What determines how much credit is assigned for work done? I'm confused, because I rarely get the amount of credit my systems are claiming, and certainly no more. I had thought it could be because my machines might be less efficient than average somehow, yet the Einstein@home work I've done (on the exact same systems) has so far been given at least the amount claimed in each case, and sometimes up to about fifty percent more. What might be the difference between the crediting processes of the two projects? Any suggestions for getting a larger proportion of the S@h credit I'm claiming or, alternatively, configuring my systems to make more realistic claims?

I'm running BOINC Menubar 5.2.13 under Mac OS X v10.3.9.

BTW, is there something wrong with the Pending Credit page here or perhaps with my browser (Safari 1.3.1)? It never shows any results, but OTOH the one at E@h works fine for me.

---

What determines how much credit is assigned for work done? I'm confused, because I rarely get the amount of credit my systems are claiming, and certainly no more.

This is one of those questions that can turn into a monster thread.

Basically, your machine(s) benchmark better than they actually crunch. That isn't good or bad -- the benchmark has to be project-neutral, and there are too many variables to benchmark with 100% accuracy.

This is the nature of benchmarks going back to the beginning of computing.

I picked this one more or less at random.

Three machines returned work. You claimed 30.73, the other two claimed 20 and 10.14. Throw out the lowest and highest, and average the remaining machine(s), and you're granted 20.

On this one you claimed 30.46, the other machines claimed 31.47 and 19.64. Throw out the top and bottom (and the bad result) and your claimed 30.46 is granted to all three machines.

Your claimed credit looked fairly reasonable to me based on what I saw.

Sometimes you'll get a gift from the other crunchers (because their claims were high) and sometimes you'll get matched up with some really low claimed credit and you'll lose a little. It evens out.

---

Basically, your machine(s) benchmark better than they actually crunch. That isn't good or bad -- the benchmark has to be project-neutral, and there are too many variables to benchmark with 100% accuracy.

Thanks for that, and the explanation that the highest and lowest claims are dropped with the remainder being averaged ... but if it's indeed the case that my benchmarks are over-optimistic, I still don't understand why there's such a discrepancy between the S@h and E@h results. Of course I haven't been using BOINC very long, so I suppose there's a chance that the pattern I'm seeing is a random anomaly ...

---

Basically, your machine(s) benchmark better than they actually crunch. That isn't good or bad -- the benchmark has to be project-neutral, and there are too many variables to benchmark with 100% accuracy.

Thanks for that, and the explanation that the highest and lowest claims are dropped with the remainder being averaged ... but if it's indeed the case that my benchmarks are over-optimistic, I still don't understand why there's such a discrepancy between the S@h and E@h results. Of course I haven't been using BOINC very long, so I suppose there's a chance that the pattern I'm seeing is a random anomaly ...

With S@H and E@H you get about the same credits per hour, the units from E@H are much larger and take a lot longer so you get more credits for them. The only veriation from there would depend on what other systems got the same unit.

Climate Prediction is even longer, I took 4 Mo. to finish one Sulpher Model (running about ½ the time) but got 19,281.57 credits for it.
Ray

---

Pizza@Home Rays Place Rays place Forums

Basically, your machine(s) benchmark better than they actually crunch. That isn't good or bad -- the benchmark has to be project-neutral, and there are too many variables to benchmark with 100% accuracy.

Thanks for that, and the explanation that the highest and lowest claims are dropped with the remainder being averaged ... but if it's indeed the case that my benchmarks are over-optimistic, I still don't understand why there's such a discrepancy between the S@h and E@h results. Of course I haven't been using BOINC very long, so I suppose there's a chance that the pattern I'm seeing is a random anomaly ...

Re-read the part about the benchmarks being an estimate of performance.

Key word is "estimate"

It measures things like math performance, and it doesn't measure other things, like cache size or memory bandwidth that may matter more for some projects than others.

Some will tell you that this is a critical flaw in BOINC that can't possibly be tolerated.

This is far more of a social issue than a technical one. You appear to be getting about the right credit.

When the enhanced science app comes out, it actually counts operations and credit is based on that. It isn't subject to variations in the benchmark.

Of course, it does alot more "accounting" work than the current science app., but too many people seem to think absolutely accurate accounting is a must.

-- Ned

---

With S@H and E@H you get about the same credits per hour, the units from E@H are much larger and take a lot longer so you get more credits for them.

It's not the difference in 'absolute' credit per WU I'm wondering about, but the difference in the ratio between the claims and granted credits.

From E@h, for the ten completed WUs that are listed at the moment, I never received less credit than claimed, and typically about 50% more. In comparison, I have eleven S@h results showing: six of them were granted the amount claimed (a higher proportion than last time I checked--I admit my original posting appears to have overstated the discrepancy), but the other five received only about 60-80% as much.

---

BTW, is there something wrong with the Pending Credit page here or perhaps with my browser (Safari 1.3.1)? It never shows any results, but OTOH the one at E@h works fine for me.

It has not worked for the 1.5 years I have been here. It's a feature that was pulled early because of the amount of database calls it does and it nearly took the systems down when it was on.

There have been talks that they would do snapshots of the pending credit and show that snapshot, but that's not been talked about for a long time. I am unsure the developers know what's on the block for that, currently. They are so busy with other things.

---

My movie https://vimeo.com/manage/videos/502242

Odysseus,
You will find as you do different projects, different combinations of hardware/OS gain differing amounts of credits, which at first glance seems strange. But with further observation there tends to be a pattern. Seti crunches faster with large L2 cache and therefore tends to favour Intel CPU's, althought the later AMD's are catching up fast. On Einstein the swing is towards AMD and very much towards G5 Macs, and CPDN likes lots of memory.
So those that chase credits tend to put their computers where they gain most credits/'time period'. And this seems to be borne out on the Stats sites that publish figures of project/cpu and project/OS.

---

From E@h, for the ten completed WUs that are listed at the moment, I never received less credit than claimed, and typically about 50% more. In comparison, I have eleven S@h results showing: six of them were granted the amount claimed (a higher proportion than last time I checked--I admit my original posting appears to have overstated the discrepancy), but the other five received only about 60-80% as much.

To be very specific here... the Einstein "Albert" application is Altivec-optimized. It runs TWICE as fast on a G4 or G5 than it does on a "matching benchmarks" Intel box. The SETI application is NOT optimized. This means that the benchmarks common to 'slower' Macs are indeed a bit "high" for SETI, and you claim "too much" credit; but the _same_ benchmark value is way _low_ for Einstein, making you claim too _little_ credit there.

Look here and you will find an Altivec-optimized SETI application from Team MacNN, that will cut your time on SETI results almost in half... and making SETI and Einstein "equal" on your Mac. Team MacNN _also_ has an optimized BOINC Client, that will raise your benchmarks to "match" (or imho exceed) what they should be for both SETI and Einstein.

Examples: without optimized SETI app, Mac claims 30 credits, gets 20. With optimized app, Mac does work in half the time, claims 15 credits, gets 20. With optimized client, Mac claims 35, still gets 20. Obviously the biggest advantage is in the optimized application, because you get almost twice as many credits per day. (Look at the "top 20 computers" list on Einstein; 6 of them are Quad G5's, thanks to the Altivec-optimized app.)

Seti crunches faster with large L2 cache and therefore tends to favour Intel CPU's, althought the later AMD's are catching up fast.

Having crunched with a considerable variety of CPUs over the last 15 months, my experience has been that large L2 caches are only significant with Intel CPUs and AMDs, for any given clock speed, are a heck of a lot faster than Intel.

---

Be lucky

Neil

Look here and you will find an Altivec-optimized SETI application from Team MacNN, that will cut your time on SETI results almost in half... and making SETI and Einstein "equal" on your Mac. Team MacNN _also_ has an optimized BOINC Client, that will raise your benchmarks to "match" (or imho exceed) what they should be for both SETI and Einstein.

Sorry If I seem clueless, but I just see a discussion forum at that link, showing a thread dating back to last August--was the optimized S@h app posted in one of the messages? I tried the "G4" link in the first message but it appears to be dead: my browser says the server inst.eecs.berkeley.edu couldn't be found.

Doesn't BOINC automatically download the programs that actually perform the work for each project? What happens when a newer 'vanilla' version comes out: will it replace the optimized one? Why doesn't BOINC choose the best version for the system it's running on to start with?

---

So those that chase credits tend to put their computers where they gain most credits/'time period'. And this seems to be borne out on the Stats sites that publish figures of project/cpu and project/OS.

All my systems are very similar--old G4 Macs--so it's not really relevant to me at the moment, but how does one go about putting computers where they gain most? AFAICT they all get their preferences from my account: there's practically nothing configurable per machine. Supposing I were to get an Intel box and didn't want it to 'waste its time' on E@h--how could I make it change its priorities? Right now my preferences say 200% S@h, 100% E@h: if I were to change that, it would affect all my systems, wouldn't it? Wouldn't I have to create a separate account?

---

All my systems are very similar--old G4 Macs--so it's not really relevant to me at the moment, but how does one go about putting computers where they gain most? AFAICT they all get their preferences from my account: there's practically nothing configurable per machine. Supposing I were to get an Intel box and didn't want it to 'waste its time' on E@h--how could I make it change its priorities? Right now my preferences say 200% S@h, 100% E@h: if I were to change that, it would affect all my systems, wouldn't it? Wouldn't I have to create a separate account?

This can be configured in your Account settings for a limited number of catagories. Go to Your Account and select View or Edit Seti@Home preferences. From there you can configure your Default, Work, School, and Home catagories. Just because they are called Work, School, and Home, doesn't mean you have to use them at those locations...configure them however you want/need.

Choose one catagory and set it for a high resource share (pick a number). Set the default for another value, set another location at another value and so on. Do the reverse thing or similar (you configure how you want it) at your account at E@H.

The last thing you need to do is to configure each machine definition to use the appropriate catagory (Work, School, Home) that gives it the resource share you want it to have. Select View Computers and choose the machine you want to configure. At the bottom of the page is a selectable Location drop-down. Set it to your desired location catagory. Be sure to do this for that computer at each project. Do an Update from that computer (or wait until it connects by itself) and it will start using the new resource shares.

What happens when a newer 'vanilla' version comes out: will it replace the optimized one?

Good question... I'm running an optimised science (seti) app. If a new "official" science application is automatically downloaded, will it overwrite the optimised one i'm using?

---

What happens when a newer 'vanilla' version comes out: will it replace the optimized one?

Good question... I'm running an optimised science (seti) app. If a new "official" science application is automatically downloaded, will it overwrite the optimised one i'm using?

No, the presence of the app_info.xml file will prevent the download of a new app automatically.

---

Be lucky

Neil

Sorry If I seem clueless, but I just see a discussion forum at that link, showing a thread dating back to last August--was the optimized S@h app posted in one of the messages? I tried the "G4" link in the first message but it appears to be dead: my browser says the server inst.eecs.berkeley.edu couldn't be found.

I just tested the link for the G4 version and it downloaded fine; the server may have been down when you tried. The link is in that first message, from when they first released the "alpha-1" version, but the link goes to whichever version is current (at this point alpha-5).

Doesn't BOINC automatically download the programs that actually perform the work for each project? What happens when a newer 'vanilla' version comes out: will it replace the optimized one? Why doesn't BOINC choose the best version for the system it's running on to start with?

BOINC does automatically update versions _if_ you have not "overridden" it by installing your own. When SETI_enhanced is released, if you've installed this G4 application, you will eventually get "no work from project" messages constantly, and you will have to manually uninstall it, at which point BOINC will go get whatever the project servers have available. As to why SETI doesn't automatically issue optimized applications like Einstein does, I have no idea.

It's up to you whether it's worth the effort to do all this or not; it cut my times on two G4's in half, and I don't mind playing with installing and deleting things, so to me it was worth it.

I just tested the link for the G4 version and it downloaded fine; the server may have been down when you tried. The link is in that first message, from when they first released the "alpha-1" version, but the link goes to whichever version is current (at this point alpha-5).

Thanks; it worked today. Now I have a folder containing a Terminal document called "seti@home-G4-a5" and a tiny file called "app_info.xml". The former doesn't seem to do anything when I run it, and there's no other installer or ReadMe. Could it be a corrupt download?

---

Thanks; it worked today. Now I have a folder containing a Terminal document called "seti@home-G4-a5" and a tiny file called "app_info.xml". The former doesn't seem to do anything when I run it, and there's no other installer or ReadMe. Could it be a corrupt download?

No - that's the right stuff. There should have been a "how to install" document around there somewhere... but it's extremely simple.

Quit BOINC. (Some instructions advise finishing the WU in progress first, as it _may_ be errored out by the change in apps, but I have had no problems. If it's at 90% though, you might want to let it finish.) Find the /Library/Application Support/BOINC Data/projects/setiathome.berkeley.edu folder. Copy seti@home-G4-a5 and app_info.xml both into this folder. Restart BOINC. That's it!

Quit BOINC. (Some instructions advise finishing the WU in progress first, as it _may_ be errored out by the change in apps, but I have had no problems. If it's at 90% though, you might want to let it finish.) Find the /Library/Application Support/BOINC Data/projects/setiathome.berkeley.edu folder. Copy seti@home-G4-a5 and app_info.xml both into this folder. Restart BOINC. That's it!

Thanks again. Beside four files that appear to be the work-units and/or results themselves, in that folder there's also a generic document called "setiathome_4.18_powerpc-apple-darwin", which I take it is the current SETI app, and an image of the S@h logo called "better_banner.jpg". (I suppose the latter has something to do with the screensaver, which I don't have.) Should I remove these, or will BOINC ignore them once the new files are there?

How can I tell how much of a WU is done when it's not actually running, i.e. when BOINC is on another project? With the old CLI version of S@h I could look in the "state.sah" file to find this information; is there an equivalent now?

---

You're using Menubar, so I don't know of a way to tell, other than the % complete when it _is_ running.

All you do is add the two new files; BOINC takes care of the rest, will delete what it doesn't need (4.18) and keep what it does (the WUs).