As far as I know there is no appreciable performance difference. While there may be others that will argue the point, I would suggest there are other factors more important than BOINC performance in choosing an OpSys. Other use of the comptuer and cost being two of the primary ones.

Hi Orakk,
Which is faster - Linux or Windows? Well, Linux is the more efficient OS, but that has close to zero effect on the science application. What really matters is the efficiency of the compiler used, which is something that only Rick Alther has the knowledge to speak about.

Of course, lack of knowledge never stops me from coming out with an opinion. Different compilers vary enormously in optimization efficiency, but that probably matters very little to us. The reason is that the FPU uses x87 instructions that keep intermediate floating point results in 80-bit format, while the SSE2 instructions use 64-bit format all the time. This results in small differences (usually) which only matter if you are doing a bit-by-bit comparison of the result. But that is what we do when we check validity. Therefore the science application has to be compiled with only a few optimizations enabled so that all the results are the same, no matter which computer they are computed on. In general, much more money has been spent developing optimizing compilers for Windows than for Linux, but without full optimization the differences in speed should be very small. So my final guess is that the Windows version will be slightly faster than the Linux version, even though I think that the Linux OS is faster than the Windows OS.

I think that we are talking about very small differences here. Use whichever OS you prefer.

Ummm. . . Linux is cheaper, isn't it?

Lawrence

Software $ is an issue, so my 3rd puter is a pure linux box so need to buy windows etc. Apart from that i think they are pretty much the same, i now run boinc on my 2 windows xp machines and boinc on my linux box, i have not noticed any significant difference.

I think the next question becomes boinc vs ud frontend.

At present i would argue that boinc is more efficient in the way work units are being sent out, ie quorum of 3 with 4th w/u sent only if error or long time delay, asfaik ( as far as i know). it seems to me that most work is being done by UD and maybe they might change there w/u allocation routine ? i think they are using the boinc "front end" Once an optimum strategy is reached then i would like to think that the entire project would run around the same efficincy, boinc might have a slight edge if it sends for next w/u just before the returning its currnet w/u but that is probably small bickys in the grand scheme. Of course the software $ is another issue, so my 3rd puter is a pure linux box so need to buy windows etc.