This is something I do not know how to do.
Does anyone have a working version for windows?
Just want to try it out and see how fast it can plot to ram.
I am running it in Linux and got a 32.8 min K32 plot. Problem is because it uses resources so well, it doesn’t lend itself to parallelism. I still get 43 plots per day, same as using pechy combined chiapos and Swar’s.
4 Likes
Thanks for the info, I guess to make it useful it would be a case of creating a script to get it to churn out plots, one after another. I think this might work better for me as I’m not using SSDs, but do have lots of ram. So I could plot temp 1 and temp 2 as ramdrives and throw everything at it. I guess I will just have to figure out how to compile it. Here to learn. 
1 Like
There may be some benefit to overlapping phases 1 & 3 which I’m going to test here in a bit. That won’t include a move from my NVMe raid 0 array to HDD. I suspect that process will slow things down enough to erase any gains.
MaxMax’s plotter changes the way we optimize plotting, in that the official plotter is very in-efficient in allocating all possible resources for a single plot. That’s why we have plot managers / parallel plotting / staggering etc.
MadMax’s plotter on the other hand will use all available resources for a single plot. If you can still create plots in parallel then it there is room for optimization of the plotter (in allocating all available). There are updates to the repo all the time… still very early in development… but very promising.
Let us know if you are able to do anything parallel with MadMax… will be a good indicator where it still can be improved
2 Likes
Thank you for making this software, it is appreciated. If I might be so cheeky, any chance of a guide to compiling for windows? Or a precompiled binary?
Yeah I bottlenecked my NVMe raid 0 so hard trying to overlap phases 1 & 3 on parallel processes that my CPU usage started to plummet. So for me at least it has zero advantage over pechy combined + Swar.
Windows build 
Please keep in mind, not an official plotter… if you can use it in a Virtual Machine… look at the source code etc.
2 Likes
I tried this, pretty cool but for high core count plotters I didn’t get any benefit.
I did do a ~2525.43 second plot. I could see how this would probably speed up a lower core machine.
Hm cant get the stotiks widows-build running. I starts up without complains and exits then.
Edit. got it! missed a restart after vc_redist - installation -.-
can you set the destination-dir in the win-version?
Huh, so only needs to be optimised ~100x more and we’ll all have to start doing K33s.
you don’t find 75% less writes to your NVME an advantage?
only a single plot lost if you have a power outage/need to restart for whatever reason?
that as long as you have enough ram for the ramdisk, it completely eliminates optimisation on a new machine/unfamiliar hardware config. you install and walk away.
MadMax repo is getting improvements at a fast pace… the Windows build is a bit behind.if a feature is missing in the Windows build, you can open an ‘issue’ on the windows repo and ask for it to be merged into the next build
There is some debate going on around that, also on keybase.
I believe that Chia said that once phase 1 goes under 10 minutes they will start to pay attention… grinding needs phase 1 to complete within 30 sec, and we’re not close to that.
maybe when it can be moved to GPU plotting we’ll be there…
something others have brought up is that if k32’s go obsolete this fast, then it could be smart to skip k33’s and go for k35’s right away for example.
in short: k32’s are just fine at this time… but there is progress being made in plotting speeds.
1 Like
IMO at the moment plot speed is everything. In the future reduced SSD wear would be great don’t get me wrong. I also can’t utilize more than 128gb with my current hardware, so for me personally, I am not seeing gains. That doesn’t mean that the server and threadripper guys won’t get an amazing boost out of this. I think we just need to test the crap out of it and see where the benefits actually lie.
How does it make 75% less writes?
Also I wonder what the result would be if you divide this in a couple of VM’s
You only get this if you are using a ramdisk with at least 110 GB for temp2 as 75% of the writes are written on temp2.
Ah ok yes I missed that sorry, using a ram disk yes that saves your ssd
this program doesn’t speed up plotting, it just helps you use more cpu if you didn’t know how to do it with the official plotter
Your comment is a bit misleading…
It greatly increases the speed of a single plot… that’s what it is designed for.
Here is a good discussion on Serial vs Parallel plotting:
Now the real question is… on Average will it be faster if you plot 10 in serial with this plotter or use the official plotter and a plot manager like SWAR to optimize the parallel plotting of 10 plots…
That depends on your system.
1 Like