Weird computer record on file
11 Jan 2011 7:17:39 UTC
Topic 195569
(moderation:
I'm starting to do some statistical analysis of the computers on file. When I get something interesting and possible valid I'll share them. But I ran across a computer's record that ..well.. can't be right.
Please take a look at computer 388,874.
Do you think there are many with invalid benchmark results?
Joe
Weird computer record on file
)
There usually are. Last time I looked it were around 5%. The benchmark isn't reliable, in particular since the client is open source and prone to manipulation of any kind. That's why we don't rely on benchmarks for granting credit.
BM
BM
RE: There usually are. Last
)
Thanks Bernd.
I'm familiar with some of the problems with benchmarks even when they are controlled as much as possible.
I'm hoping the law of large numbers allows me to see something interesting in the benchmarks but real credit and cpu seconds is perhaps better, especially if it can be calculated per application. In my limited experience, credit/cpu sec is not very constant on a single machine between applications.
My first dive into the data is looking at how families of processors compare with each other.
Regardless, I find it fun to look at numbers like these. As some anonymous philosopher once said "God is in the details too".
Joe