popularity of DC vs LL.

Deamiter · 2003-01-28, 05:13

I'd say I have my computing power split about even between DC and LL, and I'm getting about 2 DCs for every 1st LL I put out. It seems if that was normal, We'd catch up in a couple years or so. Anybody know how the DC vs LL production really compares?

garo · 2003-01-28, 06:56

Well,
The current default is to have all computers over 500Mhz to go to first time tests. In fact, in the year that I have been involved in the project, first time tests have been gaining more ground. Looking at the rate at which exponents are being assigned, for every DC assigned two or more first time tests are assigned. So no, I do not think we are going to catch up. If anything, we are falling behind.

An interesting exercise I've been doing for the past few months has been to see if the leading edge of LL is twice that of the leading edge of DC. I remember that when the first 8M DCs were assigned the leading LL tests were 15.8M. This past week, the leading edge LL has overtaken twice the leading edge of DC for the first time since I've been observing this.

eepiccolo · 2003-01-28, 15:16

I have actually been keeping track of the progress according to George's status page. Over the past two months, the rate of increase in LL'ed exponents per week has been about ten more than the rate of increase of DC'ed exponents. I only have two computers, but the slower one I devote to DCs.
I guess, however, that we really don't want the DCs to catch up to the LLs entirely, because DCs are supposed to be for slower computers anyway.

Darak · 2003-01-28, 16:05

Ok here is roughly what I get (since last server sync) -- Note: I'm completely ignoring the existance of all the 10,000,000 digit numbers since I see that is a different work type & it would greatly skew the results using my method.

Average completed LL = 16182993.84
Average completed DC = 8772462.004

Assuming the Average computer is a Athlon 1 gig machine 24/7. Also, totally disregarding the different FFT transition points (I know this is a biggy but I think we can get the general idea just the same). I may do a more accurate calucalation later when I get more time and take all the FFT ranges into account.

AVG. LL test = 25 days, 20 hours, 21 minutes ~= 620.35 hours
AVG. DC test = 6 days, 19 hours, 16 minutes ~= 163.27 hours

Total LL completed 14848
Total DC completed 8601

1051.5 LL 1.0 gig Athlon CPU Years completed since last server snyc.
160.3 DC 1.0 gig Athlon CPU Years completed since last server snyc.

So a factor of 6.6 X LL CPU Years vs DC CPU Years.

Hence I've had 100% of my CPU power on DC's for almost a year now. As exciting as finding a prime would be I sure don't want us to discover 10 - 15 years after the fact a "goof" caused us to miss one somewhere along the way. Ok plus it's cool to think I'm pulling something close to 2.01% of the total DC work load by myself.

BTW the trend of LL vs. DC isn't new and has existed for quit some time now...

binarydigits · 2003-01-28, 23:53

Darak, I don't believe the "CPU years" comparison is as important as the "number completed" because of the fact that the average DC runs so much faster than the average LL. All we need to consider is that if #DC > #LL then the gap would be narrowing, and if #LL > #DC then it is increasing. So it looks like the number of "once tested but not DC'd" exponents has increased by about 6,000 since the last server sync. Maybe I'll switch a few more of my machines from LL to DC as well. But we also need to make sure that the trend does not reverse for an extended period of time - although it would have to reverse for a VERY long time for the DCs to even come close to the LLs. I suppose the ideal situation would be to keep them roughly the same.

Hades_au · 2003-01-29, 00:04

I'm bucking the trend and doing my bit to try to ensure DC's don't fall any further behind :)

In the 18 or so months I've been with the project, I've completed 47 first-timers compared with 614 DC's. Of those 614 DC's, 164 have been completed since the last server sync.

garo · 2003-01-29, 06:02

Heh,
I've completed 260 DCs compared to zero first-timers ;)

dswanson · 2003-01-29, 06:09

Geez, from the looks of this there must be just a few of us pulling most of the weight for DCs!

My stats since starting in 1998:

560 DC (including 1.85% of those cleared since last synch)
148 LL
55 F

[edit] Whoops, double-bookkept factors found during DCs. Should be 526 DC.

ops:

QuintLeo · 2003-01-30, 23:11

3 LL
16 TF
NO DC.

On the other hand, I'm also working exclusively in the 10,000,000+ range on LLs....

Xyzzy · 2003-01-30, 23:16

Lucas-Lehmer - 9
Double Checking - 84
Factors Found - 82
DC Factors Found - 18

lycorn · 2003-01-31, 10:46

LL - 24 (3 of them 10M+ digits)
DC - 44
F - 3

2003-01-28, 05:13	#1
Deamiter Sep 2002 3²·13 Posts	popularity of DC vs LL. I'd say I have my computing power split about even between DC and LL, and I'm getting about 2 DCs for every 1st LL I put out. It seems if that was normal, We'd catch up in a couple years or so. Anybody know how the DC vs LL production really compares?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Popularity of GIMPS	MooMoo2	Lounge	8	2016-01-26 18:10
GIMPS losing popularity?	ixfd64	Lounge	8	2003-11-15 00:09

2003-01-28, 06:56	#2
garo Aug 2002 Termonfeckin, IE 2²·691 Posts	Well, The current default is to have all computers over 500Mhz to go to first time tests. In fact, in the year that I have been involved in the project, first time tests have been gaining more ground. Looking at the rate at which exponents are being assigned, for every DC assigned two or more first time tests are assigned. So no, I do not think we are going to catch up. If anything, we are falling behind. An interesting exercise I've been doing for the past few months has been to see if the leading edge of LL is twice that of the leading edge of DC. I remember that when the first 8M DCs were assigned the leading LL tests were 15.8M. This past week, the leading edge LL has overtaken twice the leading edge of DC for the first time since I've been observing this.

2003-01-28, 15:16	#3
eepiccolo Dec 2002 Frederick County, MD 2×5×37 Posts	I have actually been keeping track of the progress according to George's status page. Over the past two months, the rate of increase in LL'ed exponents per week has been about ten more than the rate of increase of DC'ed exponents. I only have two computers, but the slower one I devote to DCs. I guess, however, that we really don't want the DCs to catch up to the LLs entirely, because DCs are supposed to be for slower computers anyway.

2003-01-28, 16:05	#4
Darak Oct 2002 2³ Posts	Ok here is roughly what I get (since last server sync) -- Note: I'm completely ignoring the existance of all the 10,000,000 digit numbers since I see that is a different work type & it would greatly skew the results using my method. Average completed LL = 16182993.84 Average completed DC = 8772462.004 Assuming the Average computer is a Athlon 1 gig machine 24/7. Also, totally disregarding the different FFT transition points (I know this is a biggy but I think we can get the general idea just the same). I may do a more accurate calucalation later when I get more time and take all the FFT ranges into account. AVG. LL test = 25 days, 20 hours, 21 minutes ~= 620.35 hours AVG. DC test = 6 days, 19 hours, 16 minutes ~= 163.27 hours Total LL completed 14848 Total DC completed 8601 1051.5 LL 1.0 gig Athlon CPU Years completed since last server snyc. 160.3 DC 1.0 gig Athlon CPU Years completed since last server snyc. So a factor of 6.6 X LL CPU Years vs DC CPU Years. Hence I've had 100% of my CPU power on DC's for almost a year now. As exciting as finding a prime would be I sure don't want us to discover 10 - 15 years after the fact a "goof" caused us to miss one somewhere along the way. Ok plus it's cool to think I'm pulling something close to 2.01% of the total DC work load by myself. BTW the trend of LL vs. DC isn't new and has existed for quit some time now...

2003-01-28, 23:53	#5
binarydigits Aug 2002 2²×13 Posts	Darak, I don't believe the "CPU years" comparison is as important as the "number completed" because of the fact that the average DC runs so much faster than the average LL. All we need to consider is that if #DC > #LL then the gap would be narrowing, and if #LL > #DC then it is increasing. So it looks like the number of "once tested but not DC'd" exponents has increased by about 6,000 since the last server sync. Maybe I'll switch a few more of my machines from LL to DC as well. But we also need to make sure that the trend does not reverse for an extended period of time - although it would have to reverse for a VERY long time for the DCs to even come close to the LLs. I suppose the ideal situation would be to keep them roughly the same.

2003-01-29, 00:04	#6
Hades_au Aug 2002 11 Posts	I'm bucking the trend and doing my bit to try to ensure DC's don't fall any further behind :) In the 18 or so months I've been with the project, I've completed 47 first-timers compared with 614 DC's. Of those 614 DC's, 164 have been completed since the last server sync.

2003-01-29, 06:02	#7
garo Aug 2002 Termonfeckin, IE 2²·691 Posts	Heh, I've completed 260 DCs compared to zero first-timers ;)

2003-01-29, 06:09	#8
dswanson Aug 2002 2³×5² Posts	Geez, from the looks of this there must be just a few of us pulling most of the weight for DCs! My stats since starting in 1998: 560 DC (including 1.85% of those cleared since last synch) 148 LL 55 F [edit] Whoops, double-bookkept factors found during DCs. Should be 526 DC. ops:

2003-01-30, 23:11	#9
QuintLeo Oct 2002 Lost in the hills of Iowa 111000000₂ Posts	3 LL 16 TF NO DC. On the other hand, I'm also working exclusively in the 10,000,000+ range on LLs....

2003-01-30, 23:16	#10
Xyzzy "Mike" Aug 2002 2⁵·257 Posts	Lucas-Lehmer - 9 Double Checking - 84 Factors Found - 82 DC Factors Found - 18

2003-01-31, 10:46	#11
lycorn Sep 2002 Oeiras, Portugal 2⁶·23 Posts	LL - 24 (3 of them 10M+ digits) DC - 44 F - 3