Largest LL Test Ever Completed

jinydu · 2010-02-08, 04:31

Less than half an hour ago, an LL test for M150,000,091 finished, earning me 875.0005 GHz-Days. The test had been running for 140.2 days on a Core 2 Quad from the Caltech Math department. Unfortunately, 2^150000091 -1 is not prime.

Primenet indicates this is indeed the largest LL test ever reported. But does anyone know of a larger test ever being completed?

NBtarheel_33 · 2010-02-08, 06:39

I'd say PrimeNet is the best and most current repository of such information, so congratulations - you have the largest LL test ever completed. I was thinking about going for the record a while back, and 150M was right where I had been thinking about testing. Looks like 200M is the next frontier...

You should run some timings on that quad for a 100M-digit number. I bet it could run one in about 18 months. Or ask around the department and see if they'd let you run Prime95 on any bigger hardware (Caltech has to have some big iron processing power laying around just waiting to crunch primes!!)

And you should check and see if your department would let you run GIMPS on more computers (as curtisc does). It'd be fun to give curtisc a bit of competition.

Looks like true octocore computers will bring the testing time for a 150M exponent just about down to the time it takes a single-core to run a 47M exponent. That will really be progress! Bring on the octocores!

jinydu · 2010-02-08, 06:54

Unfortunately, that computer is the department's most powerful by far. I asked our computer administrator once whether we could get a Core i7 soon; she said "Not going to happen." On another occasion, she told me that the department's annual computer budget is $15,000; and that the secretaries alone ate half of it this year. The majority of our computers are still Pentium 4. In fact, my office had a Pentium 3 last year; mercifully it has since been replaced with a P4. Maybe I should ask other departments...

After I lost the M100,000,039 race, my first thought was to go for 100M digits. If I remember correctly, the test wouldn't have finished until 2014; so I decided to go for something smaller but still well above the then record 101M.

ewmayer · 2010-02-08, 21:58

Nice work, jinydu!

It would be interesting/useful to get an early DC of this one, since it is both really big and fairly close to the upper exponent limit for 8192K FFT length. Did you have Prime95 configured to print out intermediate residues for your run?

petrw1 · 2010-02-09, 00:05

xorbe is working on a 100M digit test...should finish this summer.

http://www.mersenneforum.org/showthr...852#post204852

mdettweiler · 2010-02-09, 00:20

Quote:

Originally Posted by petrw1

xorbe is working on a 100M digit test...should finish this summer.

http://www.mersenneforum.org/showthr...852#post204852

Think there's time to fit another sub-100M digit record test in before then?

ixfd64 · 2010-02-09, 01:33

Congratulations!

By the way, I'd say that you are quite lucky that your department allowed you to "borg" its computers!

TimSorbet · 2010-02-09, 01:35

Quote:

Originally Posted by mdettweiler

Think there's time to fit another sub-100M digit record test in before then?

Only if you can complete a 150M<p<332(ish)M exponent in under half the time he's taking for the whole number.
Suggestion: 200M http://www.mersenne.org/report_expon...&B1=Get+status
From a bit of math from http://mersenne-aries.sili.net/throughput.php?cpu=Intel%28R%29+Core%28TM%29+i7+CPU+965+%40+3.20GHz|256|8064&mhz=3600 and http://www.mersenneforum.org/showthr...p?t=59&page=10, I guesstimate that an i7@3.6 GHz running 8 threads (like xorbe) could do such an LL in 6 months (starting now, that's about August 11). I think the latest estimate for xorbe's run is June, (a bit of extrapolating suggests July 18; namely 290 days from April 21 to Feb 5, 290/.64*.36=165 days, and 165 days after Feb 5 is July 18) so you'd need hardware a decent bit faster than an i7@3.6 GHz to have a good chance at beating him to it.
(or you could lower your standards to a bit past the current 150M record and with similar hardware probably beat him easily)

Can it be done? Certainly. Is it easy to do? Not really.

joblack · 2010-02-09, 20:07

Quote:

Originally Posted by Mini-Geek

Only if you can complete a 150M<p<332(ish)M exponent in under half the time he's taking for the whole number.
Suggestion: 200M http://www.mersenne.org/report_expon...&B1=Get+status
From a bit of math from http://mersenne-aries.sili.net/throughput.php?cpu=Intel%28R%29+Core%28TM%29+i7+CPU+965+%40+3.20GHz|256|8064&mhz=3600 and http://www.mersenneforum.org/showthr...p?t=59&page=10, I guesstimate that an i7@3.6 GHz running 8 threads (like xorbe) could do such an LL in 6 months (starting now, that's about August 11).

There is no i7 oct core (they all have only 4 cores with hyperthreading). I seriously doubt that it can be done in 6 months. You need more than a year.

My quad core Q6600@3Ghz estimates over 3 years for an 100M number so the i7 is slightly faster.

ixfd64 · 2010-02-09, 20:47

A six-core i7 chip is coming out this March, though.

joblack · 2010-02-09, 21:32

Quote:

Originally Posted by ixfd64

A six-core i7 chip is coming out this March, though.

The 6-core from AMD is already out. It's really not a matter of availability but of your budget. For most of us only a quad core is in the budget so it doesn't really make sense to calculate with hypothetical 8 or 16 cores ;).

2010-02-08, 04:31	#1
jinydu Dec 2003 Hopefully Near M48 2×3×293 Posts	Largest LL Test Ever Completed Less than half an hour ago, an LL test for M150,000,091 finished, earning me 875.0005 GHz-Days. The test had been running for 140.2 days on a Core 2 Quad from the Caltech Math department. Unfortunately, 2^150000091 -1 is not prime. Primenet indicates this is indeed the largest LL test ever reported. But does anyone know of a larger test ever being completed?

2010-02-08, 06:39	#2
NBtarheel_33 "Nathan" Jul 2008 Maryland, USA 10001011011₂ Posts	Pretty sure you're the king...for now, anyway. I'd say PrimeNet is the best and most current repository of such information, so congratulations - you have the largest LL test ever completed. I was thinking about going for the record a while back, and 150M was right where I had been thinking about testing. Looks like 200M is the next frontier... You should run some timings on that quad for a 100M-digit number. I bet it could run one in about 18 months. Or ask around the department and see if they'd let you run Prime95 on any bigger hardware (Caltech has to have some big iron processing power laying around just waiting to crunch primes!!) And you should check and see if your department would let you run GIMPS on more computers (as curtisc does). It'd be fun to give curtisc a bit of competition. Looks like true octocore computers will bring the testing time for a 150M exponent just about down to the time it takes a single-core to run a 47M exponent. That will really be progress! Bring on the octocores!

2010-02-09, 00:05	#5
petrw1 1976 Toyota Corona years forever! "Wayne" Nov 2006 Saskatchewan, Canada 5²×211 Posts	xorbe is working on a 100M digit test...should finish this summer. http://www.mersenneforum.org/showthr...852#post204852 Last fiddled with by petrw1 on 2010-02-09 at 00:06

2010-02-09, 01:33	#7
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 4676₈ Posts	Congratulations! By the way, I'd say that you are quite lucky that your department allowed you to "borg" its computers! Last fiddled with by ixfd64 on 2010-02-09 at 02:06

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Largest LL tests ever completed	aketilander	Data	256	2020-06-15 12:21
Completed 29M work not showing as completed in GPU72	Chuck	GPU to 72	2	2013-02-02 03:25
Prime95 Gave Up on Completed Test, Started New One	jinydu	Lounge	5	2010-12-16 18:52
First Completed LL Test Club	Complex33	Teams	2	2003-12-10 23:40
need Pentium 4s for 5th largest prime search (largest proth)	wfgarnett3	Lounge	7	2002-11-25 06:34

2010-02-08, 06:54	#3
jinydu Dec 2003 Hopefully Near M48 1758₁₀ Posts	Unfortunately, that computer is the department's most powerful by far. I asked our computer administrator once whether we could get a Core i7 soon; she said "Not going to happen." On another occasion, she told me that the department's annual computer budget is $15,000; and that the secretaries alone ate half of it this year. The majority of our computers are still Pentium 4. In fact, my office had a Pentium 3 last year; mercifully it has since been replaced with a P4. Maybe I should ask other departments... After I lost the M100,000,039 race, my first thought was to go for 100M digits. If I remember correctly, the test wouldn't have finished until 2014; so I decided to go for something smaller but still well above the then record 101M.

2010-02-08, 21:58	#4
ewmayer ∂²ω=0 Sep 2002 República de California 26753₈ Posts	Nice work, jinydu! It would be interesting/useful to get an early DC of this one, since it is both really big and fairly close to the upper exponent limit for 8192K FFT length. Did you have Prime95 configured to print out intermediate residues for your run?

2010-02-09, 20:47	#10
ixfd64 Bemusing Prompter "Danny" Dec 2002 California 2494₁₀ Posts	A six-core i7 chip is coming out this March, though.