Fact or Friction

[storm5510]

mfaktc: It would appear that the larger number of manual assignments a person gets, the larger the end factor tends to be. I take 10, it is 72. I take 15 it is 73. This 1080 goes through them rather rapidly. So, what happens if I take 50, or more?

[Mark Rose]

Originally Posted by storm5510

mfaktc: It would appear that the larger number of manual assignments a person gets, the larger the end factor tends to be. I take 10, it is 72. I take 15 it is 73. This 1080 goes through them rather rapidly. So, what happens if I take 50, or more?

I usually queue a day's worth. Some people will grab enough for a week. If your GTX 1080 can chew through 1100 GHz-d/day, then grab the appropriate amount of GH-d.

GPU72 work tends to be slow, so I think I've only had in the range of 10k assignments there. Using the manual GPU assignment page I've had over 100k assignments.

[petrw1]

Originally Posted by storm5510

mfaktc: It would appear that the larger number of manual assignments a person gets, the larger the end factor tends to be. I take 10, it is 72. I take 15 it is 73. This 1080 goes through them rather rapidly. So, what happens if I take 50, or more?

Are you using misfit?
It works fine for me and you can set the bit level and the exponent range and whether you get from GIMPS or GPU72.

I took 2,000 assignments before I went on vacation.

[storm5510]

Originally Posted by Mark Rose

I usually queue a day's worth. Some people will grab enough for a week. If your GTX 1080 can chew through 1100 GHz-d/day, then grab the appropriate amount of GH-d.

GPU72 work tends to be slow, so I think I've only had in the range of 10k assignments there. Using the manual GPU assignment page I've had over 100k assignments.

I can do a 2⁷³ in seven minutes in the current assignment range, 181 million. Just under four minutes for a 2⁷². Many times, the end bits are not uniform. Is there a specific way you calculate this?

[chalsall]

Originally Posted by storm5510

I can do a 2⁷³ in seven minutes in the current assignment range, 181 million.

Just pointing this out... The LLTF work GPU72 is handing out, to 76 bits in the 87M range, comprises 87.4 GHzD of effort (from 75 bits). So your 1080 would be able to do about 12.6 a day.

While you wouldn't find as many factors per temporal period, this work is of more immediate need to GIMPS; 181M probably won't be LL'ed for another decade or so.

Work is also available in the 92M range to 72 bits; ~5.2 GHzD per run. At least this effort will be utilized in a couple of years.

Entirely up to you, of course. It's your kit!

[storm5510]

Originally Posted by petrw1

Are you using misfit?

No, but I have seen the name here and there.

Originally Posted by chatsall

Just pointing this out... The LLTF work GPU72 is handing out, to 76 bits in the 87M range, comprises 87.4 GHzD of effort (from 75 bits). So your 1080 would be able to do about 12.6 a day.

While you wouldn't find as many factors per temporal period, this work is of more immediate need to GIMPS; 181M probably won't be LL'ed for another decade or so.

Work is also available in the 92M range to 72 bits; ~5.2 GHzD per run. At least this effort will be utilized in a couple of years.

Entirely up to you, of course. It's your kit!

Actually, I thought it would not be needed for 15 years, or more.

I looked at the link. Very interesting. I plan to look at this in detail.

[petrw1]

Originally Posted by storm5510

No, but I have seen the name here and there.



Actually, I thought it would not be needed for 15 years, or more.

I looked at the link. Very interesting. I plan to look at this in detail.

http://www.mersenneforum.org/forumdisplay.php?f=103

Or I welcome more help here:
http://www.mersenneforum.org/showthread.php?t=22476

[storm5510]

Originally Posted by petrw1

http://www.mersenneforum.org/forumdisplay.php?f=103

Or I welcome more help here:
http://www.mersenneforum.org/showthread.php?t=22476

I created an account on GPU to 72. I really like being able to specify assignments based on GHz-d/day, and factoring limit. As a test, I requested enough to run about 8 hours. mfaktc is happily churning away at 1,100+ GHz-d/day. All the assignments are in the 92 million range.

I did a cross check on mersenne.org. These particular exponents, no attention since 2015. I found that rather astonishing.