How many Nvme drives should i use

I have a BTC 250 with 6 GPU cards.

  • I currently have 8Gb/16Gb ram but can add to 32Gb
  • CPU is i7-7700 engineering release i.e. 2.4GHz 4 Core / 8 Threads
  • mb has 1 m2 nvme slot

I was thinking to add a nvme riser.
what is the max number of plot drives i can add with these specs to use madmax with the hiveos or raveos

I’m confused about your question. Are you looking to turn your machine into a chia harvester or chia plotter?

I don’t think that with this setup you can really run more than one plot at a time with madmax (at least it wont do you much good). So one nvme is enough.
What I would recommend is to add a cheap sata ssd for the OS and then use the nvme only for plotting.

You should get the full 32GB of RAM but even then you do not have enough RAM to use madmax so you need to plot on your NVMe.

The size of your NVMe plotting drive is dependent on the size and parallel number of plots you would like to make.

I plotted 3 parallel K33s at a time on a 2TB Samsung Plus NVMe. You can plot 6 parallel K32s on a 2TB NVMe.

Depending on the number of parallel plots and what size they are you need at least a 1TB NVMe, but are better off with 2TB or even a 4TB NVMe plotting drive.

Madmax uses about .5gb of ram per thread and the machine has 8 threads, so I don’t see the point of upgrading the ram.

1 Like

gwaitsi cannot install enough RAM to use Madmax properly so gwaitsi will need to use the Chia client to plot.

I have 32GB and use over 70% of it while plotting using the Chia client.

Good for you! 20 characters

I have 60GB ssd drive for the OS, so the questions are;

  • how many plotting drives should/can I make use of i.e. 1 / 2 / 3?
    (given some CPU/Threads is also needed for GPU mining -
    i initially thought 2 CPU / 4 Threads for Rave OS - but is 1 CPU / 2 Threads enough for that)
  • based on the number of drives you recommend, what ram would i need to support the max plots per drive?

Obviously, my goal is to maximise cpu and gpu utilisation of the box


can’t use the chia client i believe, because it has the private key. and hiveos/raveos are a security risk. but madmax only needs the public key i believe

You don’t need the chia client to plot, the security risk is basically non-existent. If you want to make it a plotting machine, you’re going to be maxing out the machine’s capacity with 1 nvme drive.

1 Like
  • What is the CPU//Thread/RAM requirements per plot?
  • Is it not better to make 2 plots over 2 drives, rather than 2 plots over 1 drives?

I think you should do a bit of reading about Chia, because I can’t really understand your questions and what you’re trying to achieve.

For Madmax plotter (also build into chia client now): more threads = faster plotting. it will max out available resources. It is designed to make plots in serie, one at a time.
Stadard plotter: more complicated, 1c/2t minimum per plot, how many plots in parallel and how many treads to use for optimum output per day: trial and error

For above reason most prefer Madmax, much easier to use.
32GB should be fine for running one plot on Madmax, I used that as well.

In general yes, but also depends on where your bottleneck is (CPU/Memory/RAM/NVME speed). Like I said above Madmax was designed to make one plot at a time, just much faster than the standard plotter so you don’t need to bother with running multiple plots at the same time.

fine as long as you don’t run a chia client on it (blockchain is more than 30Gb now)


I’m assuming you’ll use Madmax for plotting.
Your magic words: 8 Threads and Intel B250 chipset
8 Threads: Madmax uses them all.
Intel B250 chipset: Supports Gen 3 PCIe.
gen 4 nvme is unnecessarily expensive for you.

Just you need 1TB Gen3 Nvme (2TB recommended. 2X TBW)
Which should I buy?
970 evo plus, mp510, sn750, firecuda 510, sx8200, crucial p5
Which should I not buy?
970 evo, 980, sn550, barracuda, sx8100, mp600 and all other nvmes

so you think no value to get two nvme drives and plot 2 or more at the same time?

Madmax uses all CPU cores. More nvme won’t do you much good. More core CPU upgrade makes more sense.

If you want an extreme experience. You can use 2 nvme and 2 parallel madmax. This is difficult to do. When phase2 is over, you run the second madmax. It will still require a CPU core upgrade.

1 Like

Something to consider gwaitsi … take all suggestions and distill what you can from it all. But, IMO, you can’t take anyone’s advice and just expect it to be best. Everyone has a setup they like, maybe they tried many setups, maybe only one or two, but it is not your setup.

You need to be adventurous, get some gear, then experiment. Sure you won’t get it perfectly best right away, or even ever, but that’s the way to success…try, try, and try again. You’ll learn and evolve towards the best you can mnake it. That’s the end goal.

Consider my setup… you can’t dup it with what you currently have. Nevertheless, you might glean something from it.

  • I never plot with only one ssd. Way to slow, too much work against one drive, compared to…
  • My fav for a MadMax instance is two nvmes, one for -t, the other for -2 … -2 is always my best ssd as it takes the most abuse. IMO, two nvmes is essential to get the best plotting speed (-t & -2). That gives the best minutes/plot except for …
  • As I have a bunch of cores, I typically run two MadMax instances, splitting the cores between instances, with plots going to two -d drives separately to eliminate writing overlap on a single HD. This gives me almost the fastest minutes/plot I can get… except…
  • I can squeeze out a couple more plots/day using a third MM instance (and two more ssds), but it is simply not worth the trouble setting it up really.

So, generally that means 4 total nvmes, two for each instance running. Alternatively, as MM plotting does not use much space, I also have a good 2tb ssd that can be shared for both instances simultaneously. So then two fast nvmes for -2, -2, and the 2tb for -t of both instances. As noted, a good, speedy ssd with high write endurance capability is essential. Don’t skimp.

So my best advice… get two good ssds and plot away with one MM instance, then if you ever upgrade to plot more, faster, you’ll have some of the gear you need already. And 16 gb memory should be just enough for that.

I agree with you but when plotting is finished you don’t need multiple nvmes anymore. So, do you need really faster plotting? Keep money and buy more hard drives

No, but I’m an ssd fanatic, so one (or 2, or 3… or) is never enough :laughing:

So, I have more suggestions. Fully filling an ssd slows it down a lot. Format 50% an ssd for plotting. Example Format 2TB as 1TB. It’s called trim. it will be fast and have long life.