[QUOTE=James Heinrich;403663]And Bit-first preview [URL="http://www.gpu72.com/account/preview/200/01/0332000000/0333000000/73/6/3566/0/"]says[/URL] "[color=red]No assignments available which match your criteria.[/color]". Not sure if that's intentional.[/QUOTE]

Thanks for pointing that out. Until recently there were still candidates at only 72 bits, so the Javascript set the pledge level to 73. This has now been updated to be 74.

What's the attraction for doing such a humongous exponent offline? or any exponent? Is something that has the electricity for that long really not going to be networked for large swaths of time?

[QUOTE=aurashift;403695]What's the attraction for doing such a humongous exponent offline? or any exponent? Is something that has the electricity for that long really not going to be networked for large swaths of time?[/QUOTE]

If I'm understanding your question correctly... Some "Newbies" think they'll find a 100M digit prime quickly (and thus get the "easy" prize money), and so request several candidates from Primenet. After a couple of hours or so they realize that it will take them YEARS to complete a single assignment, and so abandon the effort, but don't have the courtesy to unreserve it.

As I understand the situation, no recycling rules currently exist in the 332M range, and so such assignments just sit there; potentially for years.

That is why I said in the beginning to give the man a cigar if he has fire and will to do the smoking work...

I would suggest to kick the butt of the guys who do not check in for (say) 60 days. Maybe no need progress (or it does?) but at least a check in, to say "hey guys, I am alive, I know I still have this assignment". That would be at least polite. If not, kick his butt out...

I don't believe there are any "offline" assignments. Those are just "lost sheep".

[QUOTE=LaurV;403707]
I don't believe there are any "offline" assignments. Those are just "lost sheep".[/QUOTE]

That's along the lines of what I was asking. I thought maybe there were people with tinfoil hats who grab assignments and then put their cpu in a dark room supplied with only electrical outlets and let it run in there, under lock and key, until the predicted date upon which the job would be completed. Until they open that door or connect the cpu to the 'net again, that exponent will be both prime and not prime.

Seriously though, what are the odds that someone who hasn't updated in 6 months is ever going to continue meaningful work? [Really, I'm not good at math, what are the odds? :wink:]

P.S. since most of my boxes can finish an exponent in ~260 days with two threads, I've switched most workers over to that. (8-12 exponents in 250-350 days as opposed to 2 exponents in 70-90...plus it's more efficient if I have to TF. )

[QUOTE=aurashift;403731]Seriously though, what are the odds that someone who hasn't updated in 6 months is ever going to continue meaningful work? [Really, I'm not good at math, what are the odds? :wink:][/QUOTE]

I'm also not very good at math, but I would argue somewhere around not a hope in hell.

I really think we should start expiring candidates which haven't checked in in 60 days, like we do for the DC and LL ranges -- there are a tonne of assignments up there which are clearly abandoned.

Also, as an aside, I don't know if Madpoo or George did it, or if Siarhei P. saw this thread and did it himself, but 292 of his assignments became available (all at 73 bits in the 332.3 range). GPU72 now has these, ready for TF'ing work -- it really would have been a shame if these were assigned to an LL'er who then started doing TF'ing work on (shudder) a CPU....

[QUOTE=chalsall;403734]I'm also not very good at math, but I would argue somewhere around not a hope in hell.

I really think we should start expiring candidates which haven't checked in in 60 days, like we do for the DC and LL ranges -- there are a tonne of assignments up there which are clearly abandoned.

Also, as an aside, I don't know if Madpoo or George did it, or if Siarhei P. saw this thread and did it himself, but 292 of his assignments became available (all at 73 bits in the 332.3 range). GPU72 now has these, ready for TF'ing work -- it really would have been a shame if these were assigned to an LL'er who then started doing TF'ing work on (shudder) a CPU....[/QUOTE]

Not I.

FYI, UncWilly had PM'd me asking for a list of 100M digit exponents that fall under the "grandfather" rules (assigned prior to 2014-03-01) and had very little work done, along with the bit levels they were at.

I sent him the list, basically anything with no work even started yet -or- < 25% through the LL stage, -or- haven't checked in since January of this year. The < 25% done was kind of arbitrary.

I'll note here what I mentioned to him: most of those grandfathered assignments haven't done squat since they were assigned. Here's a little rundown:

[CODE]Unexpired assignments for exponents >= 332192831 made < '2014-03-01'
2053 Total
1 Double-check (M332207803 - presumably poached)
1882 with zero progress since assigned
71 with progress between 0% and 25%
20 with progress between 25% and 50%
7 with progress between 50% and 75%
10 with progress > 75%
1173 that haven't updated since 2015-01-01
1056 that haven't updated since 2014-01-01
307 that haven't updated since 2013-01-01
112 that haven't updated since 2012-01-01
770 that have checked in AFTER 2015-01-01 but still haven't shown any progress
354 that have checked in TODAY but still haven't shown any progress
72 that have checked in during the past 7 days and have actually done something since being assigned
51 of those "active" assignments are in the LL stage
13 of those "active" assignments are in Stage 1 of P-1
1 of those "active" assignments is at 1.1% doing TF to 68
7 of those "active" assignments are in the process of TF to 75[/CODE]

On the bright side, there are some that seem pretty close to finishing, like these:
[URL="http://www.mersenne.org/assignments/?exp_lo=332284963"]M332284963[/URL]
[URL="http://www.mersenne.org/assignments/?exp_lo=332204623"]M332204623[/URL]

I don't know if George has already weighed in on the expiration rules for these, but these are all grandfathered assignments we're talking about here. However, those rules are generally only strict when it comes to assignments in the current "wavefronts" of DC and LL work. These are so far out there that the expiration task basically ignores them. It's probably fair to say there could or should be some expiration even for these.

After all, 345 of these were assigned back in 2011. 315 of those haven't even started, including some that have checked in recently but still haven't done a single thing on the assignment in nearly 4 years.

Might be worth noting that of the total 2053 grandfathered assignments we're talking about here, 802 of them are assigned to just one user: "GPU Factoring"

I don't know if that's because they were checked out to GPU72 for TF work way back when and simply haven't done anything with them? Or if there's something else going on? 32 of those haven't checked in since January 2014 at the latest. The rest have been checking in regularly, just with zero progress and estimated completion dates of some time in the year 2031.

202 of the GPU Factoring assignments were made in 2011
150 in 2012
422 in 2013
28 in 2014

Sorry... long post, lots of info. Do with it what you will. :smile:

[QUOTE=Madpoo;403762]Might be worth noting that of the total 2053 grandfathered assignments we're talking about here, 802 of them are assigned to just one user: "GPU Factoring"[/QUOTE]

As explained elsewhere, George asked me to bring in those assignments not yet appropriately TF'ed. Because of the way I do this, any candidates assigned by a non registered user via the Primenet Manual Assignment form is transferred to GPU Factoring.

I consider this a (Primenet) bug; why on god's green earth should an unregistered user be allowed to reserve work? (Hint: someone malicious could easily cause havoc.)

60 days may be pushing it for people who shut down crunching during seasonally hot weather. 90 days would be a lot safer, IMHO.

[QUOTE=TObject;403769]60 days may be pushing it for people who shut down crunching during seasonally hot weather. 90 days would be a lot safer, IMHO.[/QUOTE]

My understanding is clients are supposed to report every 24 hours or so.

Whatever we converge on, what is currently happening is suboptimal.

Even 180 days would accommodate any imaginable reasonable reason to be offline for 100M exponents.