You'll have to excuse me as I'm relatively new to BOINC. Am I correct in my understanding that there is no way to acurately say which CPU/GPU is better for BOINC like with F@H due to the way that each different project uses the hardware? So without knowing which projects someone runs you won't be able to really say this is better than that etc.

I've been a long time folder and recently found BOINC. I had been working on a dedicated F@H computer and have decided I would rather put that towards several projects with BOINC. I had a CPU/GPU's in mind for when it was going to be dedicated to F@H but now I'm kind of at a loss as I haven't really been able to find anyplace that outlines how different CPU's and GPU's will perform with BOINC like they have with F@H.

Any help is appreciated.

Until someone hopefully comes up with a better answer, it might be worthwhile to have a look at http://boinc.berkeley.edu/wiki/GPU_computing

Or perhaps ask in the Number Crunching forums of those projects that interest you.

---

JOIN Team ACC

I've done some more looking and have another question. On the BOINC stats page you can look at the Host CPU breakdown for each project. Which column is the most important when determining how well those CPU's perform with that project?

Would I look at Average credit per CPU? This doesn't necessarily make sense because I could have one CPU that I have had running that project for a year and it has 10k points and another that has been running the same project for a month but only has 2k points. The one with 2k points has a lower average but will actually produce more than twice the results.

Or should I be looking at the average credit per CPU second column? This to me shows how much work you are actually getting done per second with the CPU and would mean that the CPU with the highest number here is producing the greatest results.

Perhaps the answer should be: None of the above.
For instance, with climateprediction.net, there are (or potentially are), several different model types. Each of these is tested over several months and only released for the 3 different platforms (Windows, Linux, Mac), after they have been made to produce stable, reliable, results.
During this process, both code and compiler optimisations are made, with the end result that some are faster on Linux than Windows, and others are the other way around.
So just looking at a cpu in a list won't tell you the full story.

Perhaps you should look at all of the different types of projects that are available, (medical, space, geo-physical, etc), decide what interests you the most, and then look at the web sites of some in that category to see how different processors handle the science apps.

Boinc is being used by 50+ projects at different points in their development. Depending on the requirements and the type of app and its optimization, performance will vary greatly. Even within some projects there are sub-projects each with its own needs. Boinc GPU implementation is still in its early stage and evolving quickly with project just starting to take advantage of this new computing power. Les Bayliss' suggestion is probably the best. Look for the science that holds the greatest interest for you and see what kind of fit they are for you and your hardware.

---

Boinc V 7.4.36
Win7 i5 3.33G 4GB NVidia 470