You can use this script from MadMax Discord.
Open the terminal. Go to the directory where you want to create your script.
Create a file with . sh extension.
Write the script in the file using an editor.
Make the script executable with command chmod +x .
Run the script using ./.
Ryzen 3700x
Optane 900p 280gb(260 real) as temp1 and temp2
and adata nvme just to copy plots
~38min plot ~1.5min copy
Finally I found a way to speed up the plotting. I add a 512GB NVMe and run two madmax plotter in this system with 2 ramdisks, each plotter use one ramdisk and share the same NVMe. The creation time for each plot by each plotter is about 38min, which mean I have 2 plots in 38min. BTW each plotter is using -r 32.
Ah nice! Never got Bladebit going though? I don’t have a system with 512GB to test on. Pretty sure I could probably get it going. I’m at like 160GB of RAM on one of my systems. In the process of upgrading it to 256GB. Probably gonna try out what you suggested here. I’d like to upgrade it to 512GB, but I’ve already bought 160GB of freaking 16GB dimms. I don’t feel like reselling them all and buying 32GB dimms. Lol.
Whats brand and gen ? Windows or Ubuntu? Real nice times .
ryzen 5 3600
DDR4 16GB 3200MHz
M2 500 Go 970 EVO Plus
SSD 1T with Windows 10
AsRock B550 gaming 4/ac
(all new and default config; no OC; no bios update after doing a CMOS on the mother board due to an issue)
Madmax best config so far with NTF plots: -t M2 (same for tmp2) -r 10 -u 256 -v 128
I tried many combinations for -u and -r and this is fast I can plot (tried -u 512 and -r 12)
75 min per plot
Any suggestions to improve this?
Hey mate try -
-r 16 -u 256 -v 128
I currently use r16 and get 52-54mins per plot. It does work for some reason. Also my RAM default is 2133, i have XMP turned on for 3200
thanks for helping.
First I updated the bios: no major change (within 73-77 range)
Used r16: no major change (within 73-77 range)
OC the memory to 3600: gained 2 min
so apparently this is my limit…
cheers
Mine is Ryzen 7 5700G (CPU benchmark 23663) one 500GB NVMe as -2 and one 1TB NVMe as –t. Machine cost is $730. 50 – 55 minutes / plot. Ryzen 5 3600 CPU benchmark is 17859 (PassMark - AMD Ryzen 5 3600 - Price performance comparison ) so I think 70 minutes it is about right.
I see this is an old post, but if you are still curious my results:
i7-10700K - 64GB/no RDSK/Ubuntu - T1&2 Nvme/R12/U256 - ~57 min per plot
i7-10700K -128GB/RDSK/Ubuntu - T1 Nvme/T2 110RAM/R12/U256 - ~33 min pp
Memory was DDR4-3200 - trying DDR4-3600 tonight as an experiment. Loving the results with T2 in RAM. When I drop back to STD SPD memory speed plot time is ~39 min pp - hoping the DDR4-3600 will get me close to or under 30 min pp.
Had 3 identical machines configured for parallel plotting and struggling to keep 20 ppd on each. Now getting 40+ ppd from one.
FYI - Change to DDR4-3600 showed a modest gain (average for DDR4-3200 was 33.6 mpp - average for DDR4-3600 was 32.1 mpp). Will be testing some conservative CPU OC results next.
I have 64gb of 3600 CL14 ram on a 5800x with two gen 3 nvmes (Hynix p31 and an Adata 8100 or something) and I am getting 33-34 minutes per plot. Try to tighten your timings on the ram.
i7-10700K simple OC to 4.9Ghz resulted in 31.5mpp average.
Latest changes and some advice (lesson learned). I cloned my first system to set a duplicate to do some development while leaving the first alone to plot. Even though I copied the BIOS settings, physically cloned the boot SSD and have all the same hardware in the second system, I was consistently getting ~55mpp instead of 31.5. After a lot of digging, I finally noticed that the Nvme in the second system was formatted in FUSE instead of EXT4 
Reformatted and problem solved - now both systems consistently plotted in about 31.5minutes (including background file copy script).
Went on to add a second Nvme & RAID-0. Couldn’t find how to get Intel RST to play nicely with Ubuntu so used mdadm to do software RAID-0. Wasn’t expecting any miracles at that point but final times improved a little again:
29.4min first plot / 30.6min avg for plot 2+ (including background plot move).
I’ve just been having a try with using only one temporary drive - all ramdrive, no ssds. I tried a 256GB ramdrive and it managed 1 plot but then crashed as it couldn’t copy it off to the final drive location fast enough and ran out of temp space.
Final drive is a local sas drive, so not a slow network connection.
Current attempt is with 280GB which seems to be working.
I presume 256GB would work if you set the “copy before starting next plot” argument.
Plotting 1800 seconds per plot on a Dell R620 dual e5-2670, with 384GB ram in Ubuntu 20.04.
Put somewhere -w
-w (=waits to start new plot till the copy is done)
Or just allocate more ram to the ramdrive.
Hello, I wanted to share information and data about what is happening to me with my configuration to plot chia with madmax.
I have a 128 G ram machine, a Rayzen 5 1600 6 core 12 threads processor, and two 1 Tb M.2.
I’ve been testing various settings, to get the best performance from my pc, but I can’t get below 65/70 min, and from everything I’ve been reading about this, it should be close to 32/40 min.
Next I am going to leave the performances that I have tried to use, to see if someone tried and the same thing happened to them.
a) -n 1 -r 12 -u 256 -t temp1 -2 temp2 -d destination1
b) -n 1 -r 12 -u 128 -t temp1 -2 temp2 -d destination1
c) -n 1 -r 12 -u 512 -t temp1 -2 temp2 -d destination1
d) -n 1 -r 12 -u 512 -v 128 -t temp1 -2 temp2 -d destination1
e) -n 1 -r 12 -u 512 -v 256 -t temp1 -2 temp2 -d destination1
And then, I was changing the threads (-r)
Here is your real limiter: 6 core 12 threads processor.
Use RAMDisk as temp2.