New report: Cost per Factor Found...
 

Hey all.

In order to be able to make a more informed decision on exactly how deeply we should be trial factoring in the various ranges, I have created a new report: [URL="http://gpu.mersenne.info/reports/factoring_cost/"]Cost per Factor Found[/URL] broken down by "bit depth" and 1M ranges.

It seems to me that at 53M and above we might want to consider to start going to 73, but only after clearing out everything (or, at least, most everything) below.

Thoughts?

If you make the data any clearer, you will be in danger of realizing I am talking sense.

[QUOTE=chalsall;281386]Hey all.

In order to be able to make a more informed decision on exactly how deeply we should be trial factoring in the various ranges, I have created a new report: [URL="http://gpu.mersenne.info/reports/factoring_cost/"]Cost per Factor Found[/URL] broken down by "bit depth" and 1M ranges.

It seems to me that at 53M and above we might want to consider to start going to 73, but only after clearing out everything (or, at least, most everything) below.

Thoughts?[/QUOTE]

Sorry for the cross posting, post belong here I think,

I'm really surprised by those results. We are finding way more factor then expected, excepted for the 69 bit level in DC.

Also the Factor found reported here doesn't agree with other report on the site.

LL:
[CODE]
Bit Exp. Factor Expected
69 1640 61 24
70 9949 250 142
71 6020 192 85
72 4244 112 59
73 78 3 1
[/CODE]
DC:
[CODE]
Bit Exp. Factor Expected
68 1477 32 22
69 5291 50 77
70 83 4 1
[/CODE]

[QUOTE=diamonddave;281392]Also the Factor found reported here doesn't agree with other report on the site.[/QUOTE]

Whoops... Stupid programmer error... :blush:

Fixed.

[QUOTE=chalsall;281396]Whoops... Stupid programmer error... :blush:

Fixed.[/QUOTE]

Here are my new report with the new Data.


LL:
[CODE]
Bit Exp. Factor Expected
69 1640 17 24
70 9951 106 142
71 6032 68 85
72 4271 38 59
73 78 1 1
230 311
[/CODE]

Finding only 73% of the expected number of factor.


DC:
[CODE]
Bit Exp. Factor Expected
68 1477 20 22
69 5291 45 77
70 83 3 1
68 100
[/CODE]

Finding only 68% of the expected number of factor.

Surprising results...

[QUOTE=diamonddave;281397]Finding only 68% of the expected number of factor.

Surprising results...[/QUOTE]

Adjust your expectations. Chance for a factor within 2[SUP]x[/SUP] and 2[SUP]x+1[/SUP] is 1/x [B]only if no P-1 was done[/B].

Oliver

[QUOTE=TheJudger;281398]Adjust your expectations. Chance for a factor within 2[SUP]x[/SUP] and 2[SUP]x+1[/SUP] is 1/x [B]only if no P-1 was done[/B].

Oliver[/QUOTE]

Can we have clearer data?

[QUOTE=diamonddave;281399]Can we have clearer data?[/QUOTE]

Such as?

I could do as I do on the "Available Assignments" page, and give the option of showing statistics with only "with P-1" and "without P-1". Would this be helpful?

Also, I remember back with the original PrimeNet server that some LMHers were TFing in higher ranges. The server would accept "Factors Found", but not "No Factors Found" results. I don't know what ranges and bit-levels were involved, however.

[QUOTE=chalsall;281400]Such as?

I could do as I do on the "Available Assignments" page, and give the option of showing statistics with only "with P-1" and "without P-1". Would this be helpful?

Also, I remember back with the original PrimeNet server that some LMHers were TFing in higher ranges. The server would accept "Factors Found", but not "No Factors Found" results. I don't know what ranges and bit-levels were involved, however.[/QUOTE]

Yeah the with & without P-1 would give a much clearer picture I think.

[QUOTE=TheJudger;281398]Adjust your expectations. Chance for a factor within 2[SUP]x[/SUP] and 2[SUP]x+1[/SUP] is 1/x [B]only if no P-1 was done[/B].

Oliver[/QUOTE]

For some strange reason, I have a feeling in my water that this guy knows what he's talking about.

David

[QUOTE=davieddy;281402]See above (#2)

[B]Get a brain.[/B]

(Sorry Dave. Like your posts!)[/QUOTE]

I was surprised because right now 2/3 of assignment are No P-1 assignment, it would follow that most of the result would also come from that source.

I thought that P-1 was reducing chance from something like 1/69 to 1/80. Seeing anything north of 1/90 was unexpected to me.