98.00M to 98.05M Excluded Values - Page 4

storm5510 · 2009-08-21, 02:17

A basic question: I'm doing this on a Core 2 Duo at 3.0 GHz. Is this the right type of work? Yes or No?

Something tells me I should be doing something more difficult than a lot of machines can do, like my laptop, for example.

cheesehead · 2009-08-21, 03:17

Quote:

Originally Posted by storm5510

A basic question: I'm doing this on a Core 2 Duo at 3.0 GHz. Is this the right type of work? Yes or No?

1) If it's the type of work you prefer to do, then it's the right type of work regardless of what anyone else says. Period.

2) If by "right type of work", you mean that you would be happiest doing the type of work for which your computer is deemed most efficient by GIMPS standards, then:

There's a place where you can select the system most like yours, and then see the relative efficiency (GIMPS throughput) of that system at doing various work types.

Here it is: http://mersenne-aries.sili.net/throughput.php

I selected "Intel(R) Core(TM)2 Duo CPU E8400 @ 3.00 GHz (3.0 GHz)", then I entered your system's speed as "3000" MHz (it's not GHz for that one!), clicked "Go", and got tables of relative efficiencies of various types of work at various FFT lengths and TF bit levels.

Result for your system:

Any TF throughput is noticeably higher than any FFT (L-L, P-1, ECM) work type's throughput.

You're already doing the type that produces the most GIMPS throughput on your system.

Sunfish · 2009-08-21, 08:37

Quote:

Originally Posted by storm5510

A basic question: I'm doing this on a Core 2 Duo at 3.0 GHz. Is this the right type of work? Yes or No?

Simple question with various aspects.

If you want to produce the highest possible throughput then you have to do nothing but TF to 2^63 on a 64-bit system. There is nothing that can beat it! If you follow the link provided by cheesehead you could take a look at the Core 2 Duo T7700 as an example to see the huge difference compared to any other type of work.

If you want to do what is needed most by the GIMPS project then you should dedicate one of your cores to P-1 work. P-1 is the bottleneck whereas TF work is heavily overpowered. P-1 work is the most memory intensive one and you'll need to allow Prime95 to use around 500MB RAM to select suitable B1 and B2 bounds.

If you want to have the chance of finding a prime you have to do LL testing, of course. Personally, I think that this is the most thrilling one and therefore the one I use most.

Your machine is definitely fast enough for any of those work types so the choice is yours.

storm5510 · 2009-08-22, 02:43

I attempted to transfer the LMH exponents to my old computer. It didn't fly for some reason. They were unreserved.

I'm doing P-1 based on multiple suggestions Finding a new prime doesn't much interest me, for some reason. I would rather contribute, quietly, in the background.

10metreh · 2009-08-22, 07:19

Quote:

Originally Posted by storm5510

I attempted to transfer the LMH exponents to my old computer. It didn't fly for some reason. They were unreserved.

I'm doing P-1 based on multiple suggestions Finding a new prime doesn't much interest me, for some reason. I would rather contribute, quietly, in the background.

If you feel you're not doing enough, there are always other projects to do some work for (including some that don't try to find large primes at all).

garo · 2009-08-22, 12:55

He never said he felt he wasn't doing enough. 10metreh, will you stop steering people towards "other" projects? P-1 seems just right for his hardware and GIMPS needs more P-1 testers too.

storm5510 · 2009-08-22, 16:46

I've done 65% (Stage 1) of the first pair of P-1 exponents in 18 hours. That not bad at all. If it wants to do Stage 2, there is plenty of RAM allocated.

Uncwilly · 2009-08-22, 16:55

You may want to try and get your P-1's to stagger the timeing of stage 2. Doing so will help memory usage.

storm5510 · 2009-08-22, 18:35

Quote:

Originally Posted by Uncwilly

You may want to try and get your P-1's to stagger the timeing of stage 2. Doing so will help memory usage.

Elaborate please.

CADavis · 2009-08-22, 18:49

one way is in local.txt put the line MaxHighMemWorkers=1

Uncwilly · 2009-08-22, 20:02

Stage 2 is the high memory usage stage. If 2 workers are doing stage 2 at the sme time, the both will be trying to use the 300M or more of memory at the same time. This can cause a bottleneck to appear. Stage 1 uses far less memory. If the 2 (or more) workers are not trying to do stage 2 at the same time, then, there is no bottleneck. That is why many folks suggest doing 1 or 2 LL, 1 P-1, and 2 or 1 T-F on a quad core.

2009-08-21, 02:17	#34
storm5510 Random Account Aug 2009 Not U. + S.A. 3·953 Posts	A basic question: I'm doing this on a Core 2 Duo at 3.0 GHz. Is this the right type of work? Yes or No? Something tells me I should be doing something more difficult than a lot of machines can do, like my laptop, for example. Last fiddled with by storm5510 on 2009-08-21 at 02:19 Reason: Because...

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Timing for different B1 values?	CRGreathouse	GMP-ECM	8	2018-05-12 05:57
Erroneous values of s_n?	CuriousKit	Math	15	2016-01-31 11:57
reserving a few k values	Trilo	Riesel Prime Search	7	2015-09-27 23:20
reserving a few k values	Trilo	Riesel Prime Search	0	2013-08-25 14:47
B1/B2 values	esakertt	Information & Answers	2	2012-11-14 13:11

2009-08-22, 02:43	#37
storm5510 Random Account Aug 2009 Not U. + S.A. 3·953 Posts	I attempted to transfer the LMH exponents to my old computer. It didn't fly for some reason. They were unreserved. I'm doing P-1 based on multiple suggestions Finding a new prime doesn't much interest me, for some reason. I would rather contribute, quietly, in the background.

2009-08-22, 12:55	#39
garo Aug 2002 Termonfeckin, IE 2⁴×173 Posts	He never said he felt he wasn't doing enough. 10metreh, will you stop steering people towards "other" projects? P-1 seems just right for his hardware and GIMPS needs more P-1 testers too.

2009-08-22, 16:46	#40
storm5510 Random Account Aug 2009 Not U. + S.A. 3×953 Posts	I've done 65% (Stage 1) of the first pair of P-1 exponents in 18 hours. That not bad at all. If it wants to do Stage 2, there is plenty of RAM allocated.

2009-08-22, 16:55	#41
Uncwilly 6809 > 6502 """"""""""""""""""" Aug 2003 101×103 Posts 3·7·17·31 Posts	You may want to try and get your P-1's to stagger the timeing of stage 2. Doing so will help memory usage.

2009-08-22, 18:49	#43
CADavis "werdwerdus" Jul 2005 Des Moines, Iowa, US 10101101₂ Posts	one way is in local.txt put the line MaxHighMemWorkers=1

2009-08-22, 20:02	#44
Uncwilly 6809 > 6502 """"""""""""""""""" Aug 2003 101×103 Posts 10101100111011₂ Posts	Stage 2 is the high memory usage stage. If 2 workers are doing stage 2 at the sme time, the both will be trying to use the 300M or more of memory at the same time. This can cause a bottleneck to appear. Stage 1 uses far less memory. If the 2 (or more) workers are not trying to do stage 2 at the same time, then, there is no bottleneck. That is why many folks suggest doing 1 or 2 LL, 1 P-1, and 2 or 1 T-F on a quad core.