2019 Tour de Primes

[Gilthanis]

Last year's thread - https://hardforum.com/threads/tour-de-primes-2018-primegrid.1952746/

http://www.primegrid.com/forum_thread.php?id=8409#125223

Tour de Primes 2019

Welcome to the 11th annual Tour de Primes. 2 is the first prime number...and the only even one. This makes it unique among prime numbers. Therefore, February is declared Prime month...being the 2nd month of the year. And there's no better way to pay homage to a prime number than to go out and find one. More precisely, a Top 5000 prime.

For the month of February, an informal competition is offered. There are no challenge points to be gained... just a simple rare jersey at the end of the month to add to your badge list. No pressure or stress other than what you put on yourself.

For 2019, we're bring back the badges introduced in 2018:

• 
  Red Jersey -- discoverer of largest prime
  
• 
  Yellow Jersey -- prime count leader (tiebreaker will be prime score)
  
• 
  Green Jersey -- points (prime score) leader
  
• 
  Polk-a-dot Jersey -- on the 19th of February we'll have a "Mountain Stage" and award the Polk-a-dot Jersey to the one who finds the most primes on that day (tiebreaker will be prime score for that day).
  
• 
  Prime badge -- awarded to everyone who finds an eligible prime during the month of February. This is a counter badge, so if you find more than one prime it will show how many you've found, up to 99.
  
• 
  Mega prime badge -- awarded to everyone who finds a mega prime during February. This is a counter badge.
  
• 
  Mountain Stage prime badge -- awarded to everyone who finds an eligible prime during the Mountain Stage. This is a counter badge.
  
• 
  Mountain Stage mega prime badge -- awarded to everyone who finds a mega prime during the Mountain Stage. This is a counter badge.

Results will be available at http://www.primegrid.com/challenge/tdp_2019.php.

As with the last few years, for all primes (BOINC and PRPNet) we're using the new reporting system whereby the prime's date of discovery determines whether it's eligible for the Tour de Primes. Prior to 2014, the date of verification for BOINC primes was used while the discovery date was used for PRPNet primes. The current system is more intuitive and fairer.

Note that SGS-LLR and GFN-15 are too small to be reported to the Top 5000 primes list and are therefore not eligible for the 2019 Tour de Primes.

Currently, the fastest opportunities to find Top 5000 primes is with the PPSE (LLR), and GFN-16 (65536) projects. Of course, should someone find a prime in the mega-prime searches, this would certainly give them a good shot at the green jersey. Not a guarantee, however, as in 2018 there were several mega primes found in the Tour de Primes. Overall, in 2018 we averaged close to two mega primes per week for the entire year, so you might need more than "merely" a mega prime to take home green. In 2018 there were 16 mega primes found during Tour de Primes.

All ports in PRPNet are available for the competition.

To participate in BOINC PPSE (LLR), GFN-16, or any other eligible LLR or Genefer project, all you have to do is select it in your PrimeGrid preferences. AP27 sequences are not reportable at T5K, so are not eligible for Tour de Primes.

Good Luck, have fun, and enjoy!

Previous Winners

- Robish

- [AF>Amis des Lapins] dthonon

- [AF>Amis des Lapins] dthonon

- tng*

- wdethomas

- tng*

- wdethomas

- tng*

- boss

- Scott Brown

- boss

- Orange_1050

- Scott Brown

- Randall J. Scalise

- vmc

- [DPC]x-RaY99_the_one_man_team

- [DPC]x-RaY99_the_one_man_team

- Lonnie Christensen

- [DPC]x-RaY99_the_one_man_team

- [DPC]x-RaY99_the_one_man_team

- PBT_marian_boss

- Scott Brown

- Usucapio Libertatis

- Scott Brown

- shanky123

- [BOINCstats] LostBoy

- Snf*

- lennart

- lennart

- [SG]marodeur6

- lennart

- s-yama

- lennart

Full rankings can be seen here: 2009 | 2010 | 2011 | 2012 | 2013 | 2014 | 2015 | 2016 | 2017 | 2018

Totals by Year

• 
  2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 Total Primes found 212 309 766 646 238 254 254 346 246 256 Total Score 457.30 2663.27 21803.72 20727.04 15614.83 20982.77 30268.46 43656.50 40201.94 55393.50

Tips and Strategies:

Tip #1: He (or she) who finds the prime FIRST is the discover of the prime. It's a competition between you and your wingman. While having a fast computer helps, your computer is only useful when it's running a task. If you have a cache of tasks sitting on your computer waiting to run, chances are your wingman will return the task before you've even started it. Setting both BOINC cache settings to "0 days" is strongly recommended. People with slow computers find primes all the time because their wingman downloaded the task yesterday but won't start running it until tomorrow. Set your cache to 0 days!

If you're going to ignore tip #1, don't even bother reading the rest. Seriously.

Tips for LLR:

• Your mileage may vary. What works for me may not work for you. Before TdP starts, take some time and experiment and see what works best on your computer.
  
• If you have an Intel CPU with hyperthreading, either turn off the hyperthreading in the BIOS, or set BOINC to use 50% of the processors. (But see below for exceptions.)
  • If you're using a GPU for other tasks, it may be beneficial to leave hyperthreading on in the BIOS and instead tell BOINC to use 50% of the CPU's. This will allow one of the hyperthreads to service the GPU.
  
  
  
• Use LLR's multithreaded mode. It requires a little bit of setup, but it's worth the effort. Follow these steps:
  • Create an app_config.xml file in the directory C:\ProgramData\BOINC\projects\www.primegrid.com\ (or wherever your BOINC data directory is located). For a quad core CPU, the file should contain the following contents. Change the two occurrences of "4" to the number of actual cores your computer has. The example below is for PPSE. Change the app name (2 places) to whatever LLR app you're running. The app names are listed on your task selection page.
    
    <app_config> <app> <name>llrPPSE</name> <fraction_done_exact/> <report_results_immediately/> </app> <app_version> <app_name>llrPPSE</app_name> <cmdline>-t 4</cmdline> <avg_ncpus>4</avg_ncpus> </app_version> </app_config>
    
    
  • After creating the file, click on "Options/Read config files" or restart BOINC or reboot.
    
  • The first time BOINC downloads (in this example) a PPSE-LLR task, it may act a little strange and download 4 tasks instead of 1. The run times on this first set of tasks may look a bit strange too. This is normal. This will also occur anytime BOINC downloads more than one task at a time. This can be avoided by setting "Use at most [ 1 ] % of the CPUs" before you download PPSE tasks. After one task was downloaded, increase the percentage.
    
  • Some people have observed that when using multithreaded LLR, hyperthreading is actually beneficial. I don't use hyperthreading myself, but I encourage you to experiment and see what works best for you.

Tips for GFN:

• Only run GFN on a GPU. Use your CPU for LLR tasks where it will be much more efficient.
  
• Unless you have a really slow GPU and a really fast CPU, leave a CPU core free to service the GPU. You'll want the more powerful GPU running at full speed, even if it slows down the CPU somewhat. A hyperthread should be sufficient if your CPU supports hyperthreading. For example, on a 4 core CPU (without hyperthreading), you could set BOINC to "use 75% of the CPUs" to reserve one core for the GPU.

 

[RFGuy_KCCO]

I’ve decided to participate in this challenge this year. Hoping to find a mega-prime!

 

[EXT64]

So is the multi-threading on LLRs only meant for smaller CPUs (quad/hex), or does it scale up well to larger systems?

 

[RFGuy_KCCO]

So is the multi-threading on LLRs only meant for smaller CPUs (quad/hex), or does it scale up well to larger systems?

I am new to multi-threaded LLR tasks, so I am unsure how well they actually scale with more threads. I am currently testing it with Sophie Germain tasks and find my TR's seem to scale better than my 4790K boxes, but I am still testing. Haven't tested the 9900K at all yet.

I have a feeling each subproject behaves differently, as well, so what works well for one may not necessarily work for another.

Because I have badge and hour (WU Prop) goals, I plan to run all three PPS subprojects and the SR5 subproject for the Tour de Primes. I will be using 2 threads per task as a starting point on the PPS tasks and 4 per task for the SR5 tasks.

Note - I always enable HT/SMT, so I speak in threads, not cores.

 

[Tornlogic]

It scales, but there becomes a point of diminishing returns in efficiency. On a side note, some say that enabling Hyperthreading (when multi-threading is being used) helps when doing tasks in parallel with a fast GPU, however, I have always seen performance drop when I enable Hyperthreading, so I keep it off. In other words I have NEVER seen Hyperthreading help when it comes to LLR. I think it has something to do with the amount of cache your CPU uses for each thread.... I'm not sure. So my quad core systems I have all 4 cores multi-threading. With my 8 core system I have 7 cores multi-threading, although there isn't much difference in efficiency between 6 and 7 cores. 5 and 6, yes there is a substantial difference, but not 6 and 7. I turn that last core on because, why not. I'm not doing anything else with it. Besides, my basement is freaking COLD!! and it helps keep it warm.

 

[Tornlogic]

RF, so maybe it depends what CPU you have wether one should run Hyperthreading or not? I found running SOB, I took a major hit in performance running with HT turned on. Now I'm running Woodall LLR for a badge, and I tried hyperthreading again, and still can't reach the performance I get with it turned off. Same with another subproject I tried before the SoB challenge. This is all with a 9900K.

Another test I tried with SOB was, (With HT turned off) running 2 WUs giving each WU 3 cores and another test giving each WU 4 cores. I found both of those test to be a complete bomb. It was better to push all 7 cores to one WU.

 

[RFGuy_KCCO]

The only problem with that for me is that I run other projects on my TR boxes and my 9900K box, so I leave HT turned on. My 4790K boxes (x3), though, are dedicated to PG (and sometimes SRBase), but I still have HT turned on because I also crunch GPU tasks on those boxes. On those, I set the BOINC CPU usage to 50% and run only 4 threads of PG work. It is slower than if I were to turn off HT altogether, but faster than running 8 threads of PG work. I am willing to make this compromise because the GPU work is important to me. I have 4 2080's between those three boxes, after all.

I believe all of the LLR tasks will always run faster with HT turned off, if one is willing to do that. I just run multiple projects and so am unwilling to turn it off.

 

[RFGuy_KCCO]

I am new to multi-threaded LLR tasks, so I am unsure how well they actually scale with more threads. I am currently testing it with Sophie Germain tasks and find my TR's seem to scale better than my 4790K boxes, but I am still testing. Haven't tested the 9900K at all yet.

I have a feeling each subproject behaves differently, as well, so what works well for one may not necessarily work for another.

Because I have badge and hour (WU Prop) goals, I plan to run all three PPS subprojects and the SR5 subproject for the Tour de Primes. I will be using 2 threads per task as a starting point on the PPS tasks and 4 per task for the SR5 tasks.

Note - I always enable HT/SMT, so I speak in threads, not cores.

Update to the above... It looks like with HT turned on the Sophie Germain tasks do not scale at 2 threads per task. With one thread per task, my TR's were completing the tasks in ~31:00, so to be worth running two threads per task, it would have to halve this time to ~15:30 per task when using two threads. However, I am seeing completion times closer to ~20:00 per task when using two threads, so I am going back to one thread per task for now. I would bet my results would be different if HT was turned off.

 

[Bill1024]

Not sure if you guys know this or not, but the Sophie German primes will not count for the Tour de Prime challenge.
The're now too small for the list of 5000 largest primes.

 

[RFGuy_KCCO]

Not sure if you guys know this or not, but the Sophie German primes will not count for the Tour de Prime challenge.
The're now too small for the list of 5000 largest primes.

Right, I know that. I was trying to work on a badge in that project before the start of the TdP on 2/1. Thank you for looking out, though!

 

[Bill1024]

Ahhh ok, I would of hated to see you guys chasing SG and them not counting for the TDF.
Carry on!

 

[Tornlogic]

Yea, thanks for looking out for us. I'm working on badges too before the next challenge starts. So many challenges, it's hard to get specific ones in between challenges.

 

[Gilthanis]

The event begins tomorrow

 

[Bill1024]

The event begins tomorrow

No, it starts in one hour fifty four minutes from now. Or 7PM EST or 00 universal time

It is after 10PM the 31st in Univeral time.

 

[Bill1024]

This challenge has started. Good luck everyone.