PrimeGrid Challenges 2019

Page 15 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,648
4,589
75
Current challenge: Prime Sierpinski Problem (PSP) LLR, December 12-21 (04:19 UTC)

Happy new year! Here's the (tentative) list of this year's PrimeGrid challenges:

Code:
#  Date             Time UTC  Project  Duration  Challenge
-----------------------------------------------------------------------------------------------------------------
1   7-22 January    05:43:00  SoB-LLR  15 days   Conjunction of Venus & Jupiter Challenge
2   5-10 March      18:00:00  GCW-LLR  5 days    Year of the Pig(ging out on our CPU cycles :P) Challenge
3  24-31 May        00:00:00  TRP-LLR  7 days    Hans Ivar Riesel's 90th Birthday Challenge
4  15-20 July       20:17:00  PPS-LLR  5 days    50th Anniversary of the Moon Landing Challenge
5   3-10 August     00:00:00  ESP-LLR  7 days    Lennart Vogel Honorary Challenge
6  21-26 September  11:00:00  AP27     5 days    Oktoberfest Challenge
7  10-15 October    18:00:00  PPS-DIV  5 days    World Maths Day Challenge
8  24-29 October    00:00:00  321-LLR  5 days    50 years First ARPANET Connection Challenge
9   1-11 November   18:04:00  PSP-LLR  10 days   Transit of Mercury Across the Sun Challenge
10 12-22 December   04:19:00  GFN-21+  10 days   Aussie, Aussie, Aussie! Oi! Oi! Oi! Summer Solstice Challenge

What you need:
  • One or more fast x86 processors, preferably with lots of cores. (Even slow ones might do!)
  • Windows (Vista or later 64-bit, or XP or later 32-bit), Linux, or MacOS 10.4+.
  • BOINC, attached to PrimeGrid (http://www.primegrid.com/).
  • Your PrimeGrid Preferences with only the above project(s) selected in the Projects section.
  • Patience! All of these projects run long, slow WUs, at least on your CPU. As a result, no challenge is less than five days long. :eek:

What may help LLR (all but two of the challenges):
  • An Intel Sandy Bridge or later ("Core series" other than first-generation) processor with AVX may be 20-70% faster than with the default application. Sadly, that does not include Pentium or Celeron processors, or AMD processors.
  • In most challenges - probably all of these since their WUs are so large - it helps to enable multi-core processing with app_config.xml. Leave hyper-threading on if you do this!
  • Faster RAM might help on many challenges, as long as it's stable.
What may help in other challenges:
  • A GPU helps in two challenges.
  • Juggling in some extra WUs may help in challenges where you run more than one WU on the CPU at a time. (Or, switching to use all cores on one WU at the end may work equally well.)
  • Turning on hyper-threading may help.

What won't help (but won't hurt either):
  • A large amount of RAM.
  • Any Android devices.

What won't help (and will hurt, sort of):
  • Unstable processors. (Invalid work will be deducted! :eek: If Prime95 worked recently on your processor, it should be stable.)
  • Work not downloaded anduploaded within the challenge. (It's not counted.) Should you not be able to be in front of one or more computers at that time, there are several options:
    • You can often set BOINC's network connection preferences to wait until a minute or two after challenge time.
    • And for short work units, you can just set the queue level very low (0.01 days). This also makes it more likely that you will be a prime finder rather than a double-checker. But you might want to raise their queue size after the challenge is underway.

Welcome and good luck to all! :)

P.S. If no one has posted stats lately, try tracking your stats with my user script. With that installed, visit the current challenge's Team stats link for TeAm stats.
 
Last edited:

StefanR5R

Elite Member
Dec 10, 2016
6,600
10,402
136
SG are using comparably many CPUs in this challenge. (Strange, if you consider German electricity rates.) This makes their production rate a bit erratic.
 

shka

Junior Member
Sep 26, 2019
1
6
81
I wish SG's charts could be zoomed in. Dark Blue (TeAm) vs Light Green (SG), the race for 3rd is on, with only 10+ hours remaining!
Zoom for x axis is enabled. Hold mouse clicked for your interesting area. But before ....


Same graph but I tried to 'hide' the other colors. It is the last one-third (1/3rd) of the graph (24 hours ago till now) that give me hope. :)
... hide 'irrelevant' teams by click on team name in legend.

PS: Congratulations to 3rd place.
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,648
4,589
75
Preliminary final stats:

Rank___Credits____Username
12_____17433416___phoenicis
16_____14215188___xii5ku
25_____11324443___biodoc
26_____11284013___Howdy
29_____10499671___Lane42
45_____7451249____crashtech
71_____4536246___10esseeTony
77_____3893409____Orange Kid
89_____3335475____VirtualLarry
109____2450058____EricLP
111____2413671____emoga
118____2138747____Modular
139____1787006____SlangNRox
176____1281631____[H]Skillz
177____1269502____Ken_g6
326____238537_____zzuupp
566____12129______geecee

Rank__Credits____Team
1_____169324883___Czech National Team
2_____135788198___Sicituradastra.
3_____95564391___TeAm AnandTech
4_____95208607___SETI.Germany
5_____88646818___[H]ard|OCP
6_____68185195___Aggie The Pew

Wow, I woke up on Saturday and had an AP25; I wake up today and we get third! :smile:

But I couldn't quite stay ahead of @Skillz. Guess I should have used more CPU time.
 

VirtualLarry

No Lifer
Aug 25, 2001
56,575
10,214
126
Wow, we did get 3rd place? Awesome!

Thanks, everybody, and especially @Ken g6 for stats! I was on the edge of my seat the last 3-4 days. (And no, I wasn't drinking!)

Thanks to [H] and SG, for making me/us work so hard during this race. I had nearly every machine going. (If I had another HDMI cable, I would have had an R7 260X in the race too, but I had all of my "big guns" running.)

Edit: And thanks to [H]Skillz, we couldn't have pulled off the 3rd-place win without your contribution too!

A little worried about geecee, his contribution came in pretty early in the race, and then not much more. I hope that he isn't in part of the country that had a natural disaster or something.
 
Last edited:

StefanR5R

Elite Member
Dec 10, 2016
6,600
10,402
136
This script is an offline test with a static work unit, which is good for consistency.

Downsides: a) Those of us who aren't PG insiders need to be told how to set the 'k*2^n+1' target. b) The command line switches shown in the original post (maybe there are more with additional flexibility) only allow to specify to test with either all CPUs or with a given number of CPUs, and either 1...all or a given number of program threads. Of course this latter downside can be eliminated by simply kicking off a whole batch of tests with fixed # of CPUs and # of program threads.

When I test (for throughput, usually), I pick a WU a few days before a challenge (downside: the leading edge moves during a challenge possibly quite a bit), and I already know in which range the optimum number of concurrent tasks for a given processor model should be, thus can reduce the number of tests accordingly. And if these are large WUs, I don't test until 100% completion because there isn't enough time for that, and electricity is far too expensive.

The feature of the python script to automatically re-run tests until a given statistical confidence is reached is great for really precise testing. But for testing in order to prepare for a competition, it is neither practical (due to a tight schedule) nor necessary. (A single test per #tasks x #threads setting is enough, if the system isn't doing anything else in the background --- which it must not do anyway even if you are rerunning the tests for higher confidence.)
 
  • Like
Reactions: Orange Kid

StefanR5R

Elite Member
Dec 10, 2016
6,600
10,402
136
So... PrimeGrid's Fermat Divisor Search project is about identifying primes which divide Fermat numbers. Fermat numbers are an infinitely long sequence, of which F_0 ... F_4 are known to be prime, F_5...F_32 are known to be composite, and the jury is still out on F_33 and (edit: most of) above.

This project once again uses the LLR program which implements a primality test. What we are testing here are not the Fermat numbers themselves, but numbers which are known(?) to be divisors of Fermat numbers *if* they turn out to be prime. Since this property of the numbers we are testing is known(?) beforehand, we don't(?) need to test Fermat numbers directly.

Is my understanding correct?

Edit, or is there still another test performed afterwards by somebody, somewhere, to then determine whether the found prime is a divisor of a Fermat number?

Edit 2, reading a bit further, I now understand that the majority of the Primes to be discovered within PPS-DIV will not be Fermat divisors, and that routine checks whether or not any reported prime is a Fermat divisor are in place at PrimeGrid (and have been for many years now). --- Edit 3, apropos.
 
Last edited:

VirtualLarry

No Lifer
Aug 25, 2001
56,575
10,214
126

The current LLR app uses the best instruction support of the CPU.

As a rough guide to relative peak throughput per core, per clock:

Up to 200% - Intel CPUs with 2 unit AVX-512 implementations
100%+ - AMD Zen 2 CPUs (Ryzen 3600 and higher)
100% - most Intel "-lake" CPUs excluding Celerons, Pentiums and other low end stuff
88% - Intel Haswell CPUs
50%+ - AMD Ryzen CPUs before 3600 (including Threadripper)

I say peak performance, as it can vary with multi-threading, work unit size, and architecture considerations. Specifically for Zen 2 CPUs, best throughput seems to be achieved if you do not make units cross CCX. e.g. for a 6 core CPU, running two 3-thread tasks would tend to be higher throughput than one 6-thread task. For this particular challenge lower thread counts per task are probably better than running high thread counts.

So, pre-Zen2, not so great, slower than Haswell. Good to hear that Zen2 is good to go. Intel HEDT CPUs with AVX512 should scream!

Guess that this upcoming challenge doesn't support GPUs?

And does PrimeGrid yet support the RX 5700 yet in general? I just got one in. Around 49Mh/s mining ETH, using only half the cores. I could maybe mine ETH and do DC work with it at the same time! (Ugh, RX 5700 still not supported in F@H, as of 19.9.2 drivers.)
 
Last edited:

StefanR5R

Elite Member
Dec 10, 2016
6,600
10,402
136
So, pre-Zen2, not so great, slower than Haswell.
They just have different ratios of "general purpose" execution units to FMA execution units. The LLR software depends on the latter (and on the ability to keep them fed with data).
Intel HEDT CPUs with AVX512 should scream!
The 200% figure is, as stated by mackerel, peak, and per clock. Intel doubled FMA units width on most of the SKUs of Skylake-X/SP and derivatives, but did not double cache size and cache bandwidth. Relative to how much FMA execution bandwidth is available on these processors, not very many concurrent LLR tasks fit onto them, and hence LLR will operate somewhere off of its throughput sweetspot.

Further, as soon as AVX-512 ops are used, these processors clock down notably. Though while many users keep pointing to this as something negative, I say it's a good thing as it makes AVX execution more energy efficient.
Guess that this upcoming challenge doesn't support GPUs?
Right, it doesn't ---
  • A GPU helps in two challenges.
Particularly, GPUs got/ get into business only in the September AP27 challenge and the December GFN challenge (when we are busy folding proteins).
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,648
4,589
75
Day 1 stats:

Rank___Credits____Username
80_____127854_____Howdy
97_____100274_____Ken_g6
101____92685______Orange Kid
115____76436______SlangNRox
187____34936______zzuupp

Rank__Credits____Team
13____783115_____Team 2ch
14____552880_____The Knights Who Say Ni!
15____502096_____BOINC@Poland
16____432187_____TeAm AnandTech
17____423734_____BOINC@AUSTRALIA
18____371624_____Ukraine
19____317860_____San Francisco

With Formula BOINC at the same time, I don't think we're going to make third this time. :(
 

StefanR5R

Elite Member
Dec 10, 2016
6,600
10,402
136
Third in a CPU race at PG?
We did this once, in the SR5 race in July 2018. That was because ATP's top gun was practically absent. (CNT were 1st, SG were 2nd in this one.)
 

TennesseeTony

Elite Member
Aug 2, 2003
4,334
3,803
136
www.google.com
I was really looking forward to this one, due to the original estimation that the project would only live for 3 months or so...but now that it's life expectancy is much longer....uhm, I'll probably stick to the FB Sprint for now.
 

Ken g6

Programming Moderator, Elite Member
Moderator
Dec 11, 1999
16,648
4,589
75
Day 2 stats:

Rank___Credits____Username
70_____308716_____Howdy
99_____214235_____Ken_g6
102____192670_____Orange Kid
118____152967_____SlangNRox
183____81366______zzuupp
240____45851______VirtualLarry
415____1689___10esseeTony

Rank__Credits____Team
13____1529234____Dutch Power Cows
14____1182767____The Knights Who Say Ni!
15____1111966____BOINC@Poland
16____997007_____TeAm AnandTech
17____942694_____BOINC@AUSTRALIA
18____839406_____Ukraine
19____663135_____BOINCstats

Welcome, @TennesseeTony. Let's make the Aussies "roo" the day they chased us! ;)