Claimed/granted credit

Message boards : Number crunching : Claimed/granted credit

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4

AuthorMessage
Keith Akins

Send message
Joined: 22 Oct 05
Posts: 176
Credit: 71,779
RAC: 0
Message 39868 - Posted: 25 Apr 2007, 17:02:57 UTC

First of all, let me apologize to anyone I might have unintentionally offended. I was simply trying to make the point that if my RAC were somewhat stable to what it used to run, then all would be well.

If a system, by honest means, consistantly holds an RAC of 235+ since the introduction of the new credit system over several months, then there's no reason why it shouldn't hold in this range. I kept waiting for this recent drop in RAC to even out. It kept falling. This is all I was trying to say.

Again Apologies.
ID: 39868 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 39869 - Posted: 25 Apr 2007, 17:06:27 UTC
Last modified: 25 Apr 2007, 17:08:36 UTC

In the end, credit are intended to correspond to cobblestones. Cobblestones are a measure of computing calculations performed. I am not a CPU expert, but I believe if you go to your favorite CPU manufacturer and ask them how many cobblestones per hour a given CPU can crunch... they will decline to answer the question.

The reason they will decline to answer the question is because application mix is too complex for them to presume how many cache misses there will be, and what the average utilization on the instruction stream is. They will only state how fast they run a given benchmark (i.e. a known application with a predetermined instruction path). And if you run a given benchmark more then once, you will have different results. Why? If a given CPU is capable of a given benchmark rating, then why does the benchmark result vary from one run to the next? There are still unknown factors such as the operating system handling events and etc. that pull time away from the actual execution of the benchmark.

Some CPUs are capable of having 4 instructions at various stages of execution at the same time. But only if the instructions in the program happen to request exactly the particular functions in the order the CPU can support them. Where any deviation from that perfect world exists, then you get less then 4 instructions in progress. The extent that your application runs with less instructions in the pipe will largely determine how much actual useful crunching is performed. If your application is preempted in order to run another thread, such as handling activity on the mouse, you lose the instructions you had in progress at the time, and have to use 3 or 4 clock ticks to get back to where you were when you lost the CPU. So, number of preemptions occuring is important as well.

All of these factors are why the CPU manufacturers are starting to talk less and less about GHz, and more about benchmarks. Someone that wants a gaming PC doesn't care how many clock ticks per second the CPU does, they care how much work they have done at the end of that second.

In a simplistic world, you could take CPU GHz and compute how many cobblestones per hour you will produce. But reality never works that way.

Rosetta's credit system is based on the average claims coming in for that given work unit. These average credit claims are based on the BOINC benchmarks reported for the host. As was pointed out, there are third party versions of BOINC which measure and report their benchmark results differently. These (often higher) reported credit claims are averaged in with those from the standard BOINC clients. And the specific host does not directly benefit from any credit claim that varies from the standard benchmark.

I would point out that the OTHER thing that probably occured around that same timeframe was that more and more of the installed Rosetta user-base went back to the standard BOINC version and it's (often lower) benchmarks. Therefore the average host's claimed credit was reduced, and the rolling average credit per model on the work units declined... for everyone. And I would assert that a decline due to more hosts reporting standard benchmarks with their results means that Rosetta's granted credits are now closer to the intended cobblestone measure then they were prior to that.

[edit]This could also explain why credit continues to decline slightly, as more and more hosts replace an optimized client with the standard client.
Rosetta Moderator: Mod.Sense
ID: 39869 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Nemesis
Avatar

Send message
Joined: 12 Mar 06
Posts: 149
Credit: 21,395
RAC: 0
Message 39890 - Posted: 26 Apr 2007, 2:29:47 UTC
Last modified: 26 Apr 2007, 2:52:41 UTC

Cobblestones are a mythical measure of something, invented by BOINC, based on some mythical machine doing some mythical task.

Claimed credits are based on an astoundingly badly written set of benchmarks that have no relation to actual performance of the box. Witness the *nix vs. Windoze battles over BOINC benchmarks.

The result is that the projects "twiddle" their credit granting schemes to conform to the megalomaniac David Anderson's pipedream about cross-project parity of credit. It's impossible, but the projects waste endless resources and energy trying to achieve it.

Cross-project credit parity is the worst concept to ever hit distributed computing. It's worthless, meaningless, and unobtainable. Sooner or later some project is just going to break out of the fold and start rewarding credits on their own terms, and to hell with the BOINC Emperor. Only then will cross-project parity truly die a deserved death, and each project will stand on it's own.


Nemesis n. A righteous infliction of retribution manifested by an appropriate agent.


ID: 39890 · Rating: -2 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Greg_BE
Avatar

Send message
Joined: 30 May 06
Posts: 5652
Credit: 5,622,096
RAC: 26
Message 39921 - Posted: 26 Apr 2007, 18:56:04 UTC

Excuse You!

read this:

Cobblestones

The original name for Credits.
Named after Jeff Cobb, one of the Seti administrators.

A Cobblestone is 1/100 day of CPU Time on a Reference Computer that does:

* 1,000 double-precision MIPS based on the Whetstone Benchmark.
* 1,000 VAX MIPS based on the Dhrystone Benchmark.

you tell me that this precise calculation is fictional? Thats a insult to the person who developed this measurement and spent quite a bit of time coming up with the code and math to make it all work.

I take offense at your slight/insult at another person who is putting time into looking at how linux will work with credits (even though i use windows, i have to acknowledge those who are trying to break the monolithic hold of bill gates). Read this

If your so into credits and the such and not into sharing your cpu in the name of science, then why don't you go elsewhere. I came into this project to donate my cpu time to help finding ways to get to the bottom of diseases such as cancer (my mother in law died from it) and other diseases that this project is helping to find a cure for.

Credits to me is just a way to show what I have done so far. Does credits have anything to do with anything in the real world? I don't think so. Does the results of these work units have something to with the real world? Damn straight they do.

ID: 39921 · Rating: 3 · rate: Rate + / Rate - Report as offensive    Reply Quote
Nemesis
Avatar

Send message
Joined: 12 Mar 06
Posts: 149
Credit: 21,395
RAC: 0
Message 39924 - Posted: 26 Apr 2007, 23:06:10 UTC - in response to Message 39921.  
Last modified: 26 Apr 2007, 23:43:36 UTC

Excuse You!

read this:

Cobblestones

The original name for Credits.
Named after Jeff Cobb, one of the Seti administrators.

A Cobblestone is 1/100 day of CPU Time on a Reference Computer that does:

* 1,000 double-precision MIPS based on the Whetstone Benchmark.
* 1,000 VAX MIPS based on the Dhrystone Benchmark.


Thanks for making my point. Do you *really* think that is the definition of a credit in today's BOINC world?

you tell me that this precise calculation is fictional? Thats a insult to the person who developed this measurement and spent quite a bit of time coming up with the code and math to make it all work.

I take offense at your slight/insult at another person who is putting time into looking at how linux will work with credits (even though i use windows, i have to acknowledge those who are trying to break the monolithic hold of bill gates). Read this

If your so into credits and the such and not into sharing your cpu in the name of science, then why don't you go elsewhere. I came into this project to donate my cpu time to help finding ways to get to the bottom of diseases such as cancer (my mother in law died from it) and other diseases that this project is helping to find a cure for.

Credits to me is just a way to show what I have done so far. Does credits have anything to do with anything in the real world? I don't think so. Does the results of these work units have something to with the real world? Damn straight they do.


So why are you getting all excited in a discussion about credit?

I think all the self-righteous "in it only for the science" 1%rs should have the option to turn off credit granting for their account, so they can show the world they really don't care about the credit. How many do you think would do that? I might not even run out of fingers and toes counting.
Nemesis n. A righteous infliction of retribution manifested by an appropriate agent.


ID: 39924 · Rating: -2 · rate: Rate + / Rate - Report as offensive    Reply Quote
Mod.Sense
Volunteer moderator

Send message
Joined: 22 Aug 06
Posts: 4018
Credit: 0
RAC: 0
Message 39926 - Posted: 27 Apr 2007, 1:11:01 UTC

My mention of cobblestones was a reference to how 100,000 cobblestones equals a TeraFlop, a well-defined computing term. And I think we all agree that credits are simply an approximation thereof.

...so let's not escalate this into a shouting match.
Rosetta Moderator: Mod.Sense
ID: 39926 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Nemesis
Avatar

Send message
Joined: 12 Mar 06
Posts: 149
Credit: 21,395
RAC: 0
Message 39937 - Posted: 27 Apr 2007, 7:32:43 UTC - in response to Message 39926.  

My mention of cobblestones was a reference to how 100,000 cobblestones equals a TeraFlop, a well-defined computing term. And I think we all agree that credits are simply an approximation thereof.

...so let's not escalate this into a shouting match.


Cobblestones mean absolutely NOTHING outside of BOINC. It's a BOINC invention. I can't see how it has any correlation to flops or TeraFlops.

Nemesis n. A righteous infliction of retribution manifested by an appropriate agent.


ID: 39937 · Rating: -2 · rate: Rate + / Rate - Report as offensive    Reply Quote
transient
Avatar

Send message
Joined: 30 Sep 06
Posts: 376
Credit: 10,836,395
RAC: 0
Message 39956 - Posted: 27 Apr 2007, 16:06:37 UTC - in response to Message 39937.  

My mention of cobblestones was a reference to how 100,000 cobblestones equals a TeraFlop, a well-defined computing term. And I think we all agree that credits are simply an approximation thereof.

...so let's not escalate this into a shouting match.


Cobblestones mean absolutely NOTHING outside of BOINC. It's a BOINC invention. I can't see how it has any correlation to flops or TeraFlops.


LOL, like (Tera)Flops have absolutely no meaning outside of the world of computing?

Sorry, couldn't resist this give-away.
ID: 39956 · Rating: -1 · rate: Rate + / Rate - Report as offensive    Reply Quote
Previous · 1 · 2 · 3 · 4

Message boards : Number crunching : Claimed/granted credit



©2022 University of Washington
https://www.bakerlab.org