Proposed new BOINC credit system:

http://boinc.berkeley.edu/trac/wiki/CreditNew

Comments? Please no flames, just well thought out interpretations of how it will affect things here.

Well I'll be quite honest, at this time of night, a lot of that has gone straight over my head! But it does seem that a lot of thought has been given to this. They seem to have realised some of the shortcomings of the previous/existing systems, and have made a stab at suggesting a way forward.

If machine A can do 10 times as much work as machine B in the same time then it gets 10 times the credit. That is non-negotiable.

I have no problems with project credit parity. It makes people choose a project for its scientific worth rather than which one pays the most, and is as it should be.

I also see no reason why Alpha and Beta projects should'nt be allowed to award extra credit to compensate for the general ups and downs associated with those early stages, as indeed we see here.

I reackon they are on a hiding to nothing, as it is unlikely that they will get all project admins and the majority of users to all agree on a single system. And there willl always be those that will look for a way to cheat.

However, I do see this as a welcome development, PROVIDED they are issuing this as a discussion document and not as a prelude to uniaterally imposing it.

---

Don't drink water, that's the stuff that rusts pipes

Is there a translation of this into understandable English?

---

What this system proposes will not work until they get a gpu setup for all projects that works with ati and nvidia. when that happens then bring it on. But hey I have left many comments over the last year to DA and project admins to do this but most of the time I don't even get a response. So until that happens leave the credits here as they are and they should start getting there gpu program working.

Tada.!!!!!

Time marches on but the credits keep dropping....

P.S. I still think that you can not get any comparison between projects, this is do to the science apps and the lack of software support.

DD,,

---

>Is there a translation of this into understandable English?
No, Chris S has just watched too many episodes of Yes Minister. :)

---

Cheers,

PeterV

.

Is there a translation of this into understandable English?

The "proposal"? Or something else ... not sure what you mean ...

If you mean the proposal ... in essence it means that DA has finally co-opted the calibration concept that I proposed several years ago ...

Instead of using the worthless benchmarking system you actually use the work done and average it over the systems attached and the devices contained therein. It is not clear to me yet if the implementation of the system will match the proposal, though history says that it will not ... and the proposal will not be updated to reflect the changes ...

anyway, if you have a more specific question than that I can give a shot at trying to interpret what DA is proposing ...

I posted a long list of questions on the dev mailing list and over in the thread in Collatz; though history also says that this will not result in any meaningful changes because most of the effort expended will be to tell me I don't know what I am talking about ... rather than to answer the questions raised ... :)

Oh, and if you look at the section "Operation of the Improved Benchmark System" that I wrote lo these many years ago ... you can see that it is about the same system ... issue work, time the systems, perform averaging of the FLOPS and execution times ... and smooth out the differences ...

>Is there a translation of this into understandable English?
No, Chris S has just watched too many episodes of Yes Minister. :)

Hehe, Hi Peter! Yep, I can talk fluent Sir Humphrey Appleby! :-)