How to set 2^78 as default on trial factoring

[piforbreakfast]

I am putting several more computers online this week for trial factoring and I was wondering if there was a way to configure the settings to automatically go up to 2^78 on all exponents tested, so that way I won't have to change the ranges manually in worktodo.txt. I just wanted to "set it and forget it."

[kriesel]

Originally Posted by piforbreakfast

I am putting several more computers online this week for trial factoring and I was wondering if there was a way to configure the settings to automatically go up to 2^78 on all exponents tested, so that way I won't have to change the ranges manually in worktodo.txt. I just wanted to "set it and forget it."

Probably yes, but it depends on what hardware, software, assignment mechanism & possibly OS combination(s) you're asking about, and have not indicated. (If you're planning cpu TF, please leave that to the much more capable GPUs instead, and do work that's better suited to cpus. If you're planning to TF in excess of performance-optimized guidelines, please don't.)

[piforbreakfast]

Originally Posted by kriesel

Probably yes, but it depends on what hardware, software, assignment mechanism & possibly OS combination(s) you're asking about, and have not indicated. (If you're planning cpu TF, please leave that to the much more capable GPUs instead, and do work that's better suited to cpus. If you're planning to TF in excess of performance-optimized guidelines, please don't.)

How can I tell if my operating system and hardware aren't optimized for TF? I don't have the specs on them right in front of me.

[piforbreakfast]

It looks like the hardware on each of these computers I just received is Core i5 vPro. Are those adequate for trial factoring?

[kruoli]

It would work, but would be extremely inefficient. Do these computer come with a GPU? In that case, depending on the model, the GPU might be adequate for TF.

In general, today we have no CPU that would be efficient doing TF.

[petrw1]

Originally Posted by kruoli

In general, today we have no CPU that would be efficient doing TF.

For example I have a fairly high-end i7-7820x (8-core CPU with high speed RAM).
I also have a 2080Ti GPU.

What my CPU could factor in 2 days would take my GPU 1 hour.

[Uncwilly]

The many benefit of those (10 year) old CPU's doing TF will be to heat your room.
Sure you can do TF with them if you like. As other said it would be better to use them to host a GPU and do the TF on the GPU. Even if the hardware is free, the cost of the electricity might not be worth the TF being done. If you to do it anyway you can by getting manual assignments. However, with a CPU, you should follow this chart with regards to bit depth. https://www.mersenne.ca/factorbits.php
For GPU's the yellow seen on charts like this one is what makes sense. https://www.mersenne.ca/status/tf/0/0/2/10000
You will notice that the bit depth difference is about 4. So it does not make sense with a GPU to go to 78 until you get to exponents of 408M.

If you have signed up with GPU72, there is a manual assignment form here: https://www.gpu72.com/account/getassignments/lltf/
You can set the level that you will factor too. It will give you assignments formatted that way.
Or you can use this page on PrimeNet: https://www.mersenne.org/manual_gpu_assignment/
Set the option bit level. You must be signed in to get assignments.

I suggest that other tasks you could use those machines for (in my preferred order):
P-1 (if they have a good amount of memory and banked right)
LL-DC (this will help the project along and the more help we get the there the better)
PRP-DC (there are fewer of these to do and you may wind up running out in a year or two).
CERT's (you could set them up to run certs almost exclusively)
ECM (if they have reasonable amount of memory)
PRP-CF (with software that will VDF's)

You CPUs, your "$$$$", your choice.

[kriesel]

Originally Posted by piforbreakfast

It looks like the hardware on each of these computers I just received is Core i5 vPro. Are those adequate for trial factoring?

Originally Posted by Uncwilly

The main?y benefit of those (10 year) old CPU's doing TF will be to heat your room.

(lots of good stuff cut, following lightly edited for visibility)

P-1 (if they have a good amount of memory and banked right)
LL-DC (this will help the project along and the more help we get the there the better)
PRP-DC (there are fewer of these to do and you may wind up running out in a year or two).
CERT's (you could set them up to run certs almost exclusively)
ECM (if they have reasonable amount of memory)
PRP-CF (with software that will VDF's)

Your CPUs, your $, your choice.

Not sure where Core I5 vPro = 10 years old is coming from. (Did I miss some post in another thread that says these are old used boxes?) There are 10th-generation (current production) versions coming up in search for that incomplete cpu description.

Missing at or near the top of the GIMPS cpu work priority list in my opinion is PRP/GEC/Proof as DC for LL first run exponents, a manual assignment choice recently added by George, which is faster and more certain of reliability of result than LL DC with its chance of undetected error, and not too demanding of memory quantity. Also PRP first test at the wavefront, although there's already plenty of that occurring. It would be great to have more help on the ~8-year backlog of DC. And those assignments finish quicker and don't require as much ram to run. And several good matching DC per worker are necessary to qualify a prime95/mprime instance for lower category assignments.

Note also that an instance of mprime/prime95 can run mixed loads; one worker can be allowed much of the system ram and run P-1 quite well, while any other workers do DC or first test with their moderate memory footprints. I routinely do this. P-1 can use a lot of ram for best performance, but does not require it to run. I've run P-1 from 40GB to 1GB/worker depending on exponent and available ram.
What I don't do normally is GIMPS TF on cpu; too wasteful when gpus are so much faster and more energy efficient at TF than cpus are.

[Uncwilly]

Originally Posted by kriesel

(lots of good stuff cut, following lightly edited for visibility)
Not sure where Core I5 vPro = 10 years old is coming from.

I got the 10 years from here:
https://www.digitimes.com/photogallery/showphoto.asp?ID=4791#:~:text=Intel%20Core%20i5%20vPro%20processor&text=8%20March%202011
and here
https://www.laptopmag.com/articles/intel-vpro-faq

Doing a PRP with v30.3+ instead of LL for a DC is fine. More or less the same thing.

[piforbreakfast]

Originally Posted by Uncwilly

I suggest that other tasks you could use those machines for (in my preferred order):
P-1 (if they have a good amount of memory and banked right)
LL-DC (this will help the project along and the more help we get the there the better)
PRP-DC (there are fewer of these to do and you may wind up running out in a year or two).
CERT's (you could set them up to run certs almost exclusively)
ECM (if they have reasonable amount of memory)
PRP-CF (with software that will VDF's)

You CPUs, your "$$$$", your choice.

What is the abbreviation for certs in the dropdown box on the Prime95 settings?

[LaurV]

Originally Posted by piforbreakfast

What is the abbreviation for certs in the dropdown box on the Prime95 settings?

None. There is another separate checkbox a little bit below, which is checked by default. You only have it in the most recent versions of P95, older versions don't have it.