Less than 10,000 left....

[henryzz]

Originally Posted by petrw1

Slowing down here too .... EXCEPT .... George just released all the low LL tests



As well he only assigns low-end tests to "preferred" CPUs.
http://www.mersenneforum.org/showpos...99&postcount=9

so there is no reason why this number shouldn't drop QUICKLY now.

yes i would expect it to drop quite quickly
my guess because of this is: in 150 days we shall have <100 left

[rudi_m]

Originally Posted by petrw1

Again no longer having an assignment report I can't tell for sure how many of the 33M range are less than 33,219,280 but in the past it was virtually all of them.

Yes I even checked that range completely:
- maximum of these 93 LLERR is 33217837
- maximum of these 554 No-LL is 33218797

BTW I just noticed that on 2008-12-06 and 2008-12-08
ANONYMOUS assigned 497 of these 647 exponents!
No unassigned one left.
What happened here?

I also saw that on other ranges but didnt' checked it exactly.

[petrw1]

Originally Posted by rudi_m

BTW I just noticed that on 2008-12-06 and 2008-12-08
ANONYMOUS assigned 497 of these 647 exponents!
No unassigned one left.
What happened here?

I also saw that on other ranges but didnt' checked it exactly.

Maybe this:
http://www.mersenneforum.org/showthread.php?t=11100

[rudi_m]

Originally Posted by petrw1

Maybe this:
http://www.mersenneforum.org/showthread.php?t=11100

Ah thx!

[petrw1]

Originally Posted by henryzz

yes i would expect it to drop quite quickly
my guess because of this is: in 150 days we shall have <100 left

I have a little over half-a-percent of these ... and my share should almost be done in 150 days.

[JuanTutors]

You know what would be a cool feature? If for the next 10000 primes below 10M digits, only fast computers could get them. I know that people complain that slow computers slow down progress. Of course many of us know there are really good arguments why that's not true, including things such as we wouldn't be as far as we are if it weren't for slow computers. That's why we shouldn't poach other people's primes. But would it be bad in any way if we started giving slower computers 10M+ digit primes now and started giving the available 10000 primes below 10M digits to faster computers?

[garo]

George/Scott have implemented a system in v5 that gives the smallest exponents to "trusted" computers. I don't know where the trusted boundary is at the moment but it is likely below 30M.

[petrw1]

Originally Posted by dominicanpapi82

You know what would be a cool feature? If for the next 10000 primes below 10M digits, only fast computers could get them. I know that people complain that slow computers slow down progress. Of course many of us know there are really good arguments why that's not true, including things such as we wouldn't be as far as we are if it weren't for slow computers. That's why we shouldn't poach other people's primes. But would it be bad in any way if we started giving slower computers 10M+ digit primes now and started giving the available 10000 primes below 10M digits to faster computers?

It's not the Slow computers that are dragging down that range; it is people who take assignments and before completion abandon them or work on them sparingly and report in very infrequently.

I suggest many people with Fast computers want the chance to get the larger assignmeents and better opportunities to find the next BIG prime.

Read the following post and the next few .... I think you will find that George even found a better way to clean up low-end work.

http://www.mersenneforum.org/showpos...82&postcount=7

[cheesehead]

Originally Posted by dominicanpapi82

If for the next 10000 primes below 10M digits, only fast computers could get them.

Do you want the "slow" computers to "slow down" the range _above_ 10M digits instead of the range _below_ 10M digits, or what is the motive? What would be cool about that feature?

I know that people complain that slow computers slow down progress.

... because those people are (a) near-sightedly focused on only the lowest "milestones" instead of on the big picture of overall GIMPS progress, or (b) somehow deluded into thinking that no-progress (by a particular system if a user becomes discouraged about participating) is better than slow-progress, I think. Are there other reasons?

Do some folks with "fast" systems think that if the folks with "slow" systems are denied access to assignments with the lowest exponents, those "slow-system" folks will decide to replace their "slow" systems with "fast" systems? Or what?

But would it be bad in any way if we started giving slower computers 10M+ digit primes now and started giving the available 10000 primes below 10M digits to faster computers?

Yes. It might tend to discourage participation by those with the slowest computers if they are denied assignments to the lowest exponents available. Why should they be discriminated-against if they have the same reliability record as those with "fast" computers? (Why didn't you mention reliability, anyway?)

Would it be good in any way? If so, how?

- - -

To answer my own question about mentioning reliability:

We don't have a report ranking users by their reliabilities, do we? Maybe we should, to raise the visibility of that important quality?

In general, how about creating reports for measures that show which users are most useful to GIMPS in various respects other than simply mass results due to having many systems? High reliability, high whatever-the-score-is-that-PrimeNet-uses-to-assign-juicy-assignments, ... -- things that users with only one or a few systems can aspire to rank high in.

[petrw1]

	
Start        LLERR	NO-LL
18000000	2	
21000000	1	
23000000	8	13
24000000	25	4
25000000	10	27
26000000	24	44
27000000	53	79
28000000	133	337
29000000	159	259
30000000	196	546
31000000	268	1050
32000000	321	1240
33000000	92	528

5,419 as of today.
56.6% remaining from my first update of 9,576 on April 22, 2008

Originally Posted by petrw1

Dec 12: ... so there is no reason why this number shouldn't drop QUICKLY now.

Is it???

A drop of 246 since Dec 12 ... 25 days = 10 per day.
vs. a drop of 420 from Nov 3 to Dec 12 ... 39 days = almost 11 per day.

Anyway if we can sustain 10 per day there is about 1.5 more years to go.

They are ALL currently assigned.

My 33 assignments (0.6%) in this range will all be done within 6 months.
Can everyone else who has assignments in this range aim for completion for July 4th?
Maybe we can get George to commit to fireworks to commemorate this milestone.

[joblack]

Originally Posted by cheesehead

Just a friendly reminder to impatient folks:

As I've explained in the past, the GIMPS project itself is not being held up just because some exponents below a certain milestone aren't all being actively tested, as long as all the CPUs that might've been testing them are instead testing some other higher-than-that-milestone assigned exponents. Any "delay" in reaching the milestone on which you're focused is matched by a speedup in reaching higher milestones later on.

That´s true - every Gigahertz second is fine for the project. If you force people to calculate a specific number spectra their could stop supporting the project ...