PDA

View Full Version : Moore's law and billion digit primes



jasong
06-01-2006, 11:05 PM
I was over on the Mersenne website, playing with the benchmark thingie for Prime95. Apparently it would take the computer I'm about to get in the mail 745 years to process a billion digit number. On the other hand, in about 24 years, factoring in 3.5% inflation, the same valued computer will take a little over 2 months.

Ain't Moore's Law grand? :looney:

Digital Parasite
06-02-2006, 11:09 AM
Not exactly. Moores law just refers to transistor count, not speed. So now that Intel and AMD are moving to creating multiple-cores with those extra transistors instead of creating faster speed CPUs, your Prime95 program will not get any faster, you will just be able to check multiple numbers at the same time.

If you look at the multi-threaded version of GLUCAS which does LL tests like Prime95, on current high-end hardware, the speed maxes out around 16 CPUs.

So what we really need is a new algorithm or much better implementation of it.

jasong
06-02-2006, 10:11 PM
Not exactly. Moores law just refers to transistor count, not speed. So now that Intel and AMD are moving to creating multiple-cores with those extra transistors instead of creating faster speed CPUs, your Prime95 program will not get any faster, you will just be able to check multiple numbers at the same time.

If you look at the multi-threaded version of GLUCAS which does LL tests like Prime95, on current high-end hardware, the speed maxes out around 16 CPUs.

So what we really need is a new algorithm or much better implementation of it.
You're probably right, but, honestly, I wasn't expecting to be taken seriously. I was simply bored and decided to play with my calculator a bit. Actually, all I really "discovered" was that if you take 745.5 years and divide it by 2^12, you get a little over 2 months.

riptide
07-26-2006, 01:03 PM
Not exactly. Moores law just refers to transistor count, not speed. So now that Intel and AMD are moving to creating multiple-cores with those extra transistors instead of creating faster speed CPUs, your Prime95 program will not get any faster, you will just be able to check multiple numbers at the same time.

If you look at the multi-threaded version of GLUCAS which does LL tests like Prime95, on current high-end hardware, the speed maxes out around 16 CPUs.

So what we really need is a new algorithm or much better implementation of it.
Moore was specifically talking about cost.... always the business man. People then derive what he meant by it afterwards :)

"The complexity for minimum component costs has increased at a rate of roughly a factor of two per year ... Certainly over the short term this rate can be expected to continue, if not to increase. Over the longer term, the rate of increase is a bit more uncertain, although there is no reason to believe it will not remain nearly constant for at least 10 years. That means by 1975, the number of components per integrated circuit for minimum cost will be 65,000. I believe that such a large circuit can be built on a single wafer." - Gordan Moore, 1965.

Spuer Pi on a Conroe at 2.66 Gig will outoperform a higher clocked Intel Exreme Edition P4. Its all about IPS (Instructions per second).

Digital Parasite
07-26-2006, 02:21 PM
Moore was specifically talking about cost.... always the business man. People then derive what he meant by it afterwards :)

Yes, cost is latin for transistor... :rock: :looney: