPNet. Thanks for checking it for me and replying.

Wonder what what would cause such hardware errors? Haven't noticed any stop errors in client windows or errors on any other BOINC projects. Computer is not overclocked and I recently fitted new memory, replacing some faulty 3,200 MHz rated memory that started giving errors on many different BOINC projects including an earlier PRPNet Challenge.

New RAM is on motherboard QVL list and running at default 2,133Mhz speed with relaxed timings as it refuses to run at its rated 2,666 MHz speed and tight timings. It's called HyperX Savage, looks like it's a bit too savage for PRPNet. Getting any memory to run at its rated speed and timings has proved impossible on this Skylake CPU.

Why I'm asking, is if this computer is giving errors on PRPNet but not on other BOINC projects then it would be better to avoid PRPNet on this computer and only use my other one for PRPNet. Funny thing is my other computer was giving a few errors when running multithreaded tasks in the recent Challenge on PSP. And that computer uses registered memory.
3) Message boards : Project Staging Area : 2017 PRPNet May-June Challenge (Message 108214)
Posted 874 days ago by Profile kashi
Couldn't get multithreading to work, so just ran some single core clients instead.

A few hours ago was checking one of my clients and it flashed up highlighted that it had found a prime but when I go to http://prpnet.primegrid.com:12006/user_primes.html page it says it is "hidden".

How do I unhide it please?

Done mostly Sieve on PrimeGrid BOINC and PSA subprojects so the only other primes I found were in 2010. I'm pleased to find another one.
4) Message boards : Sieving : ppsieve ATI/OpenCL testing (Message 35674)
Posted 3101 days ago by Profile kashi
Working well without invalids on both cores of my 5970 with Catalyst 11.4 and SDK 2.4. Takes about 36 minutes per task with core/memory @ 770/500.

Uses a full CPU core for each task but looks like it does that for everyone with multiple ATI GPUs. Default 0.79 avg_ncpus is too low with multiple ATI GPUs and causes overcommitment of CPU unless BOINC is configured to use less than all CPU cores. One day I will increase avg_ncpus value to 1 in an app_info.xml to work around this. For now it's working and that's good.

<core_client_version>6.12.22</core_client_version>
<![CDATA[
<stderr_txt>
Sieve started: 59664908000000000 <= p < 59664914000000000
Thread 0 starting
Detected 320 multiprocessors (1600 SPUs) on device 0.

Thread 0 completed
Sieve complete: 59664908000000000 <= p < 59664914000000000
count=155329812,sum=0x28f8861ccf35c560
Elapsed time: 2127.25 sec. (1.46 init + 2125.80 sieve) at 2822569 p/sec.
Processor time: 2149.07 sec. (1.47 init + 2147.60 sieve) at 2793912 p/sec.
Average processor utilization: 1.01 (init), 1.01 (sieve)
13:07:59 (1944): called boinc_finish
5) Message boards : Number crunching : ATI GPU workunits error (Message 29326)
Posted 3233 days ago by Profile kashi
Thank you for your reply.

Yes I discovered the hard way HD 5970 is not able to process OpenCL applications without giving errors on the 2nd core when using Stream 2.2 SDK.

Was just hoping they may have fixed it with Stream 2.3 SDK, not much chance I suppose. As it may shortly be replaced by HD 6990, perhaps HD 5970 may never be supported by OpenCL.

Also as I hadn't seen any warnings for HD 5970 owners about participating in the Winter Solstice Challenge my post had a double purpose.

To be more specific, successfully completing and reporting a task on the 2nd core of a HD 5970 does not show that that any driver (or app_info.xml, if used) issues are solved because the task will go into pending and only later will the status change to "Completed, marked as invalid".
6) Message boards : Number crunching : ATI GPU workunits error (Message 29318)
Posted 3233 days ago by Profile kashi
10.12 APP driver includes Stream 2.3 SDK? Anyone know if Stream 2.3 SDK suppports OpenCL on both cores of HD 5970?

I can't find any info on it. I wouldn't mind joining the Challenge but all tasks are invalid on the second core of my HD 5970 when using 10.11 APP driver with Stream 2.2 SDK. Tasks take about 4 minutes longer on the second core, report successfully and go into pending state but status changes to "Completed, marked as invalid" after they are checked by a wingman.

I thought OpenCL didn't support HD 5970 but saw in a thread here where someone was using one so gave it a go. Didn't notice until half a day later after wingmen starting checking pending tasks from 2nd core that all tasks completed on 2nd core were failing to validate.
7) Message boards : Number crunching : The Full Moon Challenge (Message 25201)
Posted 3372 days ago by Profile kashi
Yes all 321 Sieve tasks I have completed since the end of the Challenge are awaiting validation. Just need 26 more and will get a bronze badge when the validator starts working properly again.

I once asked how long I should wait before enquiring about fully completed work units that remained unvalidated and was told tasks are automatically rechecked once the original deadline expires. So no worries, still 5 days to go before the first ones I completed reach that.
8) Message boards : AP26 - AP27 Search : AP26 on ATI Radeon (Message 21977)
Posted 3502 days ago by Profile kashi
Yes after I read your previous post I thought of an app_info.xml to cater for the other PrimeGrid subprojects. However I then realised that it would be a quite a complicated one having to include so many different subprojects. If any of the subprojects changed or updated their application then it would quickly become out of date.

It is possible to make one I suppose for one's own usage, but if someone who is unfamiliar with how to install an app_info.xml misconfigures something then all the tasks would still be deleted. So the use of an all inclusive app_info.xml may provide false assurance.

Probably better and safer for most testers to follow the warning advice given by enderak and ensure that all tasks for all PrimeGrid subprojects are completed and the cache is empty before giving the BOINC AP26 ATI application a try.

I sometimes still muck up app_info.xml files even after using them for a while, so I almost always empty my cache first before using one. My cache is currently set to 0.2 days and I rarely use a large cache with any project so this is not a problem for me.

I'm probably not the best person to be giving advice on how to do these type of things but I can't see people struggling with something with nobody helping them for some time without at least trying to give some assistance. It is necessary to balance the level of detail given so as not to overwhelm. Also I may forget to mention some things that would not be known by someone new to a procedure because I have used it for a fair while and now take those things for granted.

Some people with greater knowledge and experience than me who would be able to help are sometimes reluctant to do so because either they do not have enough time to explain in detail or they have received ungrateful feedback in the past after the person they were trying to help had some negative experience.

For me I try not to get frustrated when something crashes that I have misconfigured or where I have followed advice given in good faith that may be incomplete or incorrect. Although I am always careful so as not to upset the running of the projects, I feel that these types of mistakes are helpful to learn something. With app_info.xml files particularly, most people who have used them extensively would have unintentionally trashed some tasks at least once, it's kind of like an initiation. :) It's quite a shock the first time it happens and any cached tasks just disappear into the ether like some kind of magic vanishing trick. It is understandable that this shock may create frustration, dismay and even anger.
9) Message boards : AP26 - AP27 Search : AP26 on ATI Radeon (Message 21966)
Posted 3502 days ago by Profile kashi
That was unfortunate. Any unfinished tasks for applications that are not in an app_info.xml will be deleted when BOINC restarts. This always happens when using anonymous platform. I did say to download an AP26 task meaning only one however I did not repeat the warning given earlier in the thread by enderak and perhaps I should have. I was only trying to help some people who posted that they were having difficulty getting the BOINC AP26 ATI version to run so that they could help to test it.

My apologies for not repeating the warning given earlier. I have only ever run one PrimeGrid subproject at a time so the thought of trashing work for other subprojects was not in my mind. I understand the frustration and disappointment this can cause, I have had the same thing happen to me and lost a lot of partly completed work the first time I used an app_info.xml file when helping to test beta applications at Einstein@home and it has also happened a few times since at other projects when I have misconfigured something.

It is the nature of helping to test new applications that you have to be prepared for tasks to crash sometimes or cause other problems. Without people prepared to test though, development of new applications or improvements would be delayed or not happen at all.
10) Message boards : AP26 - AP27 Search : AP26 on ATI Radeon (Message 21949)
Posted 3503 days ago by Profile kashi
I am definitely using the later version of the app. I had done this before, but trying again, here is exactly what I did:

With the AP26 subproject enabled in my preferences, as well as the reisel seive, I shut down Boinc and then copied the four files into "C:\ProgramData\BOINC\projects\" from the zip folder. Then I restarted Boinc, opened app_info.xml, and made the three changes suggested(<max_ncpus>0.2/max_ncpus> to <max_ncpus>0.2</max_ncpus>, then changing the 0.2 to 1.0 on the lines <avg_ncpus>1.0</avg_ncpus> and <max_ncpus>1.0</max_ncpus>)

The message in the stderr.xml is exactly the same when I launch the AP26.exe.


Firstly you need to actually download an AP26 task before you do anything else, not just select it in your preferences.

Secondly any changes you make to the app_info.xml file will not take effect until BOINC is restarted. The app_info.xml file is read by BOINC when BOINC starts. So you neeed to edit the app_info file before you restart BOINC.
You would not be aware of that if you have not used an anonymous platform application before and I did not make it clear in my earlier post.

Thirdly you place the 4 files into the PrimeGrid folder located at "C:\ProgramData\BOINC\projects\" not into the "projects" folder itself.

Finally you do not launch AP26.exe, but just start and run BOINC.


Next 10 posts
[Return to PrimeGrid main page]
DNS Powered by DNSEXIT.COM
Copyright © 2005 - 2019 Rytis Slatkevičius (contact) and PrimeGrid community. Server load 1.97, 1.96, 2.13
Generated 22 Oct 2019 | 15:37:34 UTC