For some time I've been tracking my benchmark values. Is there a master list online of processors and there average values? If not, we should make one!?

[deleted by author].

HMM, That is really cool, but I was thinking more along the lines of differnt CPU types.

I had the same thought once also, but after checking many computer id's at various projects, I think the internal benchmark of BOINC is just too inaccurate...

Not sure if thats been a problem of older BOINC versions, but I do compare my AMD X2 4600 quite often with other computers of the same type and there are quite often huge differences. Maybe the problem is that BOINC doesnt tell you: Hey, I need to do a benchmark, stop/quit all other programs, dont rip a DVD or play a game, let me do the benchmark first. (Even if the benchmark runs at high priority (what I'd expect), running cpu intensive programs at the same time will tamper the benchmark results, imho - dont know if I'm correct here).

Nevertheless, I would be happy to join you creating a master list. I have my own website (to host it there etc.), love everything thats to do with numbers and especially computers. Drop a mail at "s h a i h u l u d a t b o i n c s y n e r g y d o t c o m"

Cheers, Shai

My original post was too long and too complicated with no definite answer so I decided to delete it.

I consider BOINC benchmark useful when questimating work fetch. But too inaccurate to tell much about real performance, too inaccurate to be comparative among different OS version, CPU types, too easy to fake, with a small correlation among different BOINC applications etc.

And, I miss the usage of such data in your question...

Well, we'll give it a try. The journey is the reward. If we'll get to a dead-end, so be it. Maybe its just fun for us doing it without any use. But maybe it is a bit useful, who knows yet?

The usage could be for other people to actually compare COMPARABLE benchmark results with their own, and if they're much smaller, to look WHY they are smaller. I said COMPARABLE benchmark results, we'd of course ask people to benchmark their systems under equal conditions.

This is also for you, Kevin: We could for example avoid cheated/too low benchmark results by applying a BOINC-like system. We'd only add a certain computer type, if we would have at least (lets say for example) 5 systems (quorum of 5 ) with almost equal benchmark results.

Maybe the BOINC Benchmarking will also be more accurate in the future (dont know if there's anything planned in that direction).

Cheers, Shai aka Steffen

Ive never really gotten into trying to alter my benchmarks, they are what they are. So i havent given much thought on how people can cheat. But Shai is right, it is an adventure and the larger the database the more we can learn.

There have been talks of using project-specific benchmark WUs on the dev email lists. Those might be more accurate but could also be more sensitive to individual hardware differences beyond the actual CPU, such as memory and HDD, depending on the application. Some of that data is not available so it would be harder to make comparisons between CPUs.

This is not really a collection of benchmark values but comparation to benchmark values with credit does make sense to me.

http://boinc.netsoft-online.com/e107_plugins/boinc/get_cpcs.php