[QUOTE=Uncwilly;304503]:party:[/QUOTE]

Let's see: 1,000,000 GhzDays would only take my farm 50 years. WOW

[QUOTE=davieddy;304543]Enjoy yourself[/QUOTE]

[URL="http://www.youtube.com/watch?v=mXsmyLtpxlA"]Rest assured, we are.[/URL]

We're now (just) ahead of the LL wave front!!!
 

Hey all.

Just to report... If you look at the [URL="http://www.gpu72.com/reports/available/"]Available Trial Factoring[/URL] report and compare it with the [URL="http://www.mersenne.org/primenet/"]PrimeNet Exponent Status Distribution[/URL] report, you can see that GPU to 72 is now mostly (just) ahead of the LL wave front for new assignments. At least, as far as Trial Factoring is concerned.

I have added two new columns to the report, showing TF and P-1 assignments. As you can see, below 58M most of the assignments are for P-1, while PrimeNet is currently handing out assignments below 56M, and will probably continue to do so for some time. Of course, PrimeNet will continue to hand out assignments within the wave at around midnight UTC which need TFing to 72 and possibly P-1'ing-- Spidy will usually get these and make them available for appropriate processing.

In light of this, and based on [URL="http://mersenne-aries.sili.net/cudalucas.php?model=12&granularity=0.1"]James' analysis[/URL] which shows the cost/benefit cross over point for GPU TFing to 73 to actually be 57.5M or so (or even down to 52.3M if you consider 72.5 bits to be the cross-over), I have shifted the "release level" down to 57M.

My plan is to bring in candidates currently at 72 bits without P-1 done below 58M a few hundred at a time for those willing to take them to 73 bits. This will be done starting at the top, and working down until we meet the wave front again. (Note that all LL candidates not currently assigned in 57M are already at 72 bits.)

The "What Makes Sense" option will now assign work above 57M to 73 if that's the pledge level. As always, people are welcome and free to do whatever work type and range they like. If you don't want to do 72 to 73, just pledge to go to 72 and choose the "What Makes Sense" option, or pledge to 73 and choose the "Lowest TF Level" option, or set a specific range, etc, etc, etc...

I think this is good and impressive news. Who would have thought only ten months ago that we would have accomplished as much as we have? I certainly never imagined it! :smile:

As always, comments, (constructive) criticism, etc welcome.

:party:[QUOTE=chalsall;304753]Hey all.

Just to report... If you look at the [URL="http://www.gpu72.com/reports/available/"]Available Trial Factoring[/URL] report and compare it with the [URL="http://www.mersenne.org/primenet/"]PrimeNet Exponent Status Distribution[/URL] report, you can see that GPU to 72 is now mostly (just) ahead of the LL wave front for new assignments. At least, as far as Trial Factoring is concerned.

I have added two new columns to the report, showing TF and P-1 assignments. As you can see, below 58M most of the assignments are for P-1, while PrimeNet is currently handing out assignments below 56M, and will probably continue to do so for some time. Of course, PrimeNet will continue to hand out assignments within the wave at around midnight UTC which need TFing to 72 and possibly P-1'ing-- Spidy will usually get these and make them available for appropriate processing.

In light of this, and based on [URL="http://mersenne-aries.sili.net/cudalucas.php?model=12&granularity=0.1"]James' analysis[/URL] which shows the cost/benefit cross over point for GPU TFing to 73 to actually be 57.5M or so (or even down to 52.3M if you consider 72.5 bits to be the cross-over), I have shifted the "release level" down to 57M.

My plan is to bring in candidates currently at 72 bits without P-1 done below 58M a few hundred at a time for those willing to take them to 73 bits. This will be done starting at the top, and working down until we meet the wave front again. (Note that all LL candidates not currently assigned in 57M are already at 72 bits.)

The "What Makes Sense" option will now assign work above 57M to 73 if that's the pledge level. As always, people are welcome and free to do whatever work type and range they like. If you don't want to do 72 to 73, just pledge to go to 72 and choose the "What Makes Sense" option, or pledge to 73 and choose the "Lowest TF Level" option, or set a specific range, etc, etc, etc...

I think this is good and impressive news. Who would have thought only ten months ago that we would have accomplished as much as we have? I certainly never imagined it! :smile:

As always, comments, (constructive) criticism, etc welcome.[/QUOTE]

Since I don't have a suitable GPU, I'm just reading here to follow your progress, but ...

:party:

[QUOTE=cheesehead;304804]Since I don't have a suitable GPU, I'm just reading here to follow your progress, but ...[/QUOTE]You don't need a GPU to join the party! :smile:
I'm nearly even between factors found (under GPU[sup]72[/sup]) by GPU TF (63) vs CPU P-1 (61).

Something is definitely working because for the first time in recent memory there are plenty (10,000 ish) available LL assignments in the wave.

[QUOTE=James Heinrich;304805]You don't need a GPU to join the party! :smile:
I'm nearly even between factors found (under GPU[sup]72[/sup]) by GPU TF (63) vs CPU P-1 (61).[/QUOTE]

And then there are those of us that are doing P-1, low DC and low LL.

If at first you don;t succeed...
 

As soon as Oliver produced mfaktc it was obvious that
GPUs could do a few more bits AHEAD of the wavefront.

(See my "useless" posts thread).

All Chalsall has done is hijack the resources.

Rant over.

D

PS I do find his site both clear and informative.
Different strokes for different folks I guess.

What makes AHEAD of the wavefront any better than BEHIND the wavefront, especially when the "wavefront" tends to leave more than half of the exponents in its wake untested?

[QUOTE=Dubslow;304812]What makes AHEAD of the wavefront any better than BEHIND the wavefront, especially when the "wavefront" tends to leave more than half of the exponents in its wake untested?[/QUOTE]
Shorter time to TF to a given level.
More work saved.
Try to reduce the expired/returned expos.
And even if this didn't happen, Spidy would now be redundant had 53M+ been TFed adequately before being assigned for LL.

For Christ's sake, it is so obvious that adequate TFing should precede LL assignment.

D