Predict discovery date of the 1st 100M digit prime

[fivemack]

Originally Posted by Flatlander

I was assuming people would run 16 or 32 concurrent tests and CPUs would have huge on-die caches.

That doesn't really make sense to me; the advantage of the huge on-die cache is that the whole test fits in it, you can have all the CPUs on the die grinding away at the same FFT and not worry about memory bandwidth at all.

I would expect that within-die scalability for FFTs is pretty good, though I haven't seen any results for an FFT small enough to fit in cache on a machine with four equivalent cores on a die.

xyzzy has results on a Core2 where the 96k FFT length is 2.35ms on one core and 1.26ms on two, the 256k is 6.20ms on one and 3.31ms on two, whilst the 32768k length is 2088ms on one and 1373ms on two, so the scaling is, as expected, much better when the FFT isn't having to go off-chip

[R.D. Silverman]

Originally Posted by retina

Since we now have a 10M digit prime it is time to predict the discovery date of the first 100M digit prime.

Wild guesses are okay but it would be preferable if your guess is based upon some analysis of the various factors involved, including,

<snip>

A pointless exercize.

[retina]

Originally Posted by R.D. Silverman

A pointless exercize.

So is posting "A pointless exercize[sic]." a pointless exercise. Any anyhow, so what, I enjoy the discussion about how things will pan out in the future.

[Flatlander]

Originally Posted by fivemack

That doesn't really make sense to me; the advantage of the huge on-die cache is that the whole test fits in it, you can have all the CPUs on the die grinding away at the same FFT and not worry about memory bandwidth at all.

I was assuming that one test per core/cache would still be quicker in 2015 than sharing one test among all the cores, and that most people would choose that option.
It's only a guess guys! (In fact, I'm tempted to change my guess to an earlier date, if that's allowed.)

[xilman]

Originally Posted by R.D. Silverman

A pointless exercize.

So? It's social gossip, not scientific discussion.


Paul

[MooooMoo]

Originally Posted by R.D. Silverman

A pointless exercize.

[Uncwilly]

Originally Posted by R.D. Silverman

A pointless exercize.

No, it is not. First, advantages and benefits of a heathly "play life" are well established by the medical and psychological fields.
Secondly, the use of "play" as a safe practice in skill developement is also well documented.
Thirdly, playful thought excersizes often help to spark useful ideas and investigations.

Notice that this thread is in the "Lounge" and not in one of the Math sub-fora.

[petrw1]

On my projected retirement date posted at the "You might be addicted to GIMPS" forum.

[T.Rex]

We need a poll with about 20 years (2009...) so that anyone can vote and we can see easily the result !!
Who knows how to add a poll for 100M and for M47 ???? Do it ! Please !
Tony

[Uncwilly]

Most polls can only go to 5 options. Maybe Xyzzy could get the ghosts of the gerbils to work some magic.

[Xyzzy]

Polls have 10 options now.

Snake1: Use the power wisely. Corrupt you it will.