I would like to understand the value of the “Run Benchmarks” exercise, listed under the file tab, menu within the BOINC Client?

In particular:

Will the results or out come be up loaded to the projects I’m participating in?

If so, will this lead to an increase or decrease in the “Chunk” size of the number of Work Units that will be downloaded to my CPU in the future?

If so, will this assist the projects, I’m involved with, help determine the Work Unit partners I’ll be validating with?

A premature thanks for any and all responses.

Regards,
Tom

---

Theory of Gravitational Waves & LIGO
Paul's BOINC WIKI of Technical Info
JPL-Caltech

Tom if you have not looked at the Wiki you may want to start there...


Benchmark_Test

---

BOINC Wiki

Tom if you have not looked at the Wiki you may want to start there...

Jim – Sincerely appreciate your posting of the WIKI link, but that was my first option, for my review, when my outlined (within my posting) questions came to mind. Perhaps, I have over looked something or may not have understood what Paul was conveying to me, but I couldn’t locate the responses I needed for my three inquiries.

Will the results or out come be up loaded to the projects I’m participating in?

If so, will this lead to an increase or decrease in the “Chunk” size of the number of Work Units that will be downloaded to my CPU in the future?

If so, will this assist the projects, I’m involved with, help determine the Work Unit partners I’ll be validating with?

Thanks,
Tom

---

Theory of Gravitational Waves & LIGO
Paul's BOINC WIKI of Technical Info
JPL-Caltech

I would like to understand the value of the “Run Benchmarks” exercise, listed under the file tab, menu within the BOINC Client?

Short answer; it is there only so that if the automatic weekly running of the benchmarks happens to occur at an especially bad time (while playing three streamed videos and a game of Doom...) and gives you ridiculously low values, you can quit everything else and manually run them. The "Your computers" link on the web pages shows the benchmark values - if you notice at some point that they have changed drastically downwards, you can manually run them again to fix it, or just wait a week.

One useful thing I've found is that if I manually run the benchmarks on Sunday at 2AM, they'll continue to run every Sunday at around 2AM - when I'm not very likely to be heavily using the computer.

These are the benchmarks used to calculate your credits - running them an "extra" time won't do anything else helpful.

---

Will the results or out come be up loaded to the projects I’m participating in?

Yes, at the next scheduler request message the updated benchmarks will be sent to the project. over time all of the projects will get the information about the last run of the benchmarks.

Because of the variability of the benchmark results, this may change the reported "speed" of your system.

If so, will this lead to an increase or decrease in the “Chunk” size of the number of Work Units that will be downloaded to my CPU in the future?

It can. If you use a version of the software 4.7x or later, there are other correcting values used to attempt to mitigate the variability of the predicted processing times.

If so, will this assist the projects, I’m involved with, help determine the Work Unit partners I’ll be validating with?

No. This will still be pretty much a random draw. There were discussions on introducing code in the scheduler that would allow the projects to assign work to computers with the same approximate turn-around time, but I am not sure that this has been implemented yet by any project.

---

Everyone, ”Thanks” for your time and energy, in providing your responses to my posting(s).

Paul, personally, only an opinion, I would like to see the projects attempt to align the users as closely as possible, based on their respective turn around times. By doing so, I feel “Credits” could be issued to their contributors quicker, thus for more time consuming, work unit related projects, users would stay with their chosen projects, without skipping from one project to another, using the excuse of “Where’s my credit or why do things take so long”.

In essence, it could provide an enhanced stability to each project, which in-turn provide a greater stability to all projects and perhaps allowing the folks at S@H to catch-up some, during their “project down times”.

If so, will this assist the projects, I’m involved with, help determine the Work Unit partners I’ll be validating with?

No. This will still be pretty much a random draw. There were discussions on introducing code in the scheduler that would allow the projects to assign work to computers with the same approximate turn-around time, but I am not sure that this has been implemented yet by any project.

Guys, Again Thanks,
Tom

---

Theory of Gravitational Waves & LIGO
Paul's BOINC WIKI of Technical Info
JPL-Caltech

Tom,

My opinion also.

The projects have a competing problem, how to update the project site, their application, and the user base to new versions.

Though many are not happy with some of the ways that the 4.xy versions have, um, worked (or not worked if you did not like the version) we have had a pretty transparent upgrade path. For those of us that remember 3.x to 4.x, well, we may have a repeat in 4.x to 5.x. I hope not, but it may happen.

So, until all project upgrade to the later versions of the server software, and turn on the feature (if I recall it is a configurable feature - but I have not researched this question out fully, YET).

But, I also think that eventually to really get ahead of the power curve with this type of process is to also correct the major deficiencies with the benchmark system. So far, all the work that has been done is to aleaviate some of the worst aspects. Anyway, we are improving, and things, believe or not, are getting better.

---