GPU to 72 status...

[flashjh]

Originally Posted by James Heinrich

This is in reference to "P-1 done poorly" exponents, so yes: PrimeNet will ignore the new result if the new B1 is smaller than the old B1, even if the old P-1 test had no stage2 and/or the overall probability was lower. For example, if an exponent was done moderately-well at B1=B2=105k, and you re-do P-1 at B1=100k,B2=2M, your new result would be ignored because PrimeNet decides which P-1 is "better" solely on B1.

Right, and for the people interested in redoingthe "P-1 done poorly", what is the correct way? I don't think I'll get into it because I'm more suited for TFing, but it seems like it will require making B1 larger than the original to get any credit. Of course, if the sole purpose is to find factors, then the credit doesn't matter and PrimeNet will accept factors for those exponents.

[Dubslow]

Originally Posted by flashjh

Of course, if the sole purpose is to find factors, then the credit doesn't matter and PrimeNet will accept factors for those exponents.

But it's more efficient if you can somehow mark that Stage 2 has been performed on this exponent, which isn't recorded unless you have the higher Stage 1 as well. It's not just the credit, it's so that (e.g.) James' site can see that it's been done so someone else doesn't do it again.

[chalsall]

Originally Posted by Dubslow

But it's more efficient if you can somehow mark that Stage 2 has been performed on this exponent, which isn't recorded unless you have the higher Stage 1 as well. It's not just the credit, it's so that (e.g.) James' site can see that it's been done so someone else doesn't do it again.

Yes.

If we offer this work type (and it will be some time before we do, if ever -- we have a lot of more important work to do) the system will have to offer candidates which have had P-1 done really poorly. And the assignment form will have to calculate what the bounds Prime95/mprime will do based on the "Effort" the worker is pledging to ensure that B1 is going to be higher than the current.

Otherwise the system won't be able to detect that the work has been completed. Unless, of course, George and/or Scott update PrimeNet to record all P-1 work done. Based on how low a priority this type of work is, I doubt that will happen any time soon.

[James Heinrich]

Originally Posted by chalsall

the system will have to offer candidates which have had P-1 done really poorly.

There's no great need to incorporate this into GPU72 anytime soon, for us few who are interested in it, http://mersenne-aries.sili.net/p1small.php already offers up such really-badly-done P-1 assignments for the taking. Only downside is no GPU72 credit.

[petrw1]

http://www.gpu72.com/reports/workers/

[chalsall]

Hey all. This is still very early beta, but I thought some of you might enjoy some videos of the trial factoring efforts. Use the zoom, "<---" and "--->" links to navigate around.

Please note that at this point in time I only have webm and ogv formats; not H263. Thus, some might not be able to view them on Internet Exploder. Also, I don't yet have "1M" drill-down resolution rendered. Those should be online in the next couple of days.

[Uncwilly]

Originally Posted by chalsall

Hey all. This is still very early beta, but I thought some of you might enjoy some videos of the trial factoring efforts. Use the zoom, "<---" and "--->" links to navigate around.

I rather like this one. At the end you can see the power of a GPU at working buzz-sawing through exponents. http://www.mersenne.info/trial_facto...o/2/330000000/

[Dubslow]

Originally Posted by Uncwilly

I rather like this one. At the end you can see the power of a GPU at working buzz-sawing through exponents. http://www.mersenne.info/trial_facto...o/2/330000000/

Nice going LaurV!


Edit: Can we see the LL status line graph for 0-100M as a video?

[chalsall]

Originally Posted by Uncwilly

I rather like this one. At the end you can see the power of a GPU at working buzz-sawing through exponents.

Yeah. And because 332M is a special range, I've rendered the 1M resolution video ahead of the batch job.

BTW, for anyone not having success viewing these, I'd recommend you try using Google Chrome until I get the H264 versions online.

[Dubslow]

Originally Posted by chalsall

Yeah. And because 332M is a special range, I've rendered the 1M resolution video ahead of the batch job.

What the hax LaurV? Any chance we can get a slow-mo on that chalsall?

[chalsall]

Originally Posted by Dubslow

Can we see the LL status line graph for 0-100M as a video?

Sure....

Note that there are some glitches early on in the videos. I'll try to clean these up in the near future.

Also, the frame rate is currently 10 FPS; three seconds a month. Would 6 FPS (five seconds a month) be better?