[…] my RAC is plummeting because a host of arbitrary "wingmen" aren't able to keep pace! What's a credit whore supposed to do? (Tongue firmly in cheek, of course; but it IS annoying!)

Think of it as an investment: when the credits do come through, it’ll be like a surprise dividend. It all works out in the end … I had to detach some hosts (belonging to my employer) last autumn, but as of early December they were still yielding a little credit.

---

I don't mean to intrude (yes I do), but on the side subject of RAC, it is manifestly annoying when 77% of my recent results have yet to be credited. Actually, it is probably closer to 70% since some of them are still in process, but it really chafes my posterior that I'm cranking them out about 3 work units per day for each of my machines, but my RAC is plummeting because a host of arbitrary "wingmen" aren't able to keep pace! What's a credit whore supposed to do? (Tongue firmly in cheek, of course; but it IS annoying!)

---

Thanks, Randy, for the interesting info. I guess that's why it's almost impossible to achieve even close-to-even CPH for every machine, every workunit and every project (based on machine speed and devotion to project computation). We'll just have to like or lump the CPH disparity. The averaging effect of different workunits really helps here.

---

You're getting a boost from the shorties because you're running on Intel processors. My fastest AMD system is getting an approx. 3 cr/hr penalty with those WUs. Whereas, with some VLAR WUs, that same system will get about a 3 cr/hr boost over a more average AR.

One can easily calculate that by looking at his/her results and dividing the number of credits by the number of seconds, then multiplying that result by 3600. For example if my machine calculates a 60-credit unit in 9000 seconds that's (60/9000) * 3600 which is 24 credits per hour. Note that CPH with the same machine varies quite a bit yet from angle range to angle range, and even varies a little with units of the same angle range. Ideally, with 100% devotion, the CPH should be constant. But the minuscule staff (due to lack of contributions) hasn't had time to perfect the software yet, and since the units come with varying angle ranges anyway, there is a pretty effective averaging effect. Right now, though, we've been receiving a whole lot of "little workunits" that have been giving us an excellent CPH yield. My RAC is at about 2300 which is 15 percent over normal and the best ever since the readjustment of credit values several months ago. It's probably impossible to get CPH to be constant with a given machine anyway. Plus or minus ten percent variation would probably be an excellent achievement.

One can easily calculate that by looking at his/her results and dividing the number of credits by the number of seconds, then multiplying that result by 3600. For example if my machine calculates a 60-credit unit in 9000 seconds that's (60/9000) * 3600 which is 24 credits per hour. Note that CPH with the same machine varies quite a bit yet from angle range to angle range, and even varies a little with units of the same angle range. Ideally, with 100% devotion, the CPH should be constant. But the minuscule staff (due to lack of contributions) hasn't had time to perfect the software yet, and since the units come with varying angle ranges anyway, there is a pretty effective averaging effect. Right now, though, we've been receiving a whole lot of "little workunits" that have been giving us an excellent CPH yield. My RAC is at about 2300 which is 15 percent over normal and the best ever since the readjustment of credit values several months ago. It's probably impossible to get CPH to be constant with a given machine anyway. Plus or minus ten percent variation would probably be an excellent achievement.

---

Quick question: How do The Knights Who Say 'Ni!' at http://stats.kwsn.net/ calculate the "Credit Per CPU Hour" ?

The host stats from the project have credit per CPU second, but that's in the 0.001 to 0.02 range for mosts hosts. I'd guess that stats page just multiplies by 3600.

Credit per CPU second is calculated by BOINC using a declining average like RAC, but it's quicker to reflect recent work and is only updated when credit is granted so doesn't fade with time when no work is being done.

                                                                Joe

Quick question: How do The Knights Who Say 'Ni!' at http://stats.kwsn.net/ calculate the "Credit Per CPU Hour" ?

---