Is this updated anymore?

You asked for it, you got it!

[CODE]
Bits Exponents (11/08/03)
57 2688 (-2936)
58 362097 (-6998)
59 288546 (-21982)
60 638679 (22925)
61 27134 (1147)
62 43626 (3781)
63 8364 (2932)
64 4727 (-59)
65 9
66 8
67 6
68 7792 (131)
69 44
70 7
71 6
72 18
73 0
74 1
Total 1383752 (-1023)
[/CODE]

Another little statistics :
[CODE]
Status Average depth no factors

30 Jul 2003 58.94 bit 1395248
09 Nov 2003 59.42 bit 1383752
[/Code]
So, on the average during the last 14 weeks the factoring depth has increased by 0.48 bit. If we assume the same for the next time the average would increase by 1 bit within about 30 weeks and the cost for LMH is therefore doubled in the same time.
That's much more than predicted by Moore's Law and above the actual trend of CPU speeds. So we have to expect a gradual decline in the next year (if not much more computers involved in LMH).
Nevertheless, I guess LMH can achieve an average level of 62 bit within the next 2-2.5 years.

11496 new factors were found , ie. 821 per week or 117 per day. The GIMPS status page exhibit 18674 new factors in that time (including TF,LL,DC and others), so we can assume that at least 60% of the factors reported were obtained by LMH !

[QUOTE][i]Originally posted by hbock [/i]
[B]
[CODE]
Status Average depth no factors

30 Jul 2003 58.94 bit 1395248
09 Nov 2003 59.42 bit 1383752
[/Code]
So, on the average during the last 14 weeks the factoring depth has increased by 0.48 bit. If we assume the same for the next time the average would increase by 1 bit within about 30 weeks and the cost for LMH is therefore doubled in the same time.[/B][/QUOTE]
Since no weighting of bit depths according to difficulty is mentioned, I presume these averages gave equal weight to each exponent, regardless of size -- correct?

But we know that raising the TF bit depth for 2^971-1 from 57 to 58 will involve much more work than raising the TF bit depth for 2^9710011-1 from 57 to 58 did ... about 10,000 times more, offset by the ratio of time necessary for each trial division of 2^9710011-1 compared to time for each trial division of 2^971-1. BTW, is this ratio logarithmic by exponent, at least approximately? ~ log 9710011 / log 971 = ~ 2.3 ?

Assuming a log ratio, then what would be the average bit depths given above if the bit depth for each exponent were weighted by (log exponent)/exponent ? And what would that do to the extrapolated prediction for the 62-bit level?

After that, what about incorporating the ratio of time needed for a trial division in the 2^58 range compared to a trial division in the 2^62 and higher ranges? Iteration times increase significantly from 2^62 to 2^63, then again from 2^64 to 2^65. Of course those ratios depend on CPU type ...

Well, the average factored bit depth is just a statistical value for the database (in that case for exponents 25M-79M). Each exponent is equal weighted with 1 and therefore it doesn't say anything about the CPU time spent to reach or increase this value.

At the moment the work is distributed over all ranges. So, if we assume that each exponent is factored to a distinct value i and for more than 90% i is less 62 bit, then (regardless of the CPUs used) the time to bring this level to i+1 is increased by 2 (compared to a former step from i-1 to i and as a first approximation).
Nevertheless, because most of the work is done by raising exponents from the lowest levels to a higher one, it could be that the factoring speed will continue for awhile, even with about the same computing power. But, when all low level exponents (58/59 bit) are at the 60 bit level or more then we have to expect a significant decline, anyway. Other reason is that the organized LMH is too new and therefore the speed is not yet stable and hard to predict.
And, if we have most of the exponents at 62-64 bit the standard CPU used for LMH will maybe an (old) P4, goodness knows ...

A bit of trivia: there are now more exponents on 60 bits than below 60 bits! This is for exponents above 25M, and for all exponents.

Well done everyone!

[code]
Bits Exponents (11/17/03)
57 2688
58 333877 (-28220)
59 264974 (-23572)
60 673915 (35236)
61 31567 (4433)
62 53183 (9557)
63 8863 (499)
64 4722 (-5)
65 9
66 8
67 20 (14)
68 8074 (282)
69 49 (5)
70 7
71 6
72 18
73 0
74 1
---------------
Total 1381981 (-1771)
[/code]

[code]
Bits Exponents (12/02/03)
57 2688
58 306227 (-27650)
59 259499 (-5475)
60 652992 (-20923)
61 76122 (44555)
62 58865 (5682)
63 10320 (1457)
64 4655 (-67)
65 9
66 8
67 20
68 8370 (296)
69 50 (1)
70 7
71 6
72 18
73 0
74 1
---------------
Total 1379857 (-2124)
[/code]

Note that the range 33.7M to 33.8M is now on Primenet, and should not be reserved.

just a couple q's:

what numbers are still at 57 bits?
is someone working on those numbers?

The exponents at 57 bits are mainly, if not all, on primenet awaiting reservation.