PrimeGrid: World Record Fermat Divisor!

[RSS]

On 13 May 2013 12:41:57 UTC, PrimeGrid's Proth Prime Search project found a world record prime Fermat divisor: 57*2^2747499+1 Divides F(2747497) The prime is 827,082 digits long and will enter Chris Caldwell's The Largest Known Primes Database ranked 1st for prime Fermat divisors and 101st overall. Incidentally, it is also a new record for "weighted" prime Fermat divisors. The discovery was made by Marshall Bishop (Wintermute) of the United States using an AMD Quad-Core AMD Opteron(tm) Processor 2374 HE with 32 GB RAM running Linux. This computer took about 2 hours 28 minutes to complete the primality test using LLR. The prime was verified by [AF>Le_Pommier>MacGeneration.com] Sloughi of France using an Intel(R) Xeon(R) CPU E5645 @ 2.40GHz with 12 GB RAM running Mac OS X. This computer took about 4 hour and 46 minutes to complete the primality test using LLR. [AF>Le_Pommier>MacGeneration.com]Sloughi is a member of the L'Alliance Francophone team. For more details, please see the official announcement.

More...

[Mumps]

Interesting note that the Mac took twice as long to process as the Linux host.

[DrPop]

Interesting note that the Mac took twice as long to process as the Linux host.

I wonder why? Isn't Mac OSX a derivative of Linux in some way? And that XEON processor in there is no slouch. . Maybe they have HT enabled on the XEON, and it's not helping in this case?

[Fire$torm]

I wonder why? Isn't Mac OSX a derivative of Linux in some way? And that XEON processor in there is no slouch. . Maybe they have HT enabled on the XEON, and it's not helping in this case?

Actually Mac is a derivative of Unix as is Linux.

[Mumps]

I wonder why? Isn't Mac OSX a derivative of Linux in some way? And that XEON processor in there is no slouch. . Maybe they have HT enabled on the XEON, and it's not helping in this case?

Yes to everything. IIRC, Mac OS is a derivative of FreeBSD.

http://unix.stackexchange.com/questi...os-x-come-from
http://en.wikipedia.org/wiki/OS_X

Chances are good that it's the HT, as you surmised. The AMD wouldn't have that, so it wouldn't have to share cycles with other tasks. Actually, I'll also throw out there that the App simply may have optimization issues for the Mac as well. May be it wasn't compiled or ported well. Just like the Linux variant of NFS used to outperform the Windows version significantly, that kind of variation is not unheard of in cross-platform environs.

[zombie67]

Yes to everything. IIRC, Mac OS is a derivative of FreeBSD.

http://unix.stackexchange.com/questi...os-x-come-from
http://en.wikipedia.org/wiki/OS_X

Chances are good that it's the HT, as you surmised. The AMD wouldn't have that, so it wouldn't have to share cycles with other tasks. Actually, I'll also throw out there that the App simply may have optimization issues for the Mac as well. May be it wasn't compiled or ported well. Just like the Linux variant of NFS used to outperform the Windows version significantly, that kind of variation is not unheard of in cross-platform environs.

Usually the intel OSX apps are equal to or faster than linux or windows apps.

For linux and windows apps, they have to work with older CPUs such as pentium (the original kind), so the apps cannot require some modern instruction sets. But Intel Macs have no legacy further back than Core. That means all Intel OSX apps support SSE3 at a minimum.

Of course, this does not take into account which compiler is used, and if the apps are written to take advantage of the available instruction sets. But PG uses the intel compiler, if I recall correctly.