Milkyway: claimed on average 55.60, granted 213.76 credits per task.
Collatz: claimed on average 5.70, granted 650-800 credits per task.

BOINC = Because Our Insanity is Numbers of Credit

---

If I were to run a MW on my cpu it would take 4 or more hours. for a credit of 213.76 this compares completely with seti where I get 50-60 credits/ hour/core. same with Collatz, averaging about 700 credits but it would take a good 12-14 hours to complete on my CPU. again about 50-60 credits/hour

You're comparing credits per hour within the same project. What are the average credits per hour compared to SETI@Home? How does it compare to the fictional "golden" system that all projects are supposed to tie their credits per hour to?

the deal is that most people are just running their ATI cards on COllatz and MW and getting the massive speed benefit from it. Gipsel and others have done an excellent job in their code writing which makes the GPU about 11,000X faster than any CPU.

I assume that the seti project would see similar numbers as well.

I'm not convinced. I think people are running Collatz and MW simply because they're in it for the credits, but many are not going to outright say it.

The thing is, even with the speed-up, if the credits per hour is higher than all others, then the credit should be adjusted for such hardware so that it earns the same as all else.

---

I get about 50-60 credits/hour/core on seti as well with my AMD PII 940 OCed to 3.5GHZ. the numbers are pretty much the same. when comparing CPU time to CPU time from seti to MW or COllatz.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

But the GPUs earn far more than 50-60 credit per hour, correct? Isn't that how a user can obtain more than a million credits in a couple months? If so, then GPU credits are too high and need to be adjusted. The same should technically be done with CUDA as well, but at least there's a safeguard preventing it from becoming too large an issue (i.e. being paired up with a CPU partner).

---

the gpu WU's are identical to the CPU WU's for both MW and Collatz. So the credit assigned is also identical. Since the WU's are all the same size and crunch the same amount of data they assign virtually the same credit/WU. Seti works with the flop count and it appears that they assign the same way as well. The only thing that happens differently is that the CUDA GPU doesnt count the flops the same as the CPU which it where the discrepancy occurs.

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope

But the GPUs earn far more than 50-60 credit per hour, correct? Isn't that how a user can obtain more than a million credits in a couple months? If so, then GPU credits are too high and need to be adjusted. The same should technically be done with CUDA as well, but at least there's a safeguard preventing it from becoming too large an issue (i.e. being paired up with a CPU partner).

The mantra is "Credits should reflect the work done". I have never run MW or Collatz on my CPU so I was not aware how close the comparison was with S@H. Here on S@H the aim is to give the same credits for a WU whether it runs on a CPU or a GPU and whether using stock or op app. It would appear that the same is happening at MW but because of the nature of the tasks there just happens to be a much greater time disparity between the CPU and GPU crunch times. Note that MW does not have the variety of tasks that we have here at S@H (i.e different AR's) - all tasks are pretty well identical and there is a fixed credit per WU crunched regardless of whether on a CPU or a GPU.

F.

---

thats what I was attempting to say but not so eloquently

---

In a rich man's house there is no place to spit but his face.
Diogenes Of Sinope