[QUOTE=James Heinrich;119333]The factor I found was worth 1.4583GHz-days; the 2 days I spent looking (but not finding) the other factor was worth 0.0000GHz-days. Is that the intended value of NF-PM1?[/QUOTE]

Bug fixed. Thanks.

[QUOTE=Mini-Geek;119334]I have 1828 numbers that have found factors, and 1477 that haven't. It gives 0.0016 GHz days for all of the ones with factors, and .1815 for the ones without factors.[/QUOTE]

That could well be correct. A 55-bit factor is worth half of a 56-bit factor which is worth half of a 57-bit factor. With TF-LMH you are finding many 28-48 bit factors which from the above formula you can see are pretty much worthless. If you assume the factors found are uniformly distributed over the 28 to 57 bit range (they're not, they're actually skewed to the smaller factors) and only the last 3 or 4 bit levels will have any significant contribution to your CPU total then you can see the the factors-found CPU total being a 100 times smaller than the factors-not-found CPU total is probably right.

[quote=Prime95;119341]That could well be correct. A 55-bit factor is worth half of a 56-bit factor which is worth half of a 57-bit factor. With TF-LMH you are finding many 28-48 bit factors which from the above formula you can see are pretty much worthless. If you assume the factors found are uniformly distributed over the 28 to 57 bit range (they're not, they're actually skewed to the smaller factors) and only the last 3 or 4 bit levels will have any significant contribution to your CPU total then you can see the the factors-found CPU total being a 100 times smaller than the factors-not-found CPU total is probably right.[/quote]
Ok, I guess that's all it is, but when you consider all the computer is doing with it whether it finds a factor of 3 or no factor at all (i.e. getting assignment from server, putting it in the list, doing the actual TFing, writing the results, etc.), it really still takes a while (not 1% of the time, by any means, but shorter) total for that number...but I guess that time would be overly difficult to calculate. I'm just thinking that it's still underestimating how much time the computer actually does doing things for it vs. the time of the actual TF itself.

[QUOTE=Mini-Geek;119356]I'm just thinking that it's still underestimating how much time the computer actually does doing things for it vs. the time of the actual TF itself.[/QUOTE]

Yes, using the entire TF-LMH process to find factors below 2^4x is a grossly inefficient way to accomplish this task. The SQL lookups and TCP/IP costs far outweigh the time spent finding these trivial factors. On the other hand, it does allow us to pound the server and find bugs.

[quote=Prime95;119363]Yes, using the entire TF-LMH process to find factors below 2^4x is a grossly inefficient way to accomplish this task. The SQL lookups and TCP/IP costs far outweigh the time spent finding these trivial factors. On the other hand, it does allow us to pound the server and find bugs.[/quote]
For testing purposes this is good, but once the real new Prime95/Net goes up, I'd hate to have people doing that make it go slow from hammering the server on something so inefficient. Perhaps leave the TFing of high numbers to low limits to the LMHers, then...or use it on PrimeNet, but in some way that's not, as you put it, grossly inefficient (maybe somehow efficiently hand out a group of them at once?)...is the TF-LMH option planned to be in the final Prime95 v25/PrimeNet v5?

I decided to try P-1 factoring. I only check the work it was actually running today. When I check the status it says I'm doing a Double-Check on exponent 79299959. I'm about 0.79% in. Not sure if I should report this to Scott?

[QUOTE=Mini-Geek;119367]Is the TF-LMH option planned to be in the final Prime95 v25/PrimeNet v5?[/QUOTE]

Yes. Once the initial LMH run over all exponents is done (I think we're factoring to 2^58 now?) then LMH won't hammer the server and it won't be grossly inefficient. It's only those real puny factors that are inefficient.

[QUOTE=Bundu;119378]I decided to try P-1 factoring. I only check the work it was actually running today. When I check the status it says I'm doing a Double-Check on exponent 79299959. [/QUOTE]

Attention all trial factorers (not LMH), P-1'ers, and LLers. If you get an exponent that is not between 46,000,000 and 46,010,000 then please exit prime95, delete the exponent from the worktodo.txt file, and restart prime95.
The results database on the v5 server is very out-of-date. I manually brought the 46 million range up-to-date so that this kind of testing will result in real work.

The double-check assignment is the default backup work type if there are no exponents trial factored enough to begin P-1 factoring. M79,299,959 was marked as available for double-checking because of some other tests I was doing.

[QUOTE=Bundu;119378]I decided to try P-1 factoring. When I check the status it says I'm doing a Double-Check on exponent 79299959.[/QUOTE]

BTW, send me your prime.log file. There should have been some P-1 exponents available for assignment.

[QUOTE=Prime95;119379]I think we're factoring to 2^58 now?[/QUOTE]

Actually we're factoring to 2^57 now, at least on my system.

Can you expain the memory usage for ECM testing? Is there a difference if I say I have 128MB of memory vs 768MB, are the bounds the same, speed difference, etc). tks

[QUOTE=James Heinrich;119335]hmmm... I was (somewhat) certain this exponent was doing P-1 under the assumption that no factors < 2^66, and yet the factor found is 65.8 bits :unsure:[/QUOTE]

Bug fixed in 25.6 - nice find. SSE2 trial factoring above 2^64 is broken in 25.5.