Recreational Tests - Page 3

Dobri · 2022-08-23, 05:14

Quote:

Originally Posted by tuckerkao

I'm wondering why there was a need to recheck the trial factoring bits of M161803403 from 2⁷³ to 2⁷⁸. Kriesel's hardware is very reliable, there wouldn't have been any errors.

There is no special reason, often my automated script configures the TF tests listed in a queue to start from the lowest 2¹ value.
As concerned with reliability, the TF tests have no error correction mechanism.
With the limited error correction for LL tests, the quoted user currently has 1563 verified test results and 32 bad test results, see https://www.mersenne.org/report_ll/?...id=Kriesel&B1=.

Dobri · 2022-12-13, 05:33

Among the 51 prime exponents of known Mersenne primes, there are only 4 remaining exponents (M25964951, M30402457, M37156667, and M57885161) where a factor is yet to be found for at least one of the Mersenne numbers of either the preceding or succeeding prime exponents (or both) excluding the small consecutive prime exponents which give Mersenne primes:
M25964929, M25964957;
M30402401, M30402479;
M37156663, M37156673; and
M57885143, M57885167.

kriesel · 2022-12-13, 06:19

"Upgraded" martinette from 16GB to 64GB, to run v30.8 P-1, lost reliability, so running PRP on it for now. Can you tell when? https://www.mersenne.org/report_LL/?..._id=martinette

Dobri · 2022-12-13, 06:34

Quote:

Originally Posted by kriesel

"Upgraded" martinette from 16GB to 64GB, to run v30.8 P-1, lost reliability, so running PRP on it for now. Can you tell when? https://www.mersenne.org/report_LL/?..._id=martinette

From Post #23 (2022-08-23, 1563 verified test results and 32 bad test results) till now (2022-12-13, 1799 verified test results and 45 bad test results), it appears that there were 45 - 32 = 13 new bad test results related to all of your computers as the DC wavefront progresses.

kriesel · 2022-12-13, 06:39

Date sort this gives a 9 day period of uncertainty, 2022-10-15 to 2022-10-24, between all 31 good on martinette before "upgrade" and all 5 bad after. https://www.mersenne.org/report_LL/?..._id=martinette
(Consulting the Windows system event log shows a boot on 2022-10-22.)

storm5510 · 2022-12-15, 05:40

Quote:

Originally Posted by kriesel

Date sort this gives a 9 day period of uncertainty, 2022-10-15 to 2022-10-24, between all 31 good on martinette before "upgrade" and all 5 bad after. https://www.mersenne.org/report_LL/?..._id=martinette
(Consulting the Windows system event log shows a boot on 2022-10-22.)

Martinette? What is it you are trying to accomplish?

chalsall · 2022-12-15, 05:49

Quote:

Originally Posted by storm5510

What is it you are trying to accomplish?

I am similarly confused. I don't like being confused.

kriesel · 2022-12-15, 11:54

Dobri and tuckerkao were discussing my hardware fleet's average reliability. A significant component of the recent error rate turns out to have been caused by putting larger new memory in one laptop that happens to be named martinette; "upgraded" from 1 16 GB DIMM to 2 32 GIB DIMMs for the purpose of running all large-ram P-1 at or below the first test wavefront, made it too unreliable to run LL or P-1. Examining the GIMPS results for it indicated a stark change in reliability downward in mid October. That aligns with a boot on Oct 22, and IIRC the memory "upgrade". Coincidentally, it recently went off warranty, and the power adapter for it has now failed. Dell...
That one laptop is a few percent of my fleet but contributed a disproportionate 38% of the erroneous primality tests during the period. Identifying and taking action on low reliability hardware is a step on the path to raising reliability of results back up to where it should be.

storm5510 · 2022-12-15, 14:35

I think I get what you are doing:

Take Mersenne Prime 2³¹ as an example. The first non-Mersenne prime below it is 29 and the next above is 37. You are trying to find factors for these adjoining non-Mersenne primes.

Look at post #24. Those eight numbers have been factored to 2⁷⁶. These could be taken much higher with a decent GPU and mfaktc. Even if you were able to drop out just one, it would be worth the effort.

Dobri · 2023-01-02, 13:11

Here is an untested prime exponent containing 2023: M202362023.

kriesel · 2023-01-02, 16:17

And an exponent corresponding to 3am today, 202301023, 2023/01/02/0300, has already been factored.

2022-12-13, 06:39	#27
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 3²·11·79 Posts	Date sort this gives a 9 day period of uncertainty, 2022-10-15 to 2022-10-24, between all 31 good on martinette before "upgrade" and all 5 bad after. https://www.mersenne.org/report_LL/?..._id=martinette (Consulting the Windows system event log shows a boot on 2022-10-22.) Last fiddled with by kriesel on 2022-12-13 at 07:15

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Prime tests / BPSW tests in various number theory packages	MattcAnderson	MattcAnderson	15	2022-02-16 23:11
Hard recreational math problems	swishzzz	Math	9	2020-11-15 19:44
More P-1 tests	dave_0273	Marin's Mersenne-aries	1	2006-03-23 00:03
AI-slavery / recreational discussion	TauCeti	Soap Box	4	2004-04-07 00:00
Bad LL Tests	outlnder	Lounge	8	2002-10-21 00:12

2022-12-13, 05:33	#24
Dobri "ม้าไฟ" May 2018 534₁₀ Posts	Among the 51 prime exponents of known Mersenne primes, there are only 4 remaining exponents (M25964951, M30402457, M37156667, and M57885161) where a factor is yet to be found for at least one of the Mersenne numbers of either the preceding or succeeding prime exponents (or both) excluding the small consecutive prime exponents which give Mersenne primes: M25964929, M25964957; M30402401, M30402479; M37156663, M37156673; and M57885143, M57885167.

2022-12-13, 06:19	#25
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 3²·11·79 Posts	"Upgraded" martinette from 16GB to 64GB, to run v30.8 P-1, lost reliability, so running PRP on it for now. Can you tell when? https://www.mersenne.org/report_LL/?..._id=martinette

2022-12-15, 11:54	#30
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 3²×11×79 Posts	Dobri and tuckerkao were discussing my hardware fleet's average reliability. A significant component of the recent error rate turns out to have been caused by putting larger new memory in one laptop that happens to be named martinette; "upgraded" from 1 16 GB DIMM to 2 32 GIB DIMMs for the purpose of running all large-ram P-1 at or below the first test wavefront, made it too unreliable to run LL or P-1. Examining the GIMPS results for it indicated a stark change in reliability downward in mid October. That aligns with a boot on Oct 22, and IIRC the memory "upgrade". Coincidentally, it recently went off warranty, and the power adapter for it has now failed. Dell... That one laptop is a few percent of my fleet but contributed a disproportionate 38% of the erroneous primality tests during the period. Identifying and taking action on low reliability hardware is a step on the path to raising reliability of results back up to where it should be.

2022-12-15, 14:35	#31
storm5510 Random Account Aug 2009 Not U. + S.A. 2·1,429 Posts	I think I get what you are doing: Take Mersenne Prime 2³¹ as an example. The first non-Mersenne prime below it is 29 and the next above is 37. You are trying to find factors for these adjoining non-Mersenne primes. Look at post #24. Those eight numbers have been factored to 2⁷⁶. These could be taken much higher with a decent GPU and mfaktc. Even if you were able to drop out just one, it would be worth the effort.

2023-01-02, 13:11	#32
Dobri "ม้าไฟ" May 2018 2×3×89 Posts	Here is an untested prime exponent containing 2023: M202362023.

2023-01-02, 16:17	#33
kriesel "TF79LL86GIMPS96gpu17" Mar 2017 US midwest 3²×11×79 Posts	And an exponent corresponding to 3am today, 202301023, 2023/01/02/0300, has already been factored.