I am using ImDisk. It is free, and quite good. I didn’t have any issues with it at all, but had to give it 112 GB RAM. And yes, that is for temp2 (-2 f:\RAM_DISK).
CPUs are shared resources, as CPU is used on a thread basis, and you can have zillions of threads running. On the other hand, RAM is exclusive (I think). Whatever you allocate to your VM, it will not be shared. Therefore, if you overbook, it will crash.
Primo Cache is NOT a RAM drive, but a cache in front of whatever you point it to. What it means, it will try to keep in cache some data potentially without writing it to the destination. The reason to split that drive into two folders is to exclude writes that are done to temp1. Primo doesn’t know which writes really needs to cache, so it goes with what comes first. tmp2 folder is used for a bunch of rapid file creations/deletion, so those are good writes to cache. On the other hand, tmp1 is for bigger files that are kept around for longer. Therefore, if Primo will get stuck with tmp1 files, it will be running in a less efficient mode…
There are 2 types of plots, OG plots ( the original plots ) can only be solo farmed really, cheating can go on if you join a pool with them.
Chia created an official way to pool, but to do this you must create a nft ( smart contract ) and plot to that.
Both created the same way with 1 exception.
In your command you use -c and -f.
To make a nft plot you replace you -c with -p , -p is the pool contract address and can be found on the nft once its created.
I see , If i can get down from 92min to 30minutes it would be more than enough. Maybe i’ll just look into adding up 128GB RAM and use it with ImDisk . Probably following your setup is enough already for me. May i know what setting do you use on your Madmax ? like how many threads,bucket,parallel or queue and more if you dont mind sharing
So you mean not only farming has pool but plotting also have pool?
So in order for me to join the pool is that i need to create NFT and plot to that . As for my current setup which is use -c is actually solo plotting. Is it correct?
Turn the VM off to see if that makes a difference.
IMO you should be hitting something like 40 min plots using the most standard stetting in Maxmax and 0 tuning.
so 92 minutes, something is up with the system.
When I was plotting on nvme, I saw very little difference going from Windows to Linux.
But it’s good to try out anyway. Software and hardware combinations can have all sort off weirdness, so it might just work great for you.
Just install bog standard Ubuntu 20.04 LTS
The only thing you want to do during Ubuntu install is to check the box “install third party software”
When it finish install open a terminal and run: sudo apt-get update
Then just follow the install instructions on Madmax github page and you should be good to go.
If you still get similar results, next stop is BIOS to see if anything is wrong with CPU, memory or PCIE config.
On my first plotting rig, my Asus mainboard had completely messed up settings out of the box causing my CPU to run way too hot. Bios update fixed that.
Thanks! , i just look into the video and the explaination was real easy with him.
looking into other youtuber made me confused . Meantime i do also look into how he do parallel plotting. So i have a 18TB of HDD which in real one would have 16TB of free space.
So if i’m doing like what he did how do i do the math for calculating my NVME space (2TB Samsung 970 Pro Plus) and fill up the HDD ? by doing parallel might save some time too.
if 1 plot is 239gb , so i have a 1.6TB of free space on that NVME . Then, i can run 6 plot by grou queue at one time simultaneously. Is it correct? but how much plot can i queue at one time for each group?
Oh wow if it’s doing 92 minutes that might be the issue to. I went and stop the plotting and take a look into BIOS . Saw that the Memory was locked on 2133Mhz . meanwhile the RAM is actually 3000Mhz rated. So i changed it to 2933Mhz
As for CPU i didn’t change anything . I’m not pretty sure what to do for the CPU as it have alot of setting in it. So now i’m running it again to see if it gives the same result.
Thanks bud ! by following from the youtube u given this is my current setup for my madmax setting. Is it okay or just weird? I’m sorry if i’m not understanding it quickly might take sometime for me to get it.
I think that he was still using the original plotter (that comes with Chia install). It is (maybe was, but potentially no one bothers trying it right now) really deficient, comparing to MM, so no point to use it. That plotter was done by software engineers that had no concept of threads, so more or less everything is serialized. Thus, CPU usage is low at one physical core level, as such you can/should do parallel plots. MM on the other hand is heavily parallelized, so no point to run two instances (unless you have two CPUs - from my experience).
Both plotters are using that 239GB only in short peaks, so you can easily fit 8 parallel plots on that NVMe, although you need to stagger those plots, to better share resources. Again, just stick with one instance of MM.
Did you enable XMP? That is the proper way to set RAM speed.
As mentioned above, the best change for CPU is not around CPU per se, but rather the cooling option. Otherwise, you will get stuck with 65W only, and long plotting times.
That is good. I would still push your -r value to 22, even with your VMs (unless you are using it for everyday stuff). Also, that -u value can be either 8 or 256 - same thing (the first is the exponent).
madmax gui settings:
threads 18 (20 is slower, 14 and below is slower)
-U buckets 512
-V buckets 256
it alternates between temp1 and temp 2 drives
Final plot writes to an old samsung SSD evo and i robcopy it from that automatically to my farmer drives
[2021/09/30 02:24:33] Phase 1 took 1017.96 sec
[2021/09/30 02:33:04] Phase 2 took 511.056 sec
[2021/09/30 02:42:07] Phase 3 took 543.156 sec, wrote 21877329401 entries to final plot
[2021/09/30 02:43:51] Phase 4 took 103.065 sec, final plot size is 108836253563 bytes
[2021/09/30 02:43:51] Total plot creation time was 2175.38 sec (36.2563 min)
That’s great but you have 2 NVME which give a bit more power than me
Just enabled it already it seems like nothing change apart from getting 77min avg now
I have a big fan for the CPU currently no setting done with it and using 100% of it when running madmax. It still keeps the temp below 67c . the speed is 4.59Ghz . I guess i need to add either another NVME or bumped up the RAM to 128GB to get it down to avg 35min
Whoa! 6min per plot? . If you dont mind sharing , what setup are you using for that?
Yeah, I was just going after docs to find out why I was using -c instead of -p.