Info PrimeGrid Challenges 2024, sieve-free edition

TennesseeTony · Sep 26, 2024

@Markfw it's hard to know how to respond to the mixed news, like your mother-in-law just drove off a cliff in your new Ferrari, lol, OH NO. You've got this. PLUS, you're SLAYING the competition this round!!! Congrats!!! Minor flicker in the grid knocked some of my systems offline while away at work, no biggie. Thursday promises TROPICAL STORM weather however. Yeah! Boo! ??? Tropical storm deep into the USA? Weird.

Skillz · Sep 26, 2024

Doing a great job team. Looks like we lost a comrade though.

Markfw · Sep 26, 2024

Skillz said:
Doing a great job team. Looks like we lost a comrade though.

who did we lose ?

StefanR5R · Sep 26, 2024

Looking for somebody? I found one or another. :-)

Markfw · Sep 26, 2024

well, one reason I am lagging. An entire 7950x was on "won't get more tasks". Not sure how that happened. It back on-line.

Markfw · Sep 26, 2024

StefanR5R said:
Looking for somebody? I found one or another. :-)

I can see why somebody would want to support Ukraine, as we have no competition that I can see.

and personal people, we have 1,2,5,8,12 for the high pointers.

Ken g6 · Sep 26, 2024

PSA: All you Linux users, be sure to secure your CUPS!

[Reg] "doomsday critical Linux bug" in CUPS

https://www.theregister.com/2024/09/26/unauthenticated_rce_bug_linux/?td=readmore In short, if you're running the Unix printing system CUPS, including cups-browsed, then you may be vulnerable to attacks that could lead to your computer being commandeered over the network or internet. CUPS is...

forums.anandtech.com

Ken g6 · Sep 26, 2024

Day 4.5 stats:

Rank___Credits____Username
1______34003471___markfw
2______25539910___crashtech
5______21017328___w a h
8______14662177___Icecold
12_____11675071___cellarnoise2
31_____3569893____biodoc
33_____3505605____Orange Kid
35_____3367103____mmonnin
40_____2832489____Fardringle
137____591270_____Ken_g6
165____389933_____DROFFUNGUS
194____241113_____waffleironhead

Rank__Credits____Team
1_____124504071___TeAm AnandTech
2_____57779325___Czech National Team
3_____41927903___Antarctic Crunchers
4_____37692121___SETI.Germany

StefanR5R · Sep 27, 2024

Briefly after 01:15 UTC = less than an hour before @Ken g6 captured the latest stats, TeAm AnandTech flew by the 123,456,789 milestone. In the middle of the night for me, so I didn't watch the live transmission, but I've got it on tape. ;-)

StefanR5R · Sep 27, 2024

PRST occasionally dies with segfaults. Only on EPYC Rome though, haven't seen it on EPYC Genoa yet. (I don't have Broadwell-EP running this time.)

crashtech · Sep 27, 2024

I've had a few systems reboot themselves during the course of this challenge, but I didn't research the cause.

Markfw · Sep 27, 2024

My Milan and NO other Genoa/Zen 4 system has rebooted in the 5.5 days so far. Thats all I have running in this.

Ken g6 · Sep 27, 2024

Day 5.5 stats:

Rank___Credits____Username
1______42595337___markfw
3______31231354___crashtech
6______23495814___w a h
8______18708144___Icecold
14_____13654664___cellarnoise2
31_____4431880____biodoc
33_____4246428____mmonnin
35_____4216371____Orange Kid
39_____3759030____Fardringle
140____735448_____Ken_g6
145____639775_____DROFFUNGUS
189____338194_____waffleironhead

Rank__Credits____Team
1_____151161144___TeAm AnandTech
2_____74638737___Czech National Team
3_____52252989___Antarctic Crunchers
4_____47830901___SETI.Germany

I see they've been busy tracking down bugs in PRST. Hopefully the segfaults will get resolved soon.

PRST v13.1 is released - mersenneforum.org

Current release is v13.1 and precompiled binaries are available. Links: https://github.com/patnashev/prst/releases https://github.com/patnashev/prst https://t5k.org/bios/page.php?id=5652

www.mersenneforum.org

Markfw · Sep 27, 2024

Ken g6 said:
Day 5.5 stats:

Rank___Credits____Username
1______42595337___markfw
3______31231354___crashtech
6______23495814___w a h
8______18708144___Icecold
14_____13654664___cellarnoise2
31_____4431880____biodoc
33_____4246428____mmonnin
35_____4216371____Orange Kid
39_____3759030____Fardringle
140____735448_____Ken_g6
145____639775_____DROFFUNGUS
189____338194_____waffleironhead

Rank__Credits____Team
1_____151161144___TeAm AnandTech
2_____74638737___Czech National Team
3_____52252989___Antarctic Crunchers
4_____47830901___SETI.Germany

I see they've been busy tracking down bugs in PRST. Hopefully the segfaults will get resolved soon.

PRST v13.1 is released - mersenneforum.org

Current release is v13.1 and precompiled binaries are available. Links: https://github.com/patnashev/prst/releases https://github.com/patnashev/prst https://t5k.org/bios/page.php?id=5652

www.mersenneforum.org

why do some of us have no issues ???

StefanR5R · Sep 28, 2024

Markfw said:
why do some of us have no issues ???

In my case, only the code path which is taken on Zen 2 experiences these occasional segmentation violations. In @crashtech's case, a kernel bug or BIOS/AGESA bug or hardware fault is the cause (not, or in addition to, whichever bug in PRST).

Edit:
What I am seeing with PRST 13.0 could also be dependent on the specific candidate tested. But no error return from my single Genoa so far makes it likely that it is also specific to one or another FFT implementation.

I clicked through the workunits of all (at this time) 39 error results of my client instances which run on EPYC Rome:

Several of these workunits had only 'unsent' or 'in progress' replica tasks, plus sometimes 'abort'ed tasks.

Some workunits had same error returns from other EPYC 7002 or Ryzen 3000: 3457728*101^3457728+1, 3443928*101^3443928+1, 3380020*109^3380020+1

Workunit 3440206*101^3440206+1]3440206*101^3440206+1 had a successful return from a Threadripper 3000 = Zen 2. But this tasks was suspended and resumed once, that is, the transform did not have to run in one go. Maybe this is related, maybe not.

Workunit 3379216*109^3379216+1 completed successfully on a Ryzen 5000 = Zen 3. Ditto, workunits 3362962*109^3362962+1, 3357022*109^3357022+1, and 3346396*109^3346396+1. The latter happened to restart itself due to rounding errors, which is normal with some candidates, but the former three completed in one go. 3412000*101^3412000+1 completed on a Ryzen 6000 = Zen 3.

Workunit 3438772*101^3438772+1 had an additional seemingly unrelated error return from a Ryzen 7000: Process creation failed, execv: Text file busy.

The rest of the workunits had successful results of replica tasks on hosts with different processor architecture. For example, workunit 3386692*109^3386692+1 completed successfully on Ryzen 7000. (@crashtech's, no less.) 3437410*101^3437410+1 completed successfully on a Ryzen 9000. Ryzen 7000, 9000, and EPYC 9004 use the AVX-512 code path of the FFT, EPYC 7002 and Ryzen 3000 — but also Ryzen 5000/6000 — use the FMA3 code path. Or workunit 3442266*101^3442266+1 completed successfully on a Xeon 8160M which uses the AVX-512 implementation of the FFT too. Workunit 3434986*101^3434986+1 completed on an Apple M3 Pro but with PRST version 9.0.

All of the successful hosts in my error'ed workunits have not a single error in their results tables.

Not only have I seen this error only on my Zen 2 hosts, all of the hosts of other users which I have seen to throw this error were Zen 2 too, so far.

Homework for myself: 1. Re-run one of the faulted workunits on Zen 2 and see if it fails again. 2. If yes, Re-run on Haswell and Broadwell-EP which also uses FMA3. (Though I have not seen FMA3 faulting on other folks' Zen 3 yet.)

Icecold · Sep 28, 2024

For what it's worth, I'm currently running 4 Broadwell systems (3 of which are dual socket/CPU) and 1 single socket Haswell system and I just checked system uptime and none of them have rebooted and I don't show any errored tasks on the PG website. If there's anything else you'd like me to check let me know but I'm not having any issues at all that I'm aware of on my Haswell/Broadwell stuff other than it being quite a bit slower than newer CPU's 🤣

Orange Kid · Sep 28, 2024

I have had 3 errored tasks on my 3950x, Linux Mint.
All have since been completed.
One by an Intel Xeon, Linux Ubuntu
One by a 5950x, Windows 10
And the interesting one was a 3900x, Windows 10.
So is it simply the operating system, or hardware that is the problem?

Ken g6 · Sep 28, 2024

Day 6.5 stats:

Rank___Credits____Username
1______50537297___markfw
3______35990073___crashtech
6______25970019___w a h
9______22638739___Icecold
11_____18528227___cellarnoise2
31_____5249067____biodoc
33_____5192111____mmonnin
34_____5097865____Orange Kid
43_____4267354____Fardringle
139____896723_____Ken_g6
146____841302_____DROFFUNGUS
206____338194_____waffleironhead

Rank__Credits____Team
1_____178655677___TeAm AnandTech
2_____92444912___Czech National Team
3_____61632904___Antarctic Crunchers
4_____57722456___SETI.Germany

I might yet be caught by a fungus that's chasing me!

Markfw · Sep 28, 2024

I don't think that anybody can catch us with 9.5 hours to go, and second place is 86 million below us . We are almost double second places points. I won't let up until the morning. No "no new work" on 9 of my 10 boxes until the morning, as my slowest box of the 9 does units in 6 hours.

crashtech · Sep 29, 2024

I might let a few boxes gallop past the finish line, but most of them are already set for no new tasks.

Markfw · Sep 29, 2024

Just got home from ER again.. Fell, and might have had a concussion it was that bad. forehead vs concrete slab = very bad.

drfungus · Sep 29, 2024

Ken g6 said:
Day 6.5 stats:

Rank___Credits____Username
1______50537297___markfw
3______35990073___crashtech
6______25970019___w a h
9______22638739___Icecold
11_____18528227___cellarnoise2
31_____5249067____biodoc
33_____5192111____mmonnin
34_____5097865____Orange Kid
43_____4267354____Fardringle
139____896723_____Ken_g6
146____841302_____DROFFUNGUS
206____338194_____waffleironhead

Rank__Credits____Team
1_____178655677___TeAm AnandTech
2_____92444912___Czech National Team
3_____61632904___Antarctic Crunchers
4_____57722456___SETI.Germany

I might yet be caught by a fungus that's chasing me!

not quite i don't think but next time i'm going for yer neck!so close....

Ken g6 · Sep 29, 2024

More-or-less final stats:

Rank___Credits____Username
1______54022456___markfw
3______37494136___crashtech
6______28361574___w a h
9______24333914___Icecold
11_____19858641___cellarnoise2
22_____10307918___10esseeTony
32_____5703431____biodoc
34_____5463916____mmonnin
35_____5307867____Orange Kid
48_____4267354____Fardringle
140____948997_____Ken_g6
150____841302_____DROFFUNGUS
202____385217_____waffleironhead

Rank__Credits____Team
1_____190097513__TeAm AnandTech
2_____99710685___Czech National Team
3_____65492785___Antarctic Crunchers
4_____62921819___SETI.Germany

That's a decisive win, TeAm!

emoga · Sep 29, 2024

Congratulations on taking first place, @Markfw

I think this is your first contest win, right? You came really close a couple of other times.

Your home setup is really impressive when you fire everything up!

Markfw · Sep 29, 2024

emoga said:
Congratulations on taking first place, @Markfw View attachment 108426

I think this is your first contest win, right? You came really close a couple of other times.

Your home setup is really impressive when you fire everything up!

Thanks. I even had 1 9554 off-line waiting for Turin ! And when budget allows, I will have that and A Turin. And I had one 7950x crunching WCG.

Info PrimeGrid Challenges 2024, sieve-free edition

Elite Member

Senior member

Moderator Emeritus, Elite Member

Elite Member

Moderator Emeritus, Elite Member

Moderator Emeritus, Elite Member

Programming Moderator, Elite Member

Programming Moderator, Elite Member

Elite Member

Elite Member

Lifer

Moderator Emeritus, Elite Member

Programming Moderator, Elite Member

Moderator Emeritus, Elite Member

Elite Member

Golden Member

Elite Member

Programming Moderator, Elite Member

Moderator Emeritus, Elite Member

Lifer

Moderator Emeritus, Elite Member

Junior Member

Programming Moderator, Elite Member

Member

Moderator Emeritus, Elite Member