Newer milestone thread - Page 71

davieddy · 2013-03-05, 00:12

Quote:

Originally Posted by Uncwilly

The rates of completion of LL tests per day have been trending up since a low point around Dec 2010 - Jan 2011. This is the orange line on the attached graph, scale to left (the scale is tests completed per day). This is for the tradition range to 79.3M. Most of the data points are 7 - 10 days apart.

The raw totals at the bottom (which you are plotting) need careful interpretaion to get the DC and LL rates of completion:

The DC rate is the total + the number of DCs factored.

The LL rate is the total + the DCs + the single LLs factored.

Finally, the most telling statistic is the rate at which the expected new primes is decreasing.

David

ckdo · 2013-03-26, 06:37

Quote:

Originally Posted by ckdo

February 22, 2013: The trial factoring wavefront hits 85M.
February 26, 2013: The trial factoring wavefront hits 86M.
March 2, 2013: The trial factoring wavefront hits 87M.

That's 14 days for 3M.

March 6, 2013: The trial factoring wavefront hits 89M.
March 25, 2013: The trial factoring wavefront hits 91M.

That's 23 days for 4M.

davieddy · 2013-03-26, 07:32

Quote:

Originally Posted by ckdo

March 6, 2013: The trial factoring wavefront hits 89M.
March 25, 2013: The trial factoring wavefront hits 91M.

That's 23 days for 4M.

What significance has this to anything?

ckdo · 2013-03-26, 11:18

Whether you consider it significant that all PrimeNet organized TF will currently assign exponents that are not likely to receive an LL test within the next 15 years or some such is certainly up to you - I consider it noteworthy, at least.

TObject · 2013-03-26, 13:16

I think now would be a good time for PrimeNet to start treating manual TF assignments as GPU work and hand out higher levels (up to 74 perhaps). The lower bit lewels would be 'saved' for, presumably, CPU prime95 assignments.

Until then, I urge GPU users to volonterely upgrade their TF assignment bit levels.

petrw1 · 2013-03-26, 15:09

Quote:

Originally Posted by ckdo

March 6, 2013: The trial factoring wavefront hits 89M.
March 25, 2013: The trial factoring wavefront hits 91M.

That's 23 days for 4M.

With a disclaimer.
Can I assume you meant 87 to 91 to get your reported 4M.
88M was complete to the required limits long before we got there
And 90M is all reserved by someone.

ckdo · 2013-03-26, 16:22

No doubt about any of that. All ranges had prior work done before the wave got there. 88M happened to be complete, 90M is like 40% done and all assigned to ... me.

davieddy · 2013-03-26, 23:39

Quote:

Originally Posted by ckdo

Whether you consider it significant that all PrimeNet organized TF will currently assign exponents that are not likely to receive an LL test within the next 15 years or some such is certainly up to you - I consider it noteworthy, at least.

Currently, new LL tests are being dished out at 62M.
We currently expect the next prime discovery when this "wavefront" hits
90M in 4 years time.
ATM we are trying to ensure new LLs are TFed to 74 bits. By 90M, this will be 75 or 76 bits.

HOW MUCH IS TFing to 70 bits contributing to this effort, aside from stealing all the low-lying fruit factor finds?

David

chalsall · 2013-03-27, 00:01

Quote:

Originally Posted by davieddy

HOW MUCH IS TFing to 70 bits contributing to this effort, aside from stealing all the low-lying fruit factor finds?

Earth to David: Not your concern!

You're not donating the time, hardware, energy and associated money involved, so you have no say.

davieddy · 2013-03-27, 00:37

Quote:

Originally Posted by chalsall

Earth to David: Not your concern!

You're not donating the time, hardware, energy and associated money involved, so you have no say.

Au Contraire.
There's nothing for me to do EXCEPT shoot my mouth off, having engaged my brain first.

Aramis Wyler · 2013-03-27, 02:32

Honestly I'm surprised primenet still gives out TF work at all, while gputo72 is keeping up.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Newer X64 build needed	Googulator	Msieve	73	2020-08-30 07:47
Performance of cuda-ecm on newer hardware?	fivemack	GMP-ECM	14	2015-02-12 20:10
Cause this don't belong in the milestone thread	bcp19	Data	30	2012-09-08 15:09
Newer msieves are slow on Core i7	mklasson	Msieve	9	2009-02-18 12:58
Use of large memory pages possible with newer linux kernels	Dresdenboy	Software	3	2003-12-08 14:47

2013-03-26, 11:18	#774
ckdo Dec 2007 Cleves, Germany 2×5×53 Posts	Whether you consider it significant that all PrimeNet organized TF will currently assign exponents that are not likely to receive an LL test within the next 15 years or some such is certainly up to you - I consider it noteworthy, at least.

2013-03-26, 13:16	#775
TObject Feb 2012 3⁴×5 Posts	I think now would be a good time for PrimeNet to start treating manual TF assignments as GPU work and hand out higher levels (up to 74 perhaps). The lower bit lewels would be 'saved' for, presumably, CPU prime95 assignments. Until then, I urge GPU users to volonterely upgrade their TF assignment bit levels.

2013-03-26, 16:22	#777
ckdo Dec 2007 Cleves, Germany 1000010010₂ Posts	No doubt about any of that. All ranges had prior work done before the wave got there. 88M happened to be complete, 90M is like 40% done and all assigned to ... me.

2013-03-27, 02:32	#781
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³×53 Posts	Honestly I'm surprised primenet still gives out TF work at all, while gputo72 is keeping up.