Treading LL P-1 for 100M TF - Page 12

Uncwilly · 2013-04-23, 05:13

Quote:

Originally Posted by Aramis Wyler

You tried to do a p-1 on a 332M number once, didn't you Unc? How long did that take?

166 GHz-days, it was on a single core of a i5-2400 @ 3.10GHz running Windows64,v27.7 with about 2600 MB available to P95.

Aramis Wyler · 2013-04-23, 05:29

Quote:

Originally Posted by Uncwilly

166 GHz-days, it was on a single core of a i5-2400 @ 3.10GHz running Windows64,v27.7 with about 2600 MB available to P95.

So... about 6 weeks?

Uncwilly · 2013-04-23, 05:29

Quote:

Originally Posted by davieddy

What is your disability?

As previously brought out in the forum, dyslexia. It affects spelling (not just the random reversals, but also induces a general difficulty in spelling) among other things.
It seems that you know about this. This exchange may have been a clue.

Uncwilly · 2013-04-23, 05:37

Quote:

Originally Posted by Aramis Wyler

So... about 6 weeks?

I think that it may have been on the shorter side of that a touch. I did 2 and they got turned in Feb 20 and Mar 20. IIRC they were on the same core sequentially. I think that P-1 gives the best GHz-days credit return, as a function of actual GHz-days outlay.

chalsall · 2013-04-23, 14:42

Quote:

Originally Posted by Aramis Wyler

I suspect there there is a period between when you turn your TF'd number in to gpu72 and when gpu72 releases it.

Only a few seconds.

Quote:

Originally Posted by Aramis Wyler

You could probably use the 'highest TF level' with a min of 332M to find the one you had just turned in to 79, reserve it to 80, then p-1 the thing for primenet. When you're done with the p-1, you can simply unreserve the tf to 80.

Until we make 100M digit P-1 an official work type, the best thing to do is simply e-mail me the candidate desired, and I can manually set up the assignment.

Or, alternatively, reserve it directly from Primenet and e-mail me the candidate, and I'll set things up such that GPU72 will "watch" for its P-1 completion.

Either way works, and wouldn't take much (of my) time.

chalsall · 2013-04-23, 17:36

Quote:

Originally Posted by Aramis Wyler

I suspect there there is a period between when you turn your TF'd number in to gpu72 and when gpu72 releases it.

OK... I've figured out an easy way to facilitate this... GPU72 will now keep the 10 lowest candidates without a P-1 done.

This means two things. First, the lowest candidates (without P-1 done) are available to take to higher bit levels (read: they won't automatically get released just because they've been TFed past 73).

Second, it means that those low candidates taken to higher bit levels are available to be reserved for P-1 work.

For the record, I don't think 100M digit P-1 Makes Sense^(TM) when we're currently behind on LL P-1. But I'm just a facilitator. As always, people are free to do whatever they want with their own time, equipment and money.

kladner · 2013-04-23, 20:42

Quote:

Originally Posted by Uncwilly

I think that it may have been on the shorter side of that a touch. I did 2 and they got turned in Feb 20 and Mar 20. IIRC they were on the same core sequentially. I think that P-1 gives the best GHz-days credit return, as a function of actual GHz-days outlay.

That is encouraging.....much better than I expected. It would definitely be worth dedicating a core, even if that core is on an AMD FX instead of an i5.

Aramis Wyler · 2013-04-24, 11:38

There were no factors to 79 for 332283841 or 332283859. Puts the GHD/10 on my side up to 254.

Uncwilly · 2013-04-24, 12:31

Quote:

Originally Posted by Aramis Wyler

There were no factors to 79 for 332283841 or 332283859. Puts the GHD/10 on my side up to 254.

Cool! I just hit 246 GHz-days a moment ago.

bcp19 · 2013-04-24, 19:33

Quote:

Originally Posted by davieddy

<snip>
If it just worthwhile TFing 66M to 75, it is clearly desirable to go to 74.
It is possible with our current firepower to do this for most candidates and keep pace with LL completions. Any left over should be taken to 73. <snip>

Quote:

Originally Posted by davieddy

This so petty, but:
I said that we couldn't TF every expo to 74 and keep pace with the rate of LL completion. This remains true ATM.

<snip>David

I is confuzzled... with our current firepower we's can does MOST candidates to 74... but we's been tolds that 74 is not sustainable... MOST seems to be good in my book, is we mising something?

Quote:

I agree with BCP19 that dabbling with 100M digits is OK in small doses, as long as it isn't diverting a significant amount of firepower from the main goal of finding another prime.<snip>
David

I'se even more confuzzled... I'se doesn't 'member making this statement...

Quote:

Originally Posted by davieddy

Since you could take 25 66M exponents to 74 in the same time as one 332M exponent to 81, I am relieved to hear that there are only a few idiots (mostly people round here) who are squandering their massive iron on this currently futile exercise.

The chance of finding a prime per year by LLing 66M expos is 125 times greater than LLing 332M expos.

WAKE UP BCP, KLADNER, CHALSALL and the rest of you non-believers.
Beware of false prophets.

David

I'se doesn't think I'se can get any more confuzzled... I'se don't see any 332M assignments on any of my 'puters or GPUs. I'se done 1 332M since starting GIMPS, and I'se only did TF on it to 80 after the P-1 and concurrently with the LL run since I'se could GPU the TF from 77-80 in 3 months on the GPU I'se used, just in case I'se could save 8+ months of LL testing.

Quote:

Originally Posted by davieddy

These days I have single Core2 which suits my current needs. When I worked in IT I had access to a lot of things. I run first time LL 24/7, which puts me at ~1500 on the list to which you so tactfully linked. As such, I can feel that I am contributing sincerely to GIMPS, and empathize with the legions of run-of-the-mill participants.<snip>
D

Wowzers, I'se wrong, I'se can be more confuzzled... but since I'se not one who likes to repeat I'self, I'se gonna end up staying confuzzled.

Aramis Wyler · 2013-04-24, 19:52

Quote:

Originally Posted by chalsall

OK... I've figured out an easy way to facilitate this... GPU72 will now keep the 10 lowest candidates without a P-1 done.

This means two things. First, the lowest candidates (without P-1 done) are available to take to higher bit levels (read: they won't automatically get released just because they've been TFed past 73).

Did this go into play? I see the 2 numbers I put to 79 last night are not available for p-1.

Also! Me, I think I'd put off doing a month long p-1 on a cpu atm, what with the p-1 culu mod making progress.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
To 100M to 75 Bits	petrw1	GPU to 72	6	2017-04-17 04:03
Is LMH > 100M still alive?	TheMawn	LMH > 100M	25	2016-04-18 16:11
100m p-1 and tf	aurashift	Software	18	2016-04-14 13:48
Anyone working in 79.3M - 100M ??	markr	Lone Mersenne Hunters	21	2008-12-21 16:02
100M madness	stars10250	Hardware	8	2008-10-02 15:21

2013-04-24, 11:38	#129
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³×53 Posts	There were no factors to 79 for 332283841 or 332283859. Puts the GHD/10 on my side up to 254.