Not considering the benchmarks, or the whole credit issue, what OS allows for the absolute fastest crunching? (credit is nice, But I'm in it for the true results.)

---

<img src="http://teamstarfire.org/boinc/summary.php?name=Dwarlock&amp;team=Kansans%20Searching%20for%20a%20Planet%20With%20People%20More%20Interesting%20Than%20This%20One&amp;seti=7832952&amp;cbg=grey">

A Cray should do the job nicely!
Seriously though....any machine is okay just don't use it for anything else. The more you use it for other things the less cycles Boinc has for its use.
Just buy as fast as you can afford and set it up with the minimum of accessories.
I just got 3 new setups, 2 AMD 2900+ and an AMD 2800+, I build my own so I buy just the board and chip and then drop them into my old boxes, upgrading whatever is currently in there. I am up to a slowest cpu of 1.4ghz now. I have a couple of Celerons and a couple of Durons still to replace. The new boards and cpus cost me under $150.00US each. Got the latest at newegg.com.
Oh, I do have an OLLLLD 300mhz dual cpu server to replace too. I think the new 2800+ will replace it. Maybe get a big enough box to drop some extra hard drives in and load a Server OS on it.

---

> Not considering the benchmarks, or the whole credit issue, what OS allows for
> the absolute fastest crunching? (credit is nice, But I'm in it for the true
> results.)

The compiler is the major difference between different O/S' as far as BOINC and projects are concerned. The os affects slighly the file read/write speed and how often CPU cycles are borrowed for OS tasks...but for BOINC these are very minor.

For Windows it is MS VC++ 7.x and for any *nix (Linux, Unix, Xenix, etc) it is GCC 3.x+ (GNU C++ Compiler).

For BOINC you can compile your own, but it doesn't affect your throughput much, just changes your benchmark scores.

For Science Apps, right now all you can get access to to compile is Seti. The other project sources are proprietary.

---

I am running Linux on an AMD. I think there is not much difference in speed with what I see reported for similar hardware running Windows XP or 2000. The benchmarks differ a lot (but boinc manager compiled with better optimisations for linux corrects that by and large; also 4.25 under windows gives a more realistic result).

HOWEVER, I recently added Einstein as second project only to find out after a few weeks that this app runs much slower under Windows than under Linux. So OS and architecture have their influence but to very varying degrees

---

While there is some room for error, I created this page to answer that question for myself.

Feel free to check it out. It's not perfect but will give a good idea I think.

http://www.tlmartin.com/boincstats/

> Not considering the benchmarks, or the whole credit issue, what OS allows for
> the absolute fastest crunching? (credit is nice, But I'm in it for the true
> results.)
>

---

Using the default science apps (seti) from berkeley, my experience is that native Windows versions crunch about 10-20% faster than native Linux versions. The same was true for seti classic.

From my previous tests with seti classic, I found Windows NT4 to be very slightly faster than Win2K or WinXP, but the difference is negligable (maybe 1-3% max). Whether that still holds true for boinc versions, I don't know.

Simple answer, the Windows NT/W2K/XP line of OSes are fastest for pure seti crunching. But as Ben correctly pointed out, this has more to do with the compiler used to compile the applications than the actual OS itself.

Ned

---

*** My Guide to Compiling Optimised BOINC and SETI Clients ***
*** Download Optimised BOINC and SETI Clients for Linux Here ***

> Not considering the benchmarks, or the whole credit issue, what OS allows for
> the absolute fastest crunching? (credit is nice, But I'm in it for the true
> results.)

MS-DOS 2.11

Hi

I hate to say this, but you should go for Windows, especially if you want to
participate in multiple projects.

The LHC and Einstein linux clients are very slow. CPDN is OK on linux,
seti runs OK if you compile and optimize it by yourself.

Regards Hans

---