Curious on how you are going to offload plots to long term storage when plot 2.0 arrives

At least on my plotter, plotting times come with some jitter. So, let’s say that I mostly see around 150 secs, but from time to time there are over 220 plots as well. RAM drive is usually small, so such jitter may hose the plotter for some time.

Although, I have asked Max to write “chunked” plots that could be moved using his plot-sink. Let’s say a plot would be written in 4 chunks. Immediately, after the first chunk would be written, plot-sink could start transferring it. Once such chunk would be moved, it could be deleted. This way, RAM drive could potentially “double” in size, meaning being a bit more resilient for such jitter.

As far as that other question (how big such ramdrive needs to be), it really depends what drives you have, and where in the drive you are with writing. I recently plotted to Toshiba drives, and 3 drives were good enough to support 150 secs plots, so 4x 72 GB RAM would be needed (3x for plots being moved, plus 4th one that is being built). However, before I plotted to WD Essential (white label), and I had to use somewher 4-6 drives toward the last TB.

Are you using bladebit or gigahorse? I heard that bb cannot use ramdisk as destination because it uses direct I/O which ramdisks don’t support

I have tried RAM drive on my plotter…it is slower than SSD way, unless you have a single CPU rig supporting 512 GB RAM.

In my case, moving CPU2/GPU/SSD > 4x HDD stripped JBOD is faster than CPU2/GPU > CPU1/QP link RAM drive > 4x HDD

It may be faster if I have 768GB RAM CPU a RAM drive on the same NUMA node.