Trial Factor and P-1 assignments squatting in Cat 0 and Cat 1 ranges

[snme2pm1]

Originally Posted by kriesel

auto manufacturing, instruments, etc, metric is prevalent.

Agreed, I need my metric socket set when working on the vehicle.

[Uncwilly]

User Christian Geschwentner has 74 DC assignments in the DC Cat 0 and very low Cat 1 assignments that are manual assignments that have had 0 progress.

[slandrum]

Originally Posted by Uncwilly

User Christian Geschwentner has 74 DC assignments in the DC Cat 0 and very low Cat 1 assignments that are manual assignments that have had 0 progress.

It's not the first time they've had large blocks of exponents reserved that eventually expire out.

[S485122]

Originally Posted by kriesel

...
Manual assignments generally have a more committed human in the loop. Including the gpu crew.
...

A bit late replying : just as with mprime / Prime95 some "manual assignments" users forget their assignments, abandon them, make errors, stop for whatever reason. Even people going into loops to get low exponents sometimes forget these reservations / assignments or report them when they have already been assigned to another user. Few people are continuously keeping an eye on all their assignments, manual or not, as a look at the Active Assignments page will demonstrate. Another point is that not everybody cleans up after experimenting with the Manual Assignmnents pages.

Originally Posted by Uncwilly

User Christian Geschwentner has 74 DC assignments in the DC Cat 0 and very low Cat 1 assignments that are manual assignments that have had 0 progress.

User LordJulius was able to reserve 30 exponents for manual trial factoring yesterday. Those exponents were category 1 since the end of October 2021.

The solution has been discussed earlier :
- apply the same rules for all type of work for exponents in the "categorised" ranges.
- apply the 60 days past due expiration rule on all assignments (quite a lot of which are more than 1860 days or 5 years past due.)

[Zhangrc]

Originally Posted by S485122

apply the 60 days past due expiration rule on all assignments (quite a lot of which are more than 1860 days or 5 years past due

I disagree, because these assignments are usually far above Cat 4. For example, trial factoring in 230M range have at least 180 days to go, and PRP in these ranges have even more time to go.

[kriesel]

In the lower Cat numbers, the expiration period is a multiple of how long a run should take, allowing for queuing behind other assignments, occasional system or application stoppages, etc. even with slower hardware. In the upper reaches, 60. days is a fraction of how long it takes to run one nonstop, even on the faster available GPUs and software. 100Mdigit is ~15. days on a Radeon VII, near 1G exponent they take ~150. days. Manual GPU assignments are typically not checking in progress, either, so although they may be 40-90% done when you would expire them, the PrimeNet server still shows them at 0.0%.

[S485122]

Originally Posted by Zhangrc

I disagree, because these assignments are usually far above Cat 4. For example, trial factoring in 230M range have at least 180 days to go, and PRP in these ranges have even more time to go.

Originally Posted by kriesel

In the lower Cat numbers, the expiration period is a multiple of how long a run should take, allowing for queuing behind other assignments, occasional system or application stoppages, etc. even with slower hardware. In the upper reaches, 60. days is a fraction of how long it takes to run one nonstop, even on the faster available GPUs and software. 100Mdigit is ~15. days on a Radeon VII, near 1G exponent they take ~150. days. Manual GPU assignments are typically not checking in progress, either, so although they may be 40-90% done when you would expire them, the PrimeNet server still shows them at 0.0%.

I talked about the first assignment rule : "Since PrimeNet began, the server has recycled exponents where the client computer is 60 days past due in updating the server. " It concerns the past due communicating with the server date : the date at which the assignment must be updated or extended, not the expected completion date.

One gets 1050 days (a bit less than 3 years) to complete a manual PRP on the assignment. And this without extending the assignment.

To see what I refer to look at "active" assignments outside of the hard worked ranges.

[LordJulius]

Originally Posted by S485122

User LordJulius was able to reserve 30 exponents for manual trial factoring yesterday. Those exponents were category 1 since the end of October 2021.

Yep, server gave me those and the last of them will be submitted in ~20min. I stay very much on top of my assignments - maybe the server has that figured and allows these Cat ones to me because of quick turn around? Just taking advantage of my RTX3060; did get a factor for one of them, 59944949.


I also agree that handing out DC/FTC exponents that don't *need* more TF or PM1 really isn't justified.


Per the 2-5 years expired/overdue/never-updated assignments, I'm working on clearing those out (80M-110M) with 2 other machines, since the server apparently has completely forgotten about them.


Seems for both above cases the implementation of existing rules need to be verified, then once that's working, see what needs changing/updating.

[slandrum]

The user GPU Factoring has hundreds of TF assignments in the 108.3M range (and those assignments are moving into cat 0 range). The assignments have variable ETAs, many of them months or years out.

The highlighted exponent is currently the edge of cat 0.

[Zhangrc]

I'm doing several of them (77 to 78).
Maybe I should unreserve them, but the lowest untested is 107.1M, so I still have time to go.

[Uncwilly]

That user is the PrimeNet interface for GPU72. Some of those may be P-1's not actual TF's. Post over in the GPU72 status thread. You will get a response from chalsall there.