[QUOTE=Gordon;412888]4 instances is currently holding at ~50%[/QUOTE]

I haven't tried GMP-ECM on numbers this big, but in my experience ECM does benefit from hyperthreading; that is, you might try running 2-3 stage1's on prime95 with 5-6 stage 2's on ECM.

[QUOTE=VBCurtis;412889]I haven't tried GMP-ECM on numbers this big, but in my experience ECM does benefit from hyperthreading; that is, you might try running 2-3 stage1's on prime95 with 5-6 stage 2's on ECM.[/QUOTE]

Windows task manager shows the load spread fairly evenly across all "8" cores, am generating some more stage 1 results and will try running 2 more instances of gmp-ecm.

Before I had pointed out to me the full correct syntax for the parameters (and so redoing stage 1 again) I was also using prime95 to do ecm work via primenet, if you will remember...

M85027 - Step 2 took 1306804ms

just running the 4 x gmp-ecm

M85121 - Step 2 took 829410ms

Nice!
This is how we all learn :wink:

[QUOTE=Gordon;412896]Windows task manager shows the load spread fairly evenly across all "8" cores [/QUOTE]

OK, that explains the 50% load with 4 instances. I always forget HT, since I never had such thing on my machines.

[QUOTE=VBCurtis;412889]I haven't tried GMP-ECM on numbers this big, but in my experience ECM does benefit from hyperthreading; that is, you might try running 2-3 stage1's on prime95 with 5-6 stage 2's on ECM.[/QUOTE]

6 instances of gmp-ecm

cpu - 78%

ram - 23-26 gig

see attached

At ~15 minutes per curve, throughout is 24*4*6 = 496 curves/day

[QUOTE=Gordon;412968]6 instances of gmp-ecm

cpu - 78%

At ~15 minutes per curve, throughout is 24*4*6 = 496 curves/day[/QUOTE]

If you run 2 prime95's for stage1 at the same time, you'll sustain both stages at 400-450 curves/day. Given your timings, looks like 1 Prime95 can do stage 1 4 times faster than GMP-ECM does stage 2. So, 2 prime95 + 6 GMP-ECM will only slowly build up extra stage1 residues.

You could consider decreasing the B2 value that GMP-ECM uses to be roughly 3x the time of stage 1 so that 2 prime95's + 6 GMP-ECM's are in balance, or decrease it further such that stage 2 time is 5/3rds stage 1 time (and run 3x P95 + 5x GMP-ECM).

I suggest some experimenting to find which B2 leads to the highest fraction of a t35 per day, rather than the highest number of curves. There's lots of knobs to tinker with! If you try this, be warned that GMP-ECM has only coarse control over B2; it might round your requests to 6G, or 9G, or 12G, etc.

[QUOTE=VictordeHolland;412885]You might want to experiment with the B2 values a bit, you are now spending 6x longer in stage 2 than in stage 1 (stg1 201 sec vs. stg2 1271 sec). The 'rule of thumb' is spending the same time in stage 2 as in stage 1.[/QUOTE]

I've read on here that time spent in stage 2 should be about 0.7x time in stage 1 (RDS?), that would imply reducing B2 from it's current 1B to about 100M.

I don't care about the memory usage - 6 ecm's at 1B is only 24gb of ram, I understand the lower bound means likely to find smaller factors, but they will churn through much quicker.

A search on the internet doesn't seem to turn up much guidance other than rule of thumb B2=100*B1, anyone got any pointers to some more detailed analysis?

[QUOTE=Gordon;413286]I've read on here that time spent in stage 2 should be about 0.7x time in stage 1 (RDS?), that would imply reducing B2 from it's current 1B to about 100M.

I don't care about the memory usage - 6 ecm's at 1B is only 24gb of ram, I understand the lower bound means likely to find smaller factors, but they will churn through much quicker.

A search on the internet doesn't seem to turn up much guidance other than rule of thumb B2=100*B1, anyone got any pointers to some more detailed analysis?[/QUOTE]

The 100*B1 rule of thumb is for the lower-memory, slower, non-GMP-ECM method for stage 2. That's why Prime95 still uses it; P95 uses that low-memory algorithm.

The empirical solution is to use the -v flag with GMP-ECM, and try a variety of B2 values to see which value minimizes the expected time to complete the T-level you're interested in (I think t35, B1 = 1M?). The problem is that GMP-ECM has no way to know how long Stage 1 took, so that process doesn't work with your combination of programs. If I were you, I'd run GMP-ECM on a single curve for both stage 1 and stage 2, adjust for the time stage 1 takes on P95 (for M1277, it's about 10% faster at B1 = 11M), and see what B1 minimizes time to complete T35.

I happen to enjoy doing such experiments, and I have sufficient memory to do so; if you like I'll run the tests for you and report my results and the process I used (for public-nitpicking purposes, 'cause "trust me" isn't good enough). Let me know what specific candidate you'd like me to test (the results should hold across a pretty wide range of inputs, because the options for B2 are rather coarse).

[QUOTE=VBCurtis;413305]and I have sufficient memory to do so[/QUOTE]
[offtopic][thinking]I have a very good memory too, but I don't remember where I put it...[/thinking][/offtopic]

[QUOTE=LaurV;413310][offtopic][thinking]I have a very good memory too, but I don't remember where I put it...[/thinking][/offtopic][/QUOTE]I have lots of very good memories also. I'm now waiting for Alzheimer's to set in so that I can recall them.

M40157713 has a factor: 281921279054741252950391
77.89 bits

Trial factoring