Triple test of m2 to m499979

[Eivind]

Hi

I wantet to see how fast a standard coporation desktop computer is compared to the early mersenne prime hunting, so i did a search from m2 to m499979.

The list of primes in the worktodo file was generated from a home grown code, but tested against the "official" list from primes.utm.edu. A ",70" was then added to all lines to avoid LL testing.

The result file (ZIP compressed) is 1,2MB so it exides the max limit for upload. Is any one intersted in the file? Should I mail it to George?


-Eivind

[tha]

Can you be more specific about what exactly you did? Which exponents did you test? All exponents in that range that have no known factor? And what test did you do on those exponents? a Lucas-Lehmer test?

[Prime95]

Originally Posted by Eivind

The result file (ZIP compressed) is 1,2MB so it exides the max limit for upload. Is any one intersted in the file? Should I mail it to George?

Yes. I'll add it to the database as triple and quadruple checks.

[cheesehead]

Originally Posted by Eivind

A ",70" was then added to all lines to avoid LL testing.

I presume you mean that the ",70" was added to avoid trial factoring (by indicating that TF had already been performed unsuccessfully up to 2⁷⁰ for each exponent), so that only the LL tests were performed.

[Eivind]

Hi

Which exponents did you test?

All primes from 2 til 499979.

The times was about 20 sec/candidate in the beginning and 315 sec/candidate in the end.

presume you mean that the ",70" was added to avoid trial factoring (by indicating that TF had already been performed unsuccessfully up to 2^70 for each exponent), so that only the LL tests were performed.

Your right - i mixed up the FT/LL, sorry. I seems like the FT took verry long time for the first candidats. I think Primes95 has some time comsuming initiationg for FT so I just forced the program to jump right to the LL test.


My conclusion is that a standard desktop computer with Primes95 outpreforms the stade of the art hardware 13-15 years agao (Slowinski, Cray-2, 1992-1996). If I continued I would find the m34 (1257787) faster than they did

The Cray-2 was supassed in 1990 so in 1992 this was still a verry powerfull machine.

So by the time my doughter turns 18 I might have a BlueGene/L or a Red Storm sitting on my desk. Jum jum I wonder, if it will be optical?

-Eivind

[davar55]

All primes from 2 til 499979.
The times was about 20 sec/candidate in the beginning and 315 sec/candidate in the end.
Your right - i mixed up the FT/LL, sorry. I seems like the FT took verry long time for the first candidats. I think Primes95 has some time comsuming initiationg for FT so I just forced the program to jump right to the LL test.
My conclusion is that a standard desktop computer with Primes95 outpreforms the stade of the art hardware 13-15 years agao (Slowinski, Cray-2, 1992-1996). If I continued I would find the m34 (1257787) faster than they did.
The Cray-2 was supassed in 1990 so in 1992 this was still a verry powerfull machine.
So by the time my doughter turns 18 I might have a BlueGene/L or a Red Storm sitting on my desk. Jum jum I wonder, if it will be optical?

Can someone do an overview historical comparison of PCs that would
tell us just how much P95 runs faster over the years 1995-2011 ?

[sdbardwick]

Originally Posted by davar55

Can someone do an overview historical comparison of PCs that would
tell us just how much P95 runs faster over the years 1995-2011 ?

Quick perusal of benchmarks gives data to about 1997 era processor (Pentium 120).

1024K FFT
P120 5167.51 vs. i7 x980@4050 10.44

TF65
788.08 vs. <3.41

[Uncwilly]

As of this time GIMPS "Competition" has quad checked all numbers up to 465133 @ ~8750 sec per LL.

[mdettweiler]

Originally Posted by Uncwilly

As of this time GIMPS "Competition" has quad checked all numbers up to 465133 @ ~8750 sec per LL.

On an interesting side note (which may or may not have been noted here before), it seems that he has now become aware of GIMPS if he wasn't before:

Not only are computers faster, but there's more of them. There are literally billions of machines out there, and many of these are mostly idle. This leads to some interesting approaches to computing at the bleeding edge.

For example... if you find this website just a bit creepy or just a bit too geeky, you'll be pleased to learn that this stuff's small time. As you read this there are tens of thousands of people looking for Mersenne primes, and hundreds of thousands of processors working on the task.

Have a quick look at the GIMPS Project (Great Internet Mersenne Prime Search) for an example of massive community computing at its best.

[davar55]

His compliment to GIMPS leads to this at the moment:

M(20996011) proven to be 40th Mersenne Prime   July 11, 2010.  Almost 7 years after discovering M(20996011), GIMPS finished double-checking every smaller Mersenne number.  Since no primes were found, the "40th known Mersenne prime" can now be simply called the "40th Mersenne prime".  Congratulations and thanks to all the GIMPS members that perform this less glamorous double-checking work.

Nice way to find a press release.

[cheesehead]

Originally Posted by mdettweiler

On an interesting side note (which may or may not have been noted here before), it seems that he has now become aware of GIMPS if he wasn't before:

Interesting alternative to the term "distributed computing": "massive community computing".

I like it. It sounds less impersonal. "Community" captures the sense of interaction between human participants, such as in this very forum. It doesn't define participation in terms of only the hardware/software.

Perhaps the "massive" will need to go. What's "massive" today will be "piddling" tomorrow. OTOH, as long as GIMPS is comparable to the Top 50 Supercomputer list members ...