Sorry briefer version. Lost 1st posting due to
user error. :-(

I have two computers
1 - 187636 running Linux
2 - 430400 running MS XP Pro (two processors)

Pretty routinely, computer one shows a granted credit
that is 180-190% of "claimed credit". Computer 2
just as routinely shows a granted credit that is only
80-90% of "claimed credit".

I am a little confused by this and was wondering if
anyone could explain.

Mike

---

The Boinc benchmarks under Linux give results that are about 50% of the same hardware running the Boinc benchmarks for Windows.

The granted credit approach was Rosetta's attempt to even the playing field for AMD vs Intel, Linux vs OS X vs Windows. And while it's not perfect, it balances out over time. :)

---

Rather then comparing the credit per task and the credit claim vs. the granted credit... if you instead compare the credit each machine earns per day (or per hour), you will start to see the wisdom of the credit system used by Rosetta.

Compare:
https://boinc.bakerlab.org/rosetta/results.php?hostid=430400 your host
https://boinc.bakerlab.org/rosetta/results.php?hostid=187636 friend's host

You will notice that both of you are crunching tasks for about 28,000 seconds (8hrs). But your host, with fast CPU and more memory, is earning roughly 70 credits per task, and your friend is only earning about 48.

You both put 8hours of work in to Rosetta, but in terms of models completed and actual computer power expended, you contributed more. This is reflected in your higher RAC and higher credits granted.

---

Rosetta Moderator: Mod.Sense

"The Boinc benchmarks under Linux give results that are about 50% of the same hardware running the Boinc benchmarks for Windows."

Given this statement, does a Linux machine actually preform slower than the exact same machine running whatever flavor of Windows?

---

"The Boinc benchmarks under Linux give results that are about 50% of the same hardware running the Boinc benchmarks for Windows."

Given this statement, does a Linux machine actually preform slower than the exact same machine running whatever flavor of Windows?

No. The benchmark in BOINC is based on a very old public domain benchmark. Modern compilers with the right optimizing switches can see that the benchmark's loops aren't doing anything and will remove those loops. This can produce very high benchmark numbers.

The windows version of BOINC is compiled with switches that remove more of the benchmark than the Linux version, thus it produces much higher numbers.

This kind of Wu has very low Granted Credit.

Anders n

---

this work unit cost me alot of points

---