Hi, can someone help me run plotter on Linux? Do I need to compile it or something? When I try to start it with command chia_plot it say command not found. I starter it on WIn but want to try on Linux. There should be some better instructions on git page, and some wiki.
Second question is how much ram memory do I need for ram drive if I want to use 40 and more threads? I have some old server with 40c/80t and 256GB of ram, few nvme disks in pcie risers and want to try it. Anyone can help with starting a plot in linux ? Thanks
These instructions helped. There’s a bit of brushing up to do. MadMax Plotter: Simple workaround for Windows (running plotter in Ubuntu subsystem)
So I tested Mad Max on my Office/Video Editing rig. Threadripper Pro 3955WX (16-Cores, HT off, running at 4.14GHz during Mad Max and using 12-cores), RAM is DDR4-3200 CL22 in octa-channel configuration. SSDs are 4x 980 Pro 2TB in RAID0). Total plot creation time 2560 seconds (ca. 43 minutes).
This must be slower for you than doing a lot in parallel, right ?
I am using 5950x with 128GB RAM (3600 DDR4). . I am using Ramdisk and 2TB Rocket Sabrent 4. My times are between 2250 and 2400 seconds.
I’m surprised the Threadripper isn’t out ahead. Maybe a limitation of the ram speeds?
Ryzen 1500X 4C/8T
Intel SSD D3-S4610 960GB
16GB RAM
+/- 2 hours 
./chia_plot --threads 6 --tmpdir ~/chia/tmpdir/ --tmpdir2 ~/chia/tmpdir2/ --finaldir ~/chia/final/ -p my-pool-pub-key -f my-farmer-pub-key
Multi-threaded pipelined Chia k32 plotter - d66fd5e
Final Directory: ~/chia/final/
Number of Plots: 1
Crafting plot 1 out of 1
Process ID: 17498
Number of Threads: 6
Number of Buckets: 2^8 (256)
Pool Public Key: my-pool-pub-key
Farmer Public Key: my-farmer-pub-key
Working Directory: ~/chia/tmpdir/
Working Directory 2: ~/chia/tmpdir2/
Plot Name: plot-k32-2021-06-12-16-32-xpto
[P1] Table 1 took 71.608 sec
[P1] Table 2 took 364.775 sec, found 4294910735 matches
[P1] Table 3 took 482.34 sec, found 4294826485 matches
[P1] Table 4 took 544.299 sec, found 4294585089 matches
[P1] Table 5 took 504.301 sec, found 4294240264 matches
[P1] Table 6 took 439.784 sec, found 4293535902 matches
[P1] Table 7 took 346.263 sec, found 4292000871 matches
Phase 1 took 2753.41 sec
[P2] max_table_size = 4294967296
[P2] Table 7 scan took 109.979 sec
[P2] Table 7 rewrite took 200.711 sec, dropped 0 entries (0 %)
[P2] Table 6 scan took 84.4423 sec
[P2] Table 6 rewrite took 178.031 sec, dropped 581453287 entries (13.5425 %)
[P2] Table 5 scan took 82.2283 sec
[P2] Table 5 rewrite took 175.61 sec, dropped 762174648 entries (17.7488 %)
[P2] Table 4 scan took 88.3623 sec
[P2] Table 4 rewrite took 175.36 sec, dropped 829007755 entries (19.3036 %)
[P2] Table 3 scan took 82.5012 sec
[P2] Table 3 rewrite took 171.784 sec, dropped 855222036 entries (19.9128 %)
[P2] Table 2 scan took 85.0571 sec
[P2] Table 2 rewrite took 172.035 sec, dropped 865624382 entries (20.1547 %)
Phase 2 took 1614.14 sec
Wrote plot header with 268 bytes
[P3-1] Table 2 took 189.718 sec, wrote 3429286353 right entries
[P3-2] Table 2 took 176.709 sec, wrote 3429286353 left entries, 3429286353 final
[P3-1] Table 3 took 212.881 sec, wrote 3439604449 right entries
[P3-2] Table 3 took 175.931 sec, wrote 3439604449 left entries, 3439604449 final
[P3-1] Table 4 took 215.556 sec, wrote 3465577334 right entries
[P3-2] Table 4 took 177.493 sec, wrote 3465577334 left entries, 3465577334 final
[P3-1] Table 5 took 218.02 sec, wrote 3532065616 right entries
[P3-2] Table 5 took 180.134 sec, wrote 3532065616 left entries, 3532065616 final
[P3-1] Table 6 took 227.557 sec, wrote 3712082615 right entries
[P3-2] Table 6 took 189.785 sec, wrote 3712082615 left entries, 3712082615 final
[P3-1] Table 7 took 259.2 sec, wrote 4292000871 right entries
[P3-2] Table 7 took 219.704 sec, wrote 4292000871 left entries, 4292000871 final
Phase 3 took 2445.42 sec, wrote 21870617238 entries to final plot
[P4] Starting to write C1 and C3 tables
[P4] Finished writing C1 and C3 tables
[P4] Writing C2 table
[P4] Finished writing C2 table
Phase 4 took 152.216 sec, final plot size is 108795819253 bytes
Total plot creation time was 6965.24 sec
Yeah maybe, as I said, it runs with 3200 and CL22. Currently I am running a test with 14 threads. After that I will test it with a ramdisk too. Ah, btw, it’s under W10, not Linux, if this still matters in plotting speeds with Mad Max.
Very interesting and opposite of my observation. With my 5950x, X570 tomahawk, 64gb ddr4 3200 Cl16 ram, and 2x gen3 2tb sx8100 nvme ssds I got around 2600sec. (win10) But when I watched I noticed both ssds were not usually in use at the same time. So I figured perhaps using a single for both the temp and temp2 dir would give similar times . However I found that when using only one of the nvme gen3 ssds, my time jumped from 2600 to 3600 seconds.
Which gen3 nvme ssds are you using? Linux or windows? Any over clocking? You numbers seem surprisingly good for a single gen3 nvme ssd. But congrats!
Are you overclocked or running stock Intel specs? Do you have a solid version of windows that isn’t bloated and taking up resources? My buddies 9700k is beating you on a bloated win10 not clean, overclocked to 5ghz all core and 3600 xmp ram.
I’m plotting on my overclocking test rig, 9900k 5.2ghz/48 cache and 4400c16 memory overclock extremely tight tuned ram is pushing almost 70k read write and copy speeds in Aida.
Look at your SSD its at max 100% writing capacity, its as fast as it can write. That is your bottleneck.
System Specs:
5950x
Wraith prism cooler
128gb Corsair 2666Mhz
3x Micron 7300 max 1.6Tb in raid-0
(I have a few 980 pro but no time to switch this weekend)
110gb ramdisk on Ubuntu 20
Spits out 57 plots a day/25.5 minutes per plot written to a hard drive.
3 Likes
Bucket 64 not worth it?
That bucket 256 looks good, what changes cores/RAM wise?
1 Like
Except that with MadMax a well controlled SSD with a large enough SLC cache should be able to avoid cache misses altogether, so cache size and cache controller behavior and sustained burst (temps!) is what to look for in that case.
All cores get in. Test1/3/4 using RAM disk. Not record RAM usage cause it doesn’t matters. It already enough.
Nope gen 3 will do the same you guys need to overclock all core and disable power limits you are cpu limiting yourself. I ran 1 plot on gen 3 WD black with same results. Even 4.8ghz all core no power limits will be way faster
I could only imagine how fast I could plot on my game rig 10900k 5.5ghz/5.1cache
4700-16-17-17-34-280-2t
It’s about 20% faster than 9900k rig on test bench that I’m plotting with on cpu and 80k bandwidth on ram.
2 Likes
On my upcycled Dell R720 I am going a lot slower than most of you guys, maybe 3 hours/plot but as I don’t seem to be maxing out the CPU, maybe my ram(drive) is too slow (DDR3) I am running two madmax plots in parallel. This is churning out 2 plots every 3 hours, but keeping phases 1 and 3 out of sync between the two processes, is difficult.
What are the issues switching to K33’s with this plotter? Same increase in required temp space as with the official plotter? It seems according to madmax’s github that 256 buckets is now the default, any idea what the idea behind this is?
Certainly happier with this plotting method, woke up this morning to a whole bunch of new plots, parallel plotting for me was not working all too well.
1 Like
Had to add a fan on the watercooling and another one on the NVMe area. By doing so losing just 1 plot is a big plus. 
i have a AMD Ryzen 7 with 32 gb ram. plenty of hard drive space. i can get the plotter to work but it still takes forever. what am I doing wrong… I really wanted to make this process faster.
for($i=1; $i-le 10; $i++)
{c:\users\chala\downloads\madmax4\chia_plot.exe KEYS HERE -t f:/ -2 e:/ -r 8}
can someone help me please…
Thought of trying this:
Waited till the mid of phase 2 and launched one parallel.
LOL, it took 2 hours 30 minutes to finish the one I newly lauched. The other one got the time double as well… that’s that in my case.