K33 and K34 all [P2] Table x rewrites are Single Core

[Grosik1577]

AMD 5950x 128GB 3600 2x1.6TB PM1735 Clear Linux kernel 5.12
I have problem with k33 and k34 plots. As i observed the main difference starts from
[P2] Table 6 rewrite
If it goes SLOW than, as per Htop, all [P2] Table x rewrites are Single Core 100% and other cores 10-20%
[P2] max_table_size = 8591960036
[P2] Table 7 scan took 37.1078 sec
[P2] Table 7 rewrite took 62.7559 sec, dropped 0 entries (0 %)
[P2] Table 6 scan took 43.7843 sec
[P2] Table 6 rewrite took 346.423 sec, dropped 1162409851 entries (13.5306 %)
[P2] Table 5 scan took 42.7258 sec
[P2] Table 5 rewrite took 317.952 sec, dropped 1523678697 entries (17.737 %)
[P2] Table 4 scan took 43.0958 sec
[P2] Table 4 rewrite took 315.809 sec, dropped 1657525624 entries (19.2956 %)
[P2] Table 3 scan took 41.9618 sec
[P2] Table 3 rewrite took 320.874 sec, dropped 1710002810 entries (19.9067 %)
[P2] Table 2 scan took 41.8325 sec
[P2] Table 2 rewrite took 320.281 sec, dropped 1731051201 entries (20.1521 %)
Phase 2 took 1951.04 sec
If it goes FAST than all cores are 70-90% used
[P2] max_table_size = 8591485063
[P2] Table 7 scan took 27.2468 sec
[P2] Table 7 rewrite took 56.8793 sec, dropped 0 entries (0 %)
[P2] Table 6 scan took 45.1812 sec
[P2] Table 6 rewrite took 90.4128 sec, dropped 1162401340 entries (13.5309 %)
[P2] Table 5 scan took 43.7353 sec
[P2] Table 5 rewrite took 86.8796 sec, dropped 1523755776 entries (17.7379 %)
[P2] Table 4 scan took 42.9656 sec
[P2] Table 4 rewrite took 85.4762 sec, dropped 1657598359 entries (19.2965 %)
[P2] Table 3 scan took 42.7092 sec
[P2] Table 3 rewrite took 84.9902 sec, dropped 1710037887 entries (19.9072 %)
[P2] Table 2 scan took 42.5686 sec
[P2] Table 2 rewrite took 84.6988 sec, dropped 1731088017 entries (20.1523 %)
Phase 2 took 750.125 sec
per final i have difference
Total plot creation time was 4306.52 sec (71.7754 min)
and
Total plot creation time was 2983.01 sec (49.7169 min)
Hope it helps to find the problem.
Best regards

[mladentripalo]

i have same problem,

1st plot is usually ok, core drops happen on "Crafting plot 2 of xx"

it almost look like developer "forgot" to release threads or something on 1st plot crafting and it manifests in 2nd plot.

it wouldnt be surprising since most developers only try 1 plot creation in testing.

i have similar system as OP... 5950x with 128GB using 2x980 PRO nvme as temp1 and temp 2 and ssd as staging disk (440M/s)
plotter version is buit on ubuntu 20, commit c40d2a9 (HEAD -> master, origin/master, origin/HEAD)

coommand line:
./chia_plot_k34 -n 11 -k 34 -r 32 -u 512 -v 1024 -t /mnt/nvme2/ -2 /mnt/nvme/ -s /xfr/ -d /mnt/s01/pplt2/

possible fix: (not fully tested but seems to fix it):
./chia_plot_k34 -n 11 -k 34 -r 16 -K 2 -u 512 -v 1024 -t /mnt/nvme2/ -2 /mnt/nvme/ -s /xfr/ -d /mnt/s01/pplt2/

i managed to kinda fix this P2 core drop by reducing to -r 16 and adding -K 2 to the command line, it gets a bit slower P3 but its ok.

[schinzelh]

I'm seeing the same behaviour: chia_plot_k34 plotting k34 - either goes fast or slow.

If it goes slow, the times at [P2] Table x rewrite are x-times longer and apparently only one core is used.
If it goes fast, [P2] Table x rewrite takes roughly same as [P2] Table x scan right before.