Hey all.

I finally got the Emerald PPS sieve badge, which I want to keep instead of going for double bronze, which is not as pretty. Also, sieving got a little boring.

I decided I wanted to find primes, most importantly a T5K prime (I only have one GFN15). A mega prime would also be nice.

So, I decided to do a "fun" experiment by starting from GFN15, and only moving to a higher n (or project) when I find a prime. So, in an order of 15->16->17low->17mega->18->...

Based on calculations from statistics of earlier primes, the average times to find primes for my rig (GTX1070 and GTX1060) are as follows:

GFN15: 2.3 days
GFN16: 14.25 days
GFN17low: 89.32 days
GFN17mega: 103.43 days
GFN18: 268.03 days
GFN19: 1969.68 days
GFN20+: ???

This means that on average I'd be done in about 1.3 years, but since I could be the double checker, higher primes are harder to find than the statistics say, my computer is not always on, challenges etc., its gonna probably be at least a 2-year project to reach GFN19. After that the chance to be the finder is so low that I'll probably stop there, unless I update my rig or the software gets better.

I will try to keep this thread updated every now and then, wish me luck (and patience)! :)

15->16->17low->17mega->18->...

Haha, yeah ;) Hence the quotes.

But could be that if it goes to something like +200 days I might skip the mega.

Level up!

"Dear Primefinder,

Congratulations! Our records indicate that a computer registered by you has found a unique prime number. This computer is assigned to the Generalized Fermat Prime Search n=15 (GFN-15). Since candidates on this subproject are not large enough to report to the Top 5000 Primes List, your prime is visible immediately.

Workunit 565395795 : 90158524^32768+1 (260670 digits)"

The find was after 13 hours = 0.54 days, so it came up much faster than expected.

Next is GFN16. :)

Neat project, I will definitely be checking out your progress from time to time.

I got hooked on finding primes myself during the TdP this past February after a long time sieving. I found a PPSE after 11 days and then decided to try for a GFN16. I found my first GFN16 just 10 or so hours later! So you may find one sooner than you think, just like with the GFN15 you just found.

Between the 12th and 17th of February I found 3 GFN16s.

I hear ya on the double bronze badge color, I had my systems working over time to get it to double silver. Right now, I am trying to level up my silver badges to gold. I started with the Sophie Germain Prime Search a few days ago and found my first Sophie Germain Prime on the 17th.

Good Luck to you on your experiment!

I'm not trying to be a downer as I have tried to do the 15>16>17meg as well, but the statistics are not quite correct on the stats page. Best of luck, I've gotten a 15 and 16 but not 17 yet.. (well 2 double checkers but no prime finder)

I'm not trying to be a downer as I have tried to do the 15>16>17meg as well, but the statistics are not quite correct on the stats page. Best of luck, I've gotten a 15 and 16 but not 17 yet.. (well 2 double checkers but no prime finder)

I started with the Sophie Germain Prime Search a few days ago and found my first Sophie Germain Prime on the 17th.

The server will also check if P has a twin (so check if P + 1 is also prime).

/JeppeSN

I assume you meant P+2. :)

To be even more pedantic, it's your computer, not the server, that checks if P+2 is prime. For SGS tasks we use an LLR feature that will cause it to run a second test with c changed from -1 to +1 when the 1st test results in a prime.

Thank you for the correction and explanations. I was just happy to have found a prime and it just so happened to be found while running the SGS project. Learning more as I go!

A few days and over 3000 tasks later, no GFN16 prime yet. But it is still over 10 days to the average rate, so no worries yet :)

Level up!

"38542914^65536+1 is prime!!!"

I'm very happy, because this is my first T5K prime :)

The find took almost exactly 6 days, which again makes it faster than the average of 14.25 days. Its even under the median rate of 9.88 days.

The median rate, which basically means that I have a 50% chance to finish during that time, for all the projects are:

GFN15: 1.51 days
GFN16: 9.88 days
GFN17low: 64.61 days
GFN17mega: 58.95 days
GFN18: 206.94 days
GFN19: 625.12 days
GFN20+: ???

During writing this post, I noticed my estimations for GFN19 were of by a little, and the correct numbers are: Average: 1430.58, Minimum 65.84, Maximum 5847.26.

Next up is GFN17, which will be a longer enterprise.

Btw, when moving up the GFN numbers, it becomes evident how bottlenecked the smaller tests are.

In GFN15, GTX1060 and GTX1070 have about same runtime
In GFN16, GTX1070 performs about 12 % faster
In GFN17low, GTX 1070 performs about 23 % faster

From older experience, GFN22 performs about 35-50% faster on GTX1070

I wonder if the software or settings could be optimized somehow for the lower n GFN tests to run faster on higher-end cards.

Congrats on your find!

I have noticed the same thing when running GFN15 on my two GTX1080 cards and on my GTX1060, almost the same run times between the three of them. The larger the GFN task, the more efficient my GTX1080 cards run.

The smaller GFN projects will also slow down any CPU tasks you may have running. I noticed it while running GFN15 together with PPSE tasks. Running GFN16 and PPSE together had less of an impact on the PPSE completion times.

I think I can see the change in my GCW sieve runtimes



All my tasks are in reverse order (1 is the most recent workunit). Around task 37 I switched from GFN16 to GFN17low this morning. Around task 109 I remember switching from 6 tasks running to 8 tasks, which caused a slowdown. The last 8 tasks are before I finished the GFN15. Other variations I dont remember what happened.

Unfortunately, I didn't keep detailed stats on my run times for the PPSE tasks while running GFN15, but IIRC they were impacted on average by about 15 to 25 seconds per task while running concurrently with GFN15 compared to running concurrently with GFN16. There is even a discussion somewhere in the forums in regards to this, just wish I could remember where.

I was wondering why I am the DC for very many tasks in GFN17low.

Looking at the tasks, I found out at least 10 hosts which were producing only or mostly errors on those tasks. This of course causes a lot of tasks sent out have only one valid result and one error. I think most of the machines were from the same user.

Its a little discouraging (for example 19/20 of most recent tasks are DC's), but I'm gonna keep at it for now. Hopefully I will reach up to other normal hosts running this project, so I dont have to do only DC work.

But if it continues, I might switch to GFN17mega, since these hosts dont seem to run that project.

You're not alone, I have noticed the same thing happening to me while running the GFN17low during the past week. My observations are that many of these users are anonymous, have a RAC of 0.09 or less and have nothing but errors being spit out for the GFN15 16 and 17low projects.

I was wondering how a person might inform this/these user/s that they might be running badly configured computers.

Nothing but errors (not one valid) come from the anonymous owner/s of these computers running GFN15 16 and 17low,

418880
418877
418860
473181
418865

(I hope I didn't break any etiquette by listing these computers, it's just so we might be able to inform them of the possible computer trouble they might be having)

I can say it has improved for me somewhat in the past day or so, feel free to look through my results as I have my computers visible.

You guys might want to read http://www.primegrid.com/forum_thread.php?id=4991 before you start griping about how a malfunctioning host is affecting you. The only way it affects you is having to wait longer to get BOINC credit. On GFN, it doesn't affect your prime finding in any way, since we immediately test any primes showing up using LLR (since Genefer only does a PRP test) and so you're not waiting for the wingman to show it as prime. We report the prime immediately with a single user since we've verified it as well.

I've been churning out a lot of gfn17low tasks, but nothing yet.

For the last few days I've been experiencing problems caused by the new nvidia driver version, which crash my driver in a way that makes gpu unusable until I restart the computer. Apparently there is a hotfix on the nvidia forums that I have to try out when I get home.

I'm now running pps sieve on my one functioning gpu machine, in the hope that those tasks dont crash my driver as easily.

It's been 5 weeks, and no hits yet.

Overall, it seems that its been about 400,000 b's since last prime. Its way past the mean of 214,903 and the median of 155,442. It's about halfway to the maximum separation of 810,230, which happened between primes 2639850^131072+1 and 3450080^131072+1.

If I had been doing GFN17mega instead, there was a streak of 3 primes just recently! But you can never know beforehand :)

There was a GFN17low hit yesterday, but I did not happen to be the one testing it. Let's hope there is another hit soon :)

Congratulations, Robish!

Robish, maybe the stereotypes about Irish and luck do have some ground, because you also found a second GFN18

Congrants, haha.

I was the DC on 12512992^131072+1 !

I'm contemplating on moving to GFN17high, but I dont know if that goes with my challenge rules :P Maybe at some point I will.

I moved on to GFN17mega, and received my Sapphire badge now!

Now I'm gonna switch to PPS sieve for a while to rack up some RAC :)

... but my gut (not science) is telling me that 21 will fall before the next 20 and there's that nice shiny sherrifs badge with 21 on it, up for grabs too :)

I tried to make a table:

I've now switched to GFN17 sieving for a while.

On my GTX1070 I process about 130P of sieve a day. With the current removed density, I should find 260 factors a day.

Just running GFN17, I process about 178 work units a day. If you account for the DC, sieving will potentially advance the GFN17 effort about 520 workunits a day (Since a removed candidate accounts for 2 work units). That's about 3x efficiency! No chance of finding primes though, but credit is good.

There is now only about 33 ranges of 200P left for reservation in GFN17, so I think I will work those out before returning to testing.

A little offtopic: Another thing to consider after GFN17 sieve is the GFN21 sieve, since the GFN21 challenge is in a few months. Hopefully others also join in on GFN21 sieving before that :)

I calculated that GFN21 sieve is currently 239 times faster than running testing, so everyone should try to clear out as much candidates as possible before the challenge.

The GFN17 sieve has now been fully reserved. I'll be finished with my remaining sieve tasks in a few days.

I want to change back to GFN17 testing, but I also want to help with the GFN21 sieve effort. I might try to run GFN17low and GFN21 sieve parallel and see how that works out.

Since the GFN17 is now sieved deeper, the chances of prime per task is now slightly higher and the project also advances faster, which are nice. :)

Currently doing GFN21 sieve yields 275P per day. Potentially I could sieve 17000P before the GFN21 challenge.

Because of the screen lag, I will be running GFN17low when using the computer, so should get some of that done too. Maybe I'll get that prime finally!

Still no GFN17low..

It has already been over 8 months, so I think it is already way past the mean rate with my setup (even though I've also been doing other tasks every now and then). I got the one DC, but still been rooting for a 1st prime.

With Tour de Primes coming up, I think I will switch to GFN17mega in february, and be happy with the DC.

Guess karma hit me because of the quick GFN15&16 finds :P