[QUOTE=kladner;490653]So, at this point is it more helpful for me to aim my ~800~gd/d at 75 bit or 76 bit targets?[/QUOTE]

Sorry for the lurking; been doing a massive plumbing project (seriously). When the girlfriend's away, the boy will do work which would send her into conniptions if she was around... :wink:

At this point in time it doesn't really matter all that much, thanks to many coming forward with some serious GPU compute to deal with the crunch. Thanks guys!!!

The reason to go to 75 is it would give us the option of giving the P-1'ers such candidates to work if we ran out of 76s, but that no longer looks likely. The reason to go to 76 is it "gets them off our books", and back to Primenet to manage.

TL;DR: Entirely up to you. Those taken to "only" 75 will be further worked by others in the not too distant future.

[QUOTE=LaurV;490702]Hey Chris, can you fix the individual cost per factor pages? they stop somewhere at 78M or so, and we already have a lot of activity in 92M. Maybe you can extend to 100M or so? And what about a third table with a range in 332M?[/QUOTE]

Ah, thanks for pointing that out. I'll look at having that done in the next couple of days.

[QUOTE=chalsall;490793]Sorry for the lurking; been doing a massive plumbing project (seriously). When the girlfriend's away, the boy will do work which would send her into conniptions if she was around... :wink:

At this point in time it doesn't really matter all that much, thanks to many coming forward with some serious GPU compute to deal with the crunch. Thanks guys!!!

The reason to go to 75 is it would give us the option of giving the P-1'ers such candidates to work if we ran out of 76s, but that no longer looks likely. The reason to go to 76 is it "gets them off our books", and back to Primenet to manage.

TL;DR: Entirely up to you. Those taken to "only" 75 will be further worked by others in the not too distant future.[/QUOTE]
Cool! I will upgrade the next round of assignments to 76.

noob questions
 

Hi! I'm new here. I have a few questions.

First: My understanding is that gpu72 assigns TF work to higher-exponent factors than does primenet because for some reason primenet lowballs its estimate for what's worth it---is that right?

Second: I am wondering what the difference is between P-1 work assigned by primenet and by gpu72. What is meant by "primenet is catching up to gpu72"? Where could I find graphs evincing that information, or generally information about "where the fronts are"?

Thanks!

[QUOTE=penlu;490998]Hi! I'm new here. I have a few questions.
[/quote]
Welcome!
[Quote]
First: My understanding is that gpu72 assigns TF work to higher-exponent factors than does primenet because for some reason primenet lowballs its estimate for what's worth it---is that right?[/quote]
The primenet TF limit is for CPUs. GPUs are much better in TF, so it is worthwhile to TF to a higher bitlevel.

The original goal of GPU72 was to TF all exponents to 72 bits before they would be assigned for LL testing. Now in the higher ranges and with faster GPUs we can TF even deeper to for instance 76bits.
[Quote]
Second: I am wondering what the difference is between P-1 work assigned by primenet and by gpu72. What is meant by "primenet is catching up to gpu72"? Where could I find graphs evincing that information, or generally information about "where the fronts are"?
Thanks![/QUOTE]
P-1 assignments should be the same wether you get them from Primenet or GPU72.

It means Primenet completes/assigns more exponents for LL testing per day than GPU72 completes TFing exponents.

The work distribution page on mersenne.org should give you a reasonable indication where the LL / DC and P-1 wavefronts are. For TF GPU72 has a couple of pages.

[QUOTE=penlu;490998]Hi! I'm new here. I have a few questions.[/QUOTE]

We welcome you, and your questions (sincerely).

[QUOTE=penlu;490998]First: My understanding is that gpu72 assigns TF work to higher-exponent factors than does primenet because for some reason primenet lowballs its estimate for what's worth it---is that right?[/QUOTE]

As Victor correctly answered, Primenet doesn't "lowball" so much as it is mostly concerned about CPU workers. GPUs only recently came onto the scene (read: about five or so years ago now) and with some very skilled programmers it was demonstrated as a powerful force to assist the Great Internet Mersenne *Prime* Search. When correctly balanced, finding factors helps find Primes.

[QUOTE=penlu;490998]Second: I am wondering what the difference is between P-1 work assigned by primenet and by gpu72. What is meant by "primenet is catching up to gpu72"? Where could I find graphs evincing that information, or generally information about "where the fronts are"?[/QUOTE]

Mostly there is little difference between P-1 work assigned by GPU72 and Primenet. But since GPU72 completes the optimally Trial Factored candidates before releasing them back to Primenet to manage, it holds back a few (currently 1,000) to assign to P-1 workers who choose to work through GPU72 because they tend to be a bit more disciplined (read: actually complete the work they promise to do in a timely manner).

Lastly, what was meant by "Primenet is catching up" is we were being outrun by Primenet's P-1'ers. As in, more candidates were being P-1'ed than were being TF'ed "optimally". We take some joy in doing "just in time" work, but sometimes the temporal buffer gets a bit tight...

I hope that helps answer any questions you might have. Please ask further questions if you have them.

Thanks all for the answers, that clarifies things for me!

I am understanding that primenet is beginning to approach the point at which P-1 work will go out for exponents that haven't been optimally trial factored yet---does it mean that it'd be more helpful for me to direct my CPU power to tasks other than P-1?

[QUOTE=penlu;491023]I am understanding that primenet is beginning to approach the point at which P-1 work will go out for exponents that haven't been optimally trial factored yet[/QUOTE]

No, don't presume that. Thanks to some very cool dudes who have space heaters in their offices, we're good. (That's meant to be funny, and ironic, at the same time....)

To add to it, as we go higher in the exponent list, P-1 is getting harder, and TF is getting easier. This works in our favor (well... somehow...)

[QUOTE=LaurV;491077]...higher in the exponent list... TF is getting easier[/QUOTE]* for a given TF bitlevel. When it gets too easy, we bump the level to make it hard again. :smile:

Now I have another question---why is gpu72 handing out TFing for ~93M only up to 72 bits, esp. when TF to a given bitlevel is easier at higher exponents? Are we done with <90M for now?

edit I have found the appropriate information---1. I should tell mfloop to fetch higher, 2. no