Info PrimeGrid Challenges 2024, sieve-free edition

[Markfw]

Mark, you're still in first. You're looking at the last challenge where you finished 11th.

My other computers refresh is not working correctly, thanks. But at this rate, crashtech will pass me soon.

 

[Skillz]

It's not your computer. You're looking at the wrong challenge.

Challenge statistics

This is the correct link to the current challenge. You're still in first, by a good margin.




Based on this graph you'll stay ahead of crash unless you shut down systems or he adds more.

 

[Markfw]

wah is now on the upswing. How does he look ? debating on whether to add a dual 7763. Not efficient compared to Genoa, but better than the dual 7v12 dual that I have. As a team we are killing them, I see that.

 

[Markfw]

And a lot of our team is not participating, including you and @cellarnoise

 

[waffleironhead]

Im quietly watching this race from the beach in california. First real vacation in 15 years. Have at it mates. Cheering for you all.

 

[Skillz]

wah is now on the upswing. How does he look ? debating on whether to add a dual 7763. Not efficient compared to Genoa, but better than the dual 7v12 dual that I have. As a team we are killing them, I see that.

Doubt he has enough to pass you. Looks like he was bunkering some of his systems and dumped them sometime early this morning.

 

[mmonnin03]

How does one exclusively get proof tasks? For nearing end of the competition.

Computer 1169191

 

[Markfw]

Doubt he has enough to pass you. Looks like he was bunkering some of his systems and dumped them sometime early this morning.

does lasso work on 128 core dual socket systems ?

 

[Markfw]

well after updating lasso, and rebooting the projected elapsed times are around 13 hours, the same as crashtech's dual 7763 system, so I will let it run for now.

BTW, elsed approx times for the following (all 8 cores per task, smt disabled in lasso)

9950z 4.4 hours (and one of the 79503x3d's)
7950x 5 hours
9554 Genoa 64 core 6-6.5 hours.
dual 7763 Milan 64 core 13 hours.

The jump from Milan to Genoa is HUGE(avx-512 support). I can't wait until I can fire up my 64 core Turin !

 

[Skillz]

Your 9654 system appears to not be pinning the tasks.

 

[Ken g6]

Day 1.5 stats:

Rank___Credits____Username
1______9764408____markfw
2______8253557____w a h
3______7398710____crashtech
8______3737757____cellarnoise2
11_____3018173____Icecold
19_____1857571___10esseeTony
31_____1031514____Orange Kid
33_____976007_____biodoc
38_____750892_____mmonnin
104____245125_____Fardringle
118____189272_____Ken_g6
143____94058______DROFFUNGUS

Rank__Credits____Team
1_____37317051___TeAm AnandTech
2_____12683762___Czech National Team
3_____11415652___Antarctic Crunchers
4_____9172441____SETI.Germany

Mark's still ahead of SETI Germany.

 

[StefanR5R]

9554P @400W PPT which xii5ku rented out to parsnip:
(runs at something like 3.4+ GHz on average, IIRC, can't check right now)
8 tasks x 8 threads each, each task bound to the lower 8 logical CPUs of an exclusive CCX,
31 results from the last 24 hours: 25,390 / 22,549 / 31,368 seconds elapsed time (avg/min/max) = 7:03 / 6:16 / 8:43 h:mm

I did not take the time yet to check what would happen if I used SMT. I believe I would have to tweak my PRST test script first for more precision to know for sure. Also, RAM is currently dialed down to 3600 DDR IIRC, down from 4800 DDR originally, without having tested the effect on GCW-LLR/PRST. I left it this way because PRST should be mostly accessing the CPU cache. (Too little spare time to go better informed into this challenge...)

Zen 3, such as EPYC 7003/ Milan, should still fare well in this challenge because it has enough cache if configured near to optimum. I have no idea if Windows + Lasso work decently on a 2x64c dual socket system in particular. Anyway, keep in mind that a 64-cores EPYC 7003 has quite some less Watts/core than a 64-cores EPYC 9004 at its disposal, and for this reason alone should be noticeably slower at same core count. (Besides Genoa's improved instruction set support and more tweaks to the core and newer manufacturing process node.)

It's only day 1.5 yet, but nonetheless:

Rank__Credits____Team
1_____37,317,051___TeAm AnandTech
2_____12,683,762___Czech National Team
3_____11,415,652___Antarctic Crunchers
4_____9,172,441____SETI.Germany

12.7 + 11.4 + 9.2 = 33.3 M

Mark's still ahead of SETI Germany.

:-O

 

[Skillz]

We were talking about how TAAT was beating 2 and 3 combined earlier, then wah and cellar dropped some nukes and put the team ahead of 2, 3 and 4 combined. Also, icecold got a late start. He should be coming up in the ranks, slowly, but climbing a bit just the same.

 

[StefanR5R]

How does one exclusively get proof tasks? For nearing end of the competition.

Computer 1169191

Hmm, 1 main task and 1175 verification tasks...! I don't follow closely what the PrimeGrid admins are doing, but I doubt that they changed their minds about condoning or even supporting cherry-picking between main and verification tasks. Given that these tasks were assigned to the host in many requests over a large period, it has got the looks as if the host owner figured out a loophole. But still, I have doubts.

 

[Skillz]

He was sent that main task pretty early. I wonder what would happen if you suspended the main task in BOINC and kept requesting work. Would it maybe only send proof tasks? Definitely very odd.

 

[Markfw]

Update. my dual 7763 is doing 17.5 hours. Not sure if its configured the best, and has not dropped a unit yet, but its on windows with lasso, so that the best I can do. It will start dropping in about 2 hours, so my chart will change, as these will be the first from this box. Slower than crashtech's 7763's but that's probably because they are ES chips. The 9654 running much slower than the 9554, is running almost 10 hours compared to the 9554 at 6 hours. Not quite as fast as seti germany, but only 600k behind, and once the dual 7763's start dropping, I could catch up to them again.

 

[Markfw]

But note, I go into the hospital for my surgery/procedure in about 4 hours. Be out for about 8 more, so I can't fix anything, but most of my points are from servers, and they are rock solid.

 

[Markfw]

Back past SetiGermany again ! The dual 7763 has started to dump !

 

[mmonnin03]

Hmm, 1 main task and 1175 verification tasks...! I don't follow closely what the PrimeGrid admins are doing, but I doubt that they changed their minds about condoning or even supporting cherry-picking between main and verification tasks. Given that these tasks were assigned to the host in many requests over a large period, it has got the looks as if the host owner figured out a loophole. But still, I have doubts.

It was nearly 600 proof tasks when I posted that link. They stood out like a sore thumb on the Individual stats page. I do not see a single abort and no main tasks in the queue.

I hope all goes well Markfw

 

[Skillz]

There is one main task. Its on the last page of the in progress tab.

Workunit 1140238453

This task.

 

[Markfw]

I hope all goes well Markfw

Preliminary results on the coloscopy. The bleeding was diverticulitis. Probably NO cancer !!! I find out for sure in 7 days. I have had that for years, and did not realize that it could start bleeding later. Lets hope for the best in 7 days, but the surgeon wasn't worried that it was cancer ! He didn't think it was.

 

[Ken g6]

Day 2.5 stats:

Rank___Credits____Username
1______17252298___markfw
2______14402676___w a h
3______13061895___crashtech
9______6808678____cellarnoise2
10_____6442865____Icecold
31_____1816929____Orange Kid
32_____1809234____biodoc
35_____1577917____mmonnin
59_____1030816____Fardringle
123____343641_____Ken_g6
166____144559_____DROFFUNGUS
230____44089______waffleironhead

Rank__Credits____Team
1_____67844305___TeAm AnandTech
2_____24762061___Czech National Team
3_____21176870___Antarctic Crunchers
4_____17699124___SETI.Germany

 

[Markfw]

Your 9654 system appears to not be pinning the tasks.

just saw this. The 9654 is sectioned into 12 sections (you helped me do this)

note for update: going to ER for pain. probably normal after getting polups removed.

 

[Ken g6]

Day 3.5 stats, just a little past halfway:

Rank___Credits____Username
1______26136618___markfw
2______19866227___crashtech
3______19692677___w a h
8______10904511___Icecold
10_____9962286____cellarnoise2
32_____2762093____biodoc
34_____2634935____Orange Kid
35_____2519274____mmonnin
47_____2018285____Fardringle
126____484263_____Ken_g6
184____195692_____DROFFUNGUS
202____142284_____waffleironhead

Rank__Credits____Team
1_____100427851___TeAm AnandTech
2_____41030389___Czech National Team
3_____32841235___Antarctic Crunchers
4_____27854664___SETI.Germany

 

[Markfw]

update (I give it here, since I first mentioned my procedure yesterday.) First the good news (maybe) not official yet, but the surgeon says he is pretty sure I don't have cancer, but my symptoms are of diverticulitis, which I have had for more than 10 years.

Now the bad news. I may have pneumonia and I have been in ER for the last day. I am back, but being treated for it.