Computer/project/app efficiency

Message boards : Web interfaces : Computer/project/app efficiency
Message board moderation

To post messages, you must log in.

AuthorMessage
BeauZaux

Send message
Joined: 5 Apr 20
Posts: 2
United States
Message 97267 - Posted: 5 Apr 2020, 2:28:03 UTC

Looking for an app that collects project/app completion times by computer?
I want to see which projects run best on which of my machines, configurations, and project combinations. Data can be extracted from each project's my account/tasks pages, but not sortable or easily compared to other projects. Wish I were a programmer. And yes, still a newbie, but an old tired newbie. ;-)
ID: 97267 · Report as offensive
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 97269 - Posted: 5 Apr 2020, 5:31:29 UTC
Last modified: 5 Apr 2020, 5:43:06 UTC

For GPUs and CPUs, you can just look at the power draw (watts), temperature and frequency. (As well as core count and load).

Usually when the load is high, the temperature rises, and the frequency drops. Those are good projects. They fully tax a system.

When the load is low, power consumption is low, and the project doesn't use the hardware as well.

There is no program that can measure time of tasks for measuring hardware speed, as certain projects have same tasks sometimes differing 10%, sometimes as much as 200%.

The best way to know hardware speed is by using programs like blender or pcmark.
ID: 97269 · Report as offensive
Les Bayliss
Help desk expert

Send message
Joined: 25 Nov 05
Posts: 1654
Australia
Message 97270 - Posted: 5 Apr 2020, 6:56:12 UTC - in response to Message 97267.  

You could try WUProp@Home
ID: 97270 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 97273 - Posted: 5 Apr 2020, 8:45:40 UTC
Last modified: 5 Apr 2020, 9:46:50 UTC

Yes, use WUProp for that. See https://wuprop.boinc-af.org/results/delai.py for all the options this project checks.

If you run it yourself it will send information about your hardware vs projects back for comparison.
ID: 97273 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 97274 - Posted: 5 Apr 2020, 8:57:15 UTC - in response to Message 97269.  
Last modified: 5 Apr 2020, 10:00:14 UTC

There is no program that can measure time of tasks for measuring hardware speed, as certain projects have same tasks sometimes differing 10%, sometimes as much as 200%
There is still a lot about BOINC that you have to learn. I have asked you this before and I will ask you again, when you post advice like that to post it on your personal title, "in my opinion". You want so much to have the first answer in and be right, when will you research your answer, see if what you're about to put down is correct, before putting in an answer? Can't you see that you can do much more damage than help by answering uninformed?

Even those of us who have been working with BOINC since its conception, who help on various forums, write code for it, or its documentation, will at times research if the answer we want to put down is correct. I certainly do.
ID: 97274 · Report as offensive
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 97285 - Posted: 5 Apr 2020, 14:05:58 UTC - in response to Message 97274.  
Last modified: 5 Apr 2020, 14:08:33 UTC

There is no program that can measure time of tasks for measuring hardware speed, as certain projects have same tasks sometimes differing 10%, sometimes as much as 200%
There is still a lot about BOINC that you have to learn. I have asked you this before and I will ask you again, when you post advice like that to post it on your personal title, "in my opinion". You want so much to have the first answer in and be right, when will you research your answer, see if what you're about to put down is correct, before putting in an answer? Can't you see that you can do much more damage than help by answering uninformed?

Even those of us who have been working with BOINC since its conception, who help on various forums, write code for it, or its documentation, will at times research if the answer we want to put down is correct. I certainly do.

Correction sir,
There is NO program that can measure the time it takes to finish WUs to measure performance, simply because the method js not feasible.
If same project, same task differs as much as 200% in length to finish, what program are you going to recommend? WUs are variable in length. I see that all over the place, for nearly every project!


You probably misunderstood what I wrote, and I might have worded it a bit less clear, but I don't think what I intended to write was wrong.

If anything, I certainly didn't cause any damage, by sharing what I believe is correct.
ID: 97285 · Report as offensive
robsmith
Volunteer tester
Help desk expert

Send message
Joined: 25 May 09
Posts: 1283
United Kingdom
Message 97287 - Posted: 5 Apr 2020, 14:50:20 UTC - in response to Message 97285.  

Sadly you can do harm by telling newcomers inaccurate information.
You would do well to sit back, learn and understand what BOINC is capable of, and how it works before answering or responding to questions.
ID: 97287 · Report as offensive
ProDigit

Send message
Joined: 8 Nov 19
Posts: 718
United States
Message 97334 - Posted: 7 Apr 2020, 16:47:17 UTC - in response to Message 97287.  

Sadly you can do harm by telling newcomers inaccurate information.
You would do well to sit back, learn and understand what BOINC is capable of, and how it works before answering or responding to questions.

Someone can always try to interpret what someone shares, or try to fault find it.
If what I wrote is erroneous, it would be simple to just correct it, and we'd all learn.
But as far as I'm aware, you can't measure system efficiency based on variables. (WUs are variable in length of time to finish, therefore aren't a feasible form of benchmark).

Power consumption, and system load are better indicators. Or, one could go by PPD, but projects offering higher PPD doesn't mean the tasks run more efficient on the hardware.
ID: 97334 · Report as offensive
robsmith
Volunteer tester
Help desk expert

Send message
Joined: 25 May 09
Posts: 1283
United Kingdom
Message 97336 - Posted: 7 Apr 2020, 17:14:38 UTC

Try reading what the OP asked, and what Les & Jord replied with - then you will learn.
I'm not familiar with WUProp, but your initial answer was very wrong in that it did not address the OP's question in the least - he was asking how to collate the way his system was behaving across a number of projects. I stayed out at that time because I didn't know an appropriate solution. Glancing at WUProp I see it endevours to collate the use of various parts of a system and then produce a metric for overall performance in a comparative manner.

You keep on about TPD etc as being "good" metric for system performance, let me try and explain to you (again) why that is NOT the case. You have a very planar and simplistic view that every project's applications run in the same way and use the system resources in a very similar manner. They don't, some will make extensive (if not exclusive) use of one part of a system e.g. Colatz(sp?) & PrimeGrid do a lot of integer maths that can be processed in parallel, while a protein folding application is doing some very heavy floating point matrix manipulations that can only be done in series and require substantial amounts of memory, so while both appear to be only using one part of the system heavily they are each working in an efficient manner towards the required solution. Over the years I have seen far too many programs that have been so badly written that the hog system resources, overload parts of a system while doing needless operation, and all the while return poor quality results, if not total rubbish (I will admit to having written a few myself in the past).
ID: 97336 · Report as offensive
Profile Jord
Volunteer tester
Help desk expert
Avatar

Send message
Joined: 29 Aug 05
Posts: 15477
Netherlands
Message 97349 - Posted: 8 Apr 2020, 0:38:11 UTC
Last modified: 8 Apr 2020, 0:42:35 UTC

WUProp@home is a non-intensive project that uses Internet-connected computers to collect workunits properties of BOINC projects such as computation time, memory requirements, checkpointing interval or report limit.

But then again, who are we? What do we know? What do the project developers at WUProp know? Probably nothing.
ID: 97349 · Report as offensive
BeauZaux

Send message
Joined: 5 Apr 20
Posts: 2
United States
Message 99267 - Posted: 12 Jun 2020, 18:56:21 UTC

Sorry I have not inputted, but was watching for some sort of usable answer. Wish I were a programmer, because it seems with all that is out there, this would be simple to write, taking data from BOINC, Project site, and Cross-project stat sites. I started keeping a hand written log on 4 machines for foldin@home, and can clearly see advantages in some machines. Now experimenting with equipment and configurations. BOINC logging will be a whole other matter.
Benchmarks are good, but would be better if someone made Folding or BOINC project benchmark tools.
Just throwing ideas out there that may challenge some kid to step up. ;-)
Oh, I did WUProp a while back, but didn't see the benefit. I'll have to look again.
Thanks for your comments.
ID: 99267 · Report as offensive

Message boards : Web interfaces : Computer/project/app efficiency

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.