Just the hex value will do. You only can provide one mask per application launch. If you want to launch 2 instances, you need to start 2 processes, each with their own mask.
Hi!
Are the madmax plots valid for farming with the chia gui?
Has anyone tested it yet?
Yes they are valid plots I am gui farming about 40 (and rising) of them in addition to those made by the default plotter.
1 Like
Hi everyone, first post here for me 
Here is my config and my log:
5900X, 128GB, 2x samsung PM983 1.92TB and 2x auros gen 4 2TB.
I’ve done lot of plots with chia gui then switched to mad max plotter and IT IS AMAZING:
All this under win10, 112GB of ramdisk using ImDisk and only one PM983
Final Directory: g:/
Number of Plots: 1
Crafting plot 1 out of 1
Process ID: 12220
Number of Threads: 16
Number of Buckets: 2^8 (256)
Working Directory: f:/
Working Directory 2: r:/
Plot Name: plot-k32-2021-06-19-09-08
[P1] Table 1 took 15.2352 sec
[P1] Table 2 took 99.2911 sec, found 4294972039 matches
[P1] Table 3 took 136.156 sec, found 4295104076 matches
[P1] Table 4 took 165.587 sec, found 4295008883 matches
[P1] Table 5 took 162.126 sec, found 4295013833 matches
[P1] Table 6 took 154.814 sec, found 4294920977 matches
[P1] Table 7 took 105.718 sec, found 4294820687 matches
Phase 1 took 839.118 sec
[P2] max_table_size = 4295104076
[P2] Table 7 scan took 9.28869 sec
[P2] Table 7 rewrite took 29.009 sec, dropped 0 entries (0 %)
[P2] Table 6 scan took 24.8708 sec
[P2] Table 6 rewrite took 45.8977 sec, dropped 581277741 entries (13.5341 %)
[P2] Table 5 scan took 23.7671 sec
[P2] Table 5 rewrite took 43.8244 sec, dropped 762033606 entries (17.7423 %)
[P2] Table 4 scan took 23.3956 sec
[P2] Table 4 rewrite took 43.1953 sec, dropped 828870903 entries (19.2985 %)
[P2] Table 3 scan took 23.7961 sec
[P2] Table 3 rewrite took 43.3023 sec, dropped 855196367 entries (19.911 %)
[P2] Table 2 scan took 23.5402 sec
[P2] Table 2 rewrite took 43.3859 sec, dropped 865545187 entries (20.1525 %)
Phase 2 took 379.471 sec
Wrote plot header with 268 bytes
[P3-1] Table 2 took 32.1115 sec, wrote 3429426852 right entries
[P3-2] Table 2 took 30.2505 sec, wrote 3429426852 left entries, 3429426852 final
[P3-1] Table 3 took 39.5688 sec, wrote 3439907709 right entries
[P3-2] Table 3 took 30.2947 sec, wrote 3439907709 left entries, 3439907709 final
[P3-1] Table 4 took 38.6425 sec, wrote 3466137980 right entries
[P3-2] Table 4 took 30.5245 sec, wrote 3466137980 left entries, 3466137980 final
[P3-1] Table 5 took 39.6447 sec, wrote 3532980227 right entries
[P3-2] Table 5 took 31.3563 sec, wrote 3532980227 left entries, 3532980227 final
[P3-1] Table 6 took 41.4649 sec, wrote 3713643236 right entries
[P3-2] Table 6 took 33.0071 sec, wrote 3713643236 left entries, 3713643236 final
[P3-1] Table 7 took 37.1343 sec, wrote 4294820687 right entries
[P3-2] Table 7 took 37.6865 sec, wrote 4294820687 left entries, 4294820687 final
Phase 3 took 422.588 sec, wrote 21876916691 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 127.934 sec, final plot size is 108833943493 bytes
Total plot creation time was **1769.19 sec**
Started copy to g:/plot-k32-2021-06-19-09-08-
Copy to g:/plot-k32-2021-06-19-09-08-k **466.045 sec**, 0 MB/s avg.
Multi-threaded pipelined Chia k32 plotter - aff2601
Build 0.5.0 for Windows.
Soon I’ll switch to ubuntu linux
2 Likes
Nice!
Have you won any chia yet? Or have you just farmed with the madmax plots?
I would like to try it out, is there a way to check if the plots are working and can generate chia?
1 Like
I am only at 116 plots, I have not won any coins. I am shooting for about 1000 plots, but a bit loathe to go at it hard with every machine until pooling is available, a poolable plot seems a much more useful thing to have, and K33 at that, given that there is a reasonable chance k32 will be deprecated if people continue to demonstrate their ability to madmax plot in minutes not hours. MM doesn’t do K33 yet.
1 Like
Yes!
i‘m also unsure about how long k32 will be up to date. Can you already plot k-33 oder is this just the next standard, which is coming as soon as pooling gets available?
i just saw, that k-33 is already available!
do you recommend to plot k-33 instead of k-32?
If your concern is fast plotters will make K32’s obsolete as the driver to create K33’s then there is no need. Chia has explained that even with theoretical near future speed gains, it’s 10x more economical to just store the plots.
And in any case they would tweak the plot filter to be less aggressive, and not the K plot size to counter that.
K33’s are only useful to fill up your disk space more perfectly as they are slightly larger than 2 times the K32’s
1 Like
k-33 available via madmax plotter or via the chia default plotter? You can definitely plot k33 now using the basic plotter, my farmer machine plots away quietly 3-4 K33’s day without tying up too many resources. My main plotter does not it just plots K32’s via Madmax (fury road) windows fork.
1 Like
I haven’t had any trouble with my RAM, when I get some time I was going to bump it up to 3800 and test it out. Currently at 3600, 22-22-22-52 w/1.40v. Will try 3800 @ 1.45v. For reference I’m using the MSI MEG X570 ACE Motherboard with two sets of TEAMGROUP T-Create Classic 10L DDR4 64GB Kit (2 x 32GB) 3200MHz (PC4 25600) CL22 Desktop Memory Module Ram - TTCCD464G3200HC22DC01.
I haven’t ran the Ubuntu benchmark but will check it out when I get home. It would be interesting to compare the 3 scenarios I’ve ran (XFS for all plotting drives). Using my Samsung 980 Pro 1TB (firmware updated) that I ran the initial test on it always kicks out 1130-1160 plots. I have 3x Samsung 980’s, I tried a second one and my times were 1260-1300?!? Same exact drive, purchased at the same time and firmware updated. I haven’t tried the 3rd drive yet but will today/tomorrow. Now even more interesting my 300GB 10K SAS RAID0 array puts out 1218-1260 plot times!! That I was overly happy with as then I’m not killing and NVME drives So I’ll benchmark the two Samsung 980’s and my SAS setup to see what the numbers look like.
1 Like
I’ve got a 3970x w/256GB DDR4, 2x Samsung Pro 980 2TB’s, 2x 109GB TMPFS Ram Drives and run two parallel jobs. With them running at the same time they each finish a plot every 1400-1500 seconds This gets me about 120 Plots per day:
chia_plot -n 60 -r 24 -u 512 -t /ssd/ssd1/ -2 /ssd/mem2/
chia_plot -n 60 -r 24 -u 512 -t /ssd/ssd0/ -2 /ssd/mem/
looks like madmax plotter + k32 plots are the winning strategy, right?
i‘m just a little bit insecure, because madmax plotter is so much (maybe unrealistically) faster than the normal official Chia plotter
As of today, and to the best of my knowledge, the fastest plotting strategy is to use madmax plotting one plot at a time, provided you can max your CPU out using either ramdisk or a fast nvme/ssd or both (use the ramdrive as temp2). Madmax is only fast as it segments the calculations into chunks and feeds them into all the threads simultaneously, whereas the default plotter really only ustilises 4 threads at most.
If you cannot max out your CPU running madmax because your ram is slow, or your disks are slow, then you can run multiple instances (I am running two, which gives me two plots every 10000 seconds, if I run one I get 1 plot in 6000 seconds).
However, you can also run a large number of the default plotter processes in parallel, either manually or using a manager like swar, plotting 10-15 plots in parallel to 10-15 drives at a time, which will give you perhaps slightly fewer plots per day, but certainly in the same ball park.
Your system can only process and load/save a fixed amount of data per day, madmax just offers a more optimised method of doing it, and one that you can interrupt every 2000-6000 seconds rather than only once every 12-18 hours.
1 Like
One more thing to add to @DjDemonD response is that if you have enough RAM, then you can use a ram disk with madmax that will take on 75% of the writes instead of your SSD, and this will put a lot less wear on the SSD, making plotting greener
3 Likes
@Chiafanboy
nope, 661 plot and still no chia for me.
After creation I’ve checked them with “chia plots checks” and it seems they’re good.
@SB1Racing
awesome results, really good.
Guys madmax plotter is amazing because it writes 75% of data on the ramdisk and 25% on the NMVE expanding, de facto, the life of our drives.
Madmax and buckets. I know everyone seems to be using 512 because it uses less ram saving more for ramdrives, right? But what is actually fastest? I’ve tried 64 and it was faster. I tried 32 as well but ran a second process and two ramdrives and even though I have 320GB of ram crashed my plotter down to zero free ram.
I would run the fastest config if it can get me under 5000 seconds/plot and just run one at a time.
I could see the logic of more buckets being broadly as fast as fewer if you have very fast disks, but mine seem slow maybe fewer buckets means fewer disk ops.
There seems to be a new arugment:
-v, --buckets3 arg Number of buckets for phase 3+4 (default = buckets)
So whats the logic here, more buckets or fewer is faster for phase 3&4?
What is the bug you are mentioning?
128GB memory is a common config and the previous default of 128 buckets would allocate 1GB per thread.
110GB needed + 1GB per thread + some GB for OS = 128GB
left systems with more than 16 threads out of memory.
madmax then moved the default up to 256 buckets (0.5GB per thread), and now a 24 thread system would be OK.
if you have even more threads then 512 buckets (0.25GB per thread) may work better, just have to run a few tests…
the new argument was introduced because the 4 phases don’t have the same resource needs (CPU, RAM, DISK) and this allows you to tweak them better.
again, you need to experiment what works best with your system… knowing that with just 128GB RAM a larger bucket number is probably better… otherwise your OS will be out of RAM
1 Like
It is all dependent on your system. 512 was the fastest for me. I tried 32, 64, 128, 512 and even 1024. 512 was the fastest for my setup.