GPU to 72 status... - Page 173

chalsall · 2013-02-09, 15:56

Quote:

Originally Posted by Redarm

The range of 65M - 66M seems to be locked...
could anybody verify this?

Not "locked", simply that we don't have any candidates available there at the moment.

Please see the Available Assignments report to see what's available. Choosing "What Makes Sense" (with the ranges left as default) on the assignments form will always have assignments available (as shown in the preview area).

Aramis Wyler · 2013-02-09, 16:03

It's funny someone would bring that up. I was thinking last night that the 'exponent range' you could request (for lltf, at >71 bitdepth) was 0 to 100000000 but really no matter how you fiddled with the numbers you weren't going to get anything outside of 60-65M with tf of 71-73, and could only creep into 59M at 74+.

I especially love the warning for if you set the bitdepth to 71 - Only numbers above 60M!!! - which is, of course, the same as 72 or 73.

No complaints here, but I am curious about something. How come when we watch the factoring graph movies (love those things! Thanks for updating!) the 50-53M range only appears to be about half factored. Is there a reason those aren't made available to factor? I'm not sure they should be considered double checks since I'm currently doing first time tests on 6 LLs and 5 of them are < 54M (The other is 54M).

chalsall · 2013-02-09, 16:18

Quote:

Originally Posted by Aramis Wyler

It's funny someone would bring that up. I was thinking last night that the 'exponent range' you could request (for lltf, at >71 bitdepth) was 0 to 100000000 but really no matter how you fiddled with the numbers you weren't going to get anything outside of 60-65M with tf of 71-73, and could only creep into 59M at 74+.

Yes. GPU72 only "brings in" (read: officially reserves from Primenet) a few tens of thousands of candidates at a time for "lending" to GPU72 participants.

Quote:

Originally Posted by Aramis Wyler

I especially love the warning for if you set the bitdepth to 71 - Only numbers above 60M!!! - which is, of course, the same as 72 or 73.

Thanks for reminding me -- that's legacy code from when we still had work to do below 60M, and I didn't want people wasting everyone's time by only going a step-at-a-time there.

Quote:

Originally Posted by Aramis Wyler

No complaints here, but I am curious about something. How come when we watch the factoring graph movies (love those things! Thanks for updating!) the 50-53M range only appears to be about half factored. Is there a reason those aren't made available to factor? I'm not sure they should be considered double checks since I'm currently doing first time tests on 6 LLs and 5 of them are < 54M (The other is 54M).

The graphs, charts, movies etc (which are unfortunately temporarily off-line; the server hosting that appeared to have suffered a DDOS last night (my provider, 1and1, are working on it)) don't differentiate between LL and DC for the factor-level.

However, the GPU72 Esitmated Completion report does, and I can assure you that everything which is still an LL candidate below 57M is at at least 72, and everything between 57M and 60M is at at least 73.

Aramis Wyler · 2013-02-09, 17:00

Ah, so those remnants in the 50-53M range have actually been factored - by a LL test - but the chart is only showing the ones that were factored by trial factoring, ecm, and p-1?

chalsall · 2013-02-09, 17:12

Quote:

Originally Posted by Aramis Wyler

Ah, so those remnants in the 50-53M range have actually been factored - by a LL test ...

You got the nomenclature slightly wrong... They weren't "factored" by an LL test, but instead had a successful (read: no errors) LL test (which can't find factors). Thus they are now classified as Double Check (DC) candidates.

DC candidates deserve approximately one less level of TFing since a factor found will only save approximately one test (the DC), instead of approximately two (the LL and then the DC).

Quote:

Originally Posted by Aramis Wyler

... but the chart is only showing the ones that were factored by trial factoring, ecm, and p-1?

The chart is showing the trial factoring level for all candidates, LL, DC and "cleared" (read: have had a successful LL and a matching DC proving them composite).

Note that for finding factors only TFing is "deterministic" and can be represented with a single simple integer. P-1 and ECM use very different techniques, and currently their statistics are not "visualized".

Redarm · 2013-02-09, 18:01

primenet denies TF-assignments in that range (65M - 66M) too, don't understand why :(

chalsall · 2013-02-09, 18:06

Quote:

Originally Posted by Redarm

primenet denies TF-assignments in that range (65M - 66M) too, don't understand why :(

I don't know for a fact, but it's possible George is keeping the 60M to 70M range for GPU72 to manage.

We really need to focus on the lower 60s ranges right now (you'll see why shortly), but if you really want to work in 65M, I'll bring some in for you.

chalsall · 2013-02-09, 18:24

Quote:

Originally Posted by chalsall

ALL HANDS ON DECK!!! WE'RE CURRENTLY AT DEFCON 1!!!

Is there a DEFCON 0???

Before the big announcement Primenet was only handing out about 183 candidates a day (taking into account the expires which get recycled). At that rate we had about a 30 day lead, and were very slowing gaining (going to 73).

The number of LL assignments being issued by Primenet is now around 1,000 a day!!!

Therefor, I have temporarily lowered the default TF level for LL from 73 to 72, in preparation for having to start releasing at that level. I've also lowered the number of candidates at 73 cached for P-1 work, and will continue to lower this number as the need arises.

We still have about a three day lead, but unless we can produce about 1,000 TF results to at least 72 a day, we're toast!

Also, it would be helpful if those who are doing P-1 work would complete and return them to Primenet as soon as possible.

GPU72.com out....

Redarm · 2013-02-09, 19:06

Quote:

Originally Posted by chalsall

We really need to focus on the lower 60s ranges right now

Affirmative!

flashjh · 2013-02-09, 19:08

Chris,

Do you want me to un-reserve everything and re-reserve with pledge to 72? I can turn a lot more up to 72 than I'm doing to 73.

swl551 · 2013-02-09, 19:10

Reconfiguring fetching to 72.

Did I mention I am actually on the GPU72 team now..... Must comply with the boss!

2013-02-09, 16:03	#1894
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 424₁₀ Posts	It's funny someone would bring that up. I was thinking last night that the 'exponent range' you could request (for lltf, at >71 bitdepth) was 0 to 100000000 but really no matter how you fiddled with the numbers you weren't going to get anything outside of 60-65M with tf of 71-73, and could only creep into 59M at 74+. I especially love the warning for if you set the bitdepth to 71 - Only numbers above 60M!!! - which is, of course, the same as 72 or 73. No complaints here, but I am curious about something. How come when we watch the factoring graph movies (love those things! Thanks for updating!) the 50-53M range only appears to be about half factored. Is there a reason those aren't made available to factor? I'm not sure they should be considered double checks since I'm currently doing first time tests on 6 LLs and 5 of them are < 54M (The other is 54M). Last fiddled with by Aramis Wyler on 2013-02-09 at 16:05 Reason: syntax.

2013-02-09, 19:10	#1903
swl551 Aug 2012 New Hampshire 2³·101 Posts	I'm stopping my 100 assignment DC work. Reconfiguring fetching to 72. Did I mention I am actually on the GPU72 team now..... Must comply with the boss! Last fiddled with by swl551 on 2013-02-09 at 19:11

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2013-02-09, 17:00	#1896
Aramis Wyler "Bill Staffen" Jan 2013 Pittsburgh, PA, USA 2³×53 Posts	Ah, so those remnants in the 50-53M range have actually been factored - by a LL test - but the chart is only showing the ones that were factored by trial factoring, ecm, and p-1?

2013-02-09, 18:01	#1898
Redarm Apr 2012 Berlin Germany 63₈ Posts	primenet denies TF-assignments in that range (65M - 66M) too, don't understand why :(

2013-02-09, 19:08	#1902
flashjh "Jerry" Nov 2011 Vancouver, WA 1,123 Posts	Chris, Do you want me to un-reserve everything and re-reserve with pledge to 72? I can turn a lot more up to 72 than I'm doing to 73.