GPU to 72 status... - Page 40

Dubslow · 2012-01-31, 23:57

Well, I meant catch up to these new bounds. If we're ahead of the wavefront, there's no point in reserving expos as LLs -- just as TF.

Also, could you tell me which report, as far as 72 per day? The weekly progress graph gives a good idea, though I imagine it's not strictly GPU272 workers -- pretty close estimate though. (That's like 400+ per day, though I imagine in the 55+ range less of the work is us. Perhaps 350 a day for the project?)

chalsall · 2012-02-01, 00:22

Quote:

Originally Posted by Dubslow

Well, I meant catch up to these new bounds. If we're ahead of the wavefront, there's no point in reserving expos as LLs -- just as TF.

You continue to fail to understand.

We are doing more than twice as many TF to 72 as the overall system is LLing.

We are still working within the wave.

Quote:

Originally Posted by Dubslow

Also, could you tell me which report, as far as 72 per day? The weekly progress graph gives a good idea, though I imagine it's not strictly GPU272 workers -- pretty close estimate though. (That's like 400+ per day, though I imagine in the 55+ range less of the work is us. Perhaps 350 a day for the project?)

Compare this and this, and this and this.

But, if you (or anyone) wants to do your own thing, go to it....

diamonddave · 2012-02-01, 04:30

Quote:

Originally Posted by chalsall

Compare this and this, and this and this.

Third link is broken

Edit(by CH): Thanks. Fixed.

LaurV · 2012-02-01, 06:58

Not broken, just inserted two times by mistake. Cut it in half and it will work

ckdo · 2012-02-01, 09:34

Quote:

Originally Posted by chalsall

You continue to fail to understand.

Me too. I can't second your definition of GPUto72's effectiveness. Compare the number of exponents factored with the number of (successful) LL tests run.

For the last month that gives 207+671=878 vs. 57+952+13+4495=5517 (for LLTF, i.e. 40M-60M). Under the (somewhat faulty) assumption that every factor saves 2 LL tests in this range, we would have had to factor like 3x as many candidates to keep up.

The numbers are of course even (much much) worse for DCTF: 185 candidates factored vs. 4103 that should have been. (So, do more DCTF, everyone!

)

axn · 2012-02-01, 09:56

Quote:

Originally Posted by ckdo

Me too. I can't second your definition of GPUto72's effectiveness. Compare the number of exponents factored with the number of (successful) LL tests run.

If you're going to do that, you must compare the number of exponents factored, with the number of LLs GPUto72 team could have cleared by LL. Not whole of GIMPS. This is, after all, not a competition between GPUto72 and GIMPS -- they're complementary efforts.

The point of GPUto72 was to factor LL's to 72 bits (way over the default primenet limit of 68(?) bits) and release the survivors to the rest of GIMPS for LL-ing. The whole thing was meant to tactically focus GPUs (whose strong suit is TF), to advance the GIMPS LL-wave. But for that to effectively happen, GPUto72 must be able to "keep up" with the LL-wave. Earlier in the project, it didn't have the firepower to do so; now it does.

chalsall · 2012-02-01, 14:55

Quote:

Originally Posted by axn

If you're going to do that, you must compare the number of exponents factored, with the number of LLs GPUto72 team could have cleared by LL. Not whole of GIMPS. This is, after all, not a competition between GPUto72 and GIMPS -- they're complementary efforts.

Exactly!!!

Quote:

Originally Posted by axn

The point of GPUto72 was to factor LL's to 72 bits (way over the default primenet limit of 68(?) bits) and release the survivors to the rest of GIMPS for LL-ing. The whole thing was meant to tactically focus GPUs (whose strong suit is TF), to advance the GIMPS LL-wave. But for that to effectively happen, GPUto72 must be able to "keep up" with the LL-wave. Earlier in the project, it didn't have the firepower to do so; now it does.

Indeed. And we are contining to gain firepower every week.

Additionally, as the exponents get larger, the trial factoring cost (for a particular bit level) decreases while the LL cost increases.

Lastly, my thinking had to do with the fact we're now completing more than twice as many candidates than are being LLed. This is the magic "cross-over" point where we start pulling ahead in the "wave" (because approximately half of all LL assignments are never completed).

To be clear, "Spidy" is not going to get aggressive with it's reservations. However, when it finds a candidate only TFed to 70 or below above 55M, it will keep it until it's taken to 72.

James Heinrich · 2012-02-01, 15:23

@chalsall: could you create a wiki entry for GPU to 72? Something along the lines of my site's entry that briefly describes what it is and what it does.

chalsall · 2012-02-01, 15:35

Quote:

Originally Posted by James Heinrich

@chalsall: could you create a wiki entry for GPU to 72? Something along the lines of my site's entry that briefly describes what it is and what it does.

Sure. Good idea. Thanks.

Dubslow · 2012-02-01, 16:51

http://gpu72.com/reports/workers/ll/graph/3-12/

Heehee, that's rather funny looking

chalsall · 2012-02-01, 17:00

Quote:

Originally Posted by Dubslow

Heehee, that's rather funny looking

But accurate.

Consider it modern art....

2012-01-31, 23:57	#430
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1C35₁₆ Posts	Well, I meant catch up to these new bounds. If we're ahead of the wavefront, there's no point in reserving expos as LLs -- just as TF. Also, could you tell me which report, as far as 72 per day? The weekly progress graph gives a good idea, though I imagine it's not strictly GPU272 workers -- pretty close estimate though. (That's like 400+ per day, though I imagine in the 55+ range less of the work is us. Perhaps 350 a day for the project?) Last fiddled with by Dubslow on 2012-02-01 at 00:02

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Status	Primeinator	Operation Billion Digits	5	2011-12-06 02:35
62 bit status	1997rj7	Lone Mersenne Hunters	27	2008-09-29 13:52
OBD Status	Uncwilly	Operation Billion Digits	22	2005-10-25 14:05
1-2M LLR status	paulunderwood	3*2^n-1 Search	2	2005-03-13 17:03
Status of 26.0M - 26.5M	1997rj7	Lone Mersenne Hunters	25	2004-06-18 16:46

2012-02-01, 06:58	#433
LaurV Romulan Interpreter Jun 2011 Thailand 25BF₁₆ Posts	Not broken, just inserted two times by mistake. Cut it in half and it will work

2012-02-01, 15:23	#437
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 23×149 Posts	@chalsall: could you create a wiki entry for GPU to 72? Something along the lines of my site's entry that briefly describes what it is and what it does.

2012-02-01, 16:51	#439
Dubslow Basketry That Evening! "Bunslow the Bold" Jun 2011 40<A<43 -89<O<-88 1110000110101₂ Posts	http://gpu72.com/reports/workers/ll/graph/3-12/ Heehee, that's rather funny looking