Cool! It's good to see Mac getting the support and recognition it deserves. How about starting a new thread in the BOINC forum to let Mac users know of your find, it will likely be missed by many in this thread. I think a few might be very grateful.

Yes, it's that time acoming where the blue V pill with the P additive needs to be rolled out again when the realisation sinks in that 40+ credit claims per hour won't stand.

Komick. you might want to get a later official client for PowerMAC here and be the best in the class with honours.....get the official here: BOINC: compute for science

ciao

Does that mean this this is rubbished?

8/15/2006 11:09:29 AM||Benchmark results:
8/15/2006 11:09:29 AM|| Number of CPUs: 2
8/15/2006 11:09:29 AM|| 5751 floating point MIPS (Whetstone) per CPU
8/15/2006 11:09:29 AM|| 26373 integer MIPS (Dhrystone) per CPU
8/15/2006 11:09:29 AM||Finished CPU benchmarks

It's on an "SSE2 optimized" client.

The official version's benchmarks simply disadvantages the C2D to an extent as the results of an intel X6800 EE chip are comparable with, even overshot by, AMD 3xxx or 4xxx chips;
Although in a majority of other major benchmarks, the C2DEE trounces even overclocked FX-62s by a decent margin.

But what I am seeing is this:
B00341_ 0259_ YTMA8B-5-7-9-c1 x965-providence Valid 08/15/2006 08:30:29 08/15/2006 10:04:34 1.53 102 / 22
B00338_ 0133_ YTMA8B-4-15-14-c1 x965-providence Valid 08/14/2006 16:31:44 08/15/2006 03:59:08 1.37 92 / 39
B00338_ 0046_ YTMA8B-4-10-3-c1 x965-providence Valid 08/14/2006 15:12:05 08/14/2006 16:34:22 1.29 25 / 15

In other words, before updating the client, my machine was basically looked down upon - a 25 credits already is already reasonably low for this class of processor - but then it gave me 15!
After updating, with the benchmark pasted above, ~90 was claimed, yet ~40 was given!

Edit:
And the most recent, it simply makes me mad - 22 for 102 claimed!? That doesn't even match with the previous! Can't the grid servers even recognise "larger than" and "smaller than"!?

Give me 39 for 1.4 CPU hours, give me 22 for 1.5 CPU hours on the very same machine a moment after!

If you tell me that math is being used to see how much each machine should be awarded per time spent on work unit, that seems UTTER BS.

If this is the same server that processes results, woe be us! It can't even tell big number from small number!! And that isn't an offensive remark, that is an observed fact in the comparison between wishy-wshiness and mathematics.

Before that, when I saw "90 claimed", I thought, "ah, it's finally warming up to the new kid on the block, but that's a little much", but then I get a fraction of that. A hundred for a 1.5 work unit is very off. I would have thought perhaps 45-65. But the reduction is simply illogical.

Is the grid thinking my chip is trying to skive?
It frankly is as fast as the most powerful stock-frequency CPU chip available today. Yet it does not get the crediting it deserves, in my opinion, even once the client benchmarks are in sync with it, because at the grid level, the scores are still downplayed.

I simply did not use the UD agent because my CPU score and total score was maxed out on it, and that surely would have ommitted a large amount of processing power from being recognised, on a single core alone.

I think I can appreciate the problem of overcrediting with unofficial clients, such as the one my benchmark seems crazily high on. But otherwise, the official is simply *undercrediting* the newer chips that have unprecedented processor power increments.

I think BOINC should come up with an official version that takes advantage of the latest processor technologies of each processor class available to conduct benchmarks, because those very technologies are being used to crunch the relevant research right here and right now, and they are producing results with due speed, not sitting there for show!

I am also for the integration of identity keys into its client so that the end product and the credits claimed can be registered as official, not from a rogue client.

And lastly, with the two above, WCG should tie crediting DIRECTLY to the claimed amount, not add a level of random "granting".

It's like "I like you now, heres 10 more credits."
Then the next moment, "I see you've done more work, but now I hate you, so I'm going to give you a fraction of what you should get."

We are a volunteer scientific research faculty, not a political party

I am loathe to take my machine off the grid because that is pure selfishness, but until this situation is rectified, it really does make things



--
This post has been edited for inappropriate language - nelsoc

Orca, (and anyone fitting the shoe), the only thing that comes to mind is that your BOINC BENCHMARK is hyperinflated. Understand clearly that the benchmark is an attempted representation of how fast your computer WILL complete a WU. U can benchmark all u want but if it does a WU in 1.5 hours and mine does it in 2 hours (benchmark 8.06 per CPU hour), your benchmark should be reading out a claimable per-hour credit of about 10.7 per CPU hour with some margin. If u turn in 102 for a job in 1.5 hours, and 2 others do same in 2 hours and claim 16 respective 20, yours and 16 will be dismissed and all get 20. Each time u come in with an bloated claim, and the 2 others in the quorum have a lower claim, u will get the middle of the 3.

There's a principle in BOINC which is 100,000 credits per 1 teraflop of crunched data. If u do 1 teraflop in X time, where everyone else does it on average in Y time, than u will still get 100,000 credits, only u get them 100,000 credits in whatever X on your machine stands for.

The credit system is under review and current discussion can be found in the "....fArce Majeure...." thread.

Basta per me,

Orca, it doesn't matter how shiny-fast you think your computer is. The WCG science applications are not optimised to take full advantage of every chipset feature you have available. So, using an "optimised" client will produce an unrealistic score and an unrealistic claim.

The credit system punishes this by throwing out your inflated claim.

Nothing to complain about, I think? Remember, the high and low claims are thrown out. If nobody is cheating, then the median claim will be close to the mean claim, and you will get a score very close to your claim. If someone is claiming unrealistic points, then you may see a large difference.

Also, remember that your SiSoft results combine both cores, but the BOINC score is per core.

WCG award points for how much work you do, not for how expensive your computer is. Sorry.