SSD wear leveling

Since most of us are using SSD’s for our plotting temp drive, you might find this video to be of interest:

@0:40, the video presents a chart.
For those that do not know the pecking order, from best to worst quality NAND cells, it is:
SLC – single layer cells
MLC – multi layer cells
TLC – tri level cells
QLC – quad level cells

And the NAND type will factor in to the price of the SSD.
Also, even low-end QLC SSDs will have some faster NAND cells for cache, which is why all NVMe drives seem fast (because for general use, you are always writing to the cache).

If you purchase a $7 thumb drive, then do not expect any cache. It will be likely be comprised 100% of the slow stuff.


Any idea for each TB of plot, how many TB is written as temp space on NVME? Trying to get an idea of how much life is left of my NVME.

I think its about 1.5tb - 1.8tb per plot depending what plotter you use.

I have a Samsung 970 Pro. CrystalDiskinfo shows 2,845,709GB writeen and 27% good whereas specs says its limit is 1,200 TBW. Look like it is doing pretty good.

Considering ssd are good for plotting long after expected time, isnt levelling them just wrecking more ssd’s.

It seems they will be fine for plotting long after you would want to use them for long term storage.

So why wreck more than you need to, just use one till its dead.

Thats my thoughts anyway.

…for K32 plots.
(20 characters)

My guesses, as to how the manufacturers derive their “TBW” values, are:

  1. At the time that they released their “new model / new design” to the public, they did not know the answer. It turns out it takes over a year of 24/7/365 pounding (perhaps several years of relentless pounding) to wear out the NVMe drive – and the manufacturer was not going to test it in the lab for years and hold off on selling their product. So they settled on a TBW value that they were positive all of their drives would easily pass.

  2. It makes the manufacturer look really good when their TBW value is 1,200 TBW, and folks have 2x, 3x, or more, with no issues.

  3. The manufacturer found a way to deny warranty claims. If you exceed their TBW value, you take a hike, even though exceeding the TBW value is nearly meaningless, in terms of being the reason for a drive failing.

I have only about 24 TB left to plot after that I am done with plotting for good.

Time to buy more drives!!!

Sounds like a good excuse to re-plot to higher “K” size plots.

Just to waste electric


It does use electricity, yes, but >k32 plots lighten the node’s load, esp. nice for ‘lite nodes’ and/or if this blockchain becomes busy (ier) :disappointed: Also might allow farming a bit more space. Lastly, because something to do :thinking:

I hate to plot/replot because it wastes electricity. No intention of replotting unless we are forced to replot.


That may come sooner than one would think or is being suggested. Advances in CPUs perf are coming fast and furious, not to mention PCI-e 5.0 SSDs and memory. Just watch the next 9-12 months, exciting times. K33,K34 will soon become as fast (and energy efficient) to make as k32 are now.

But why k32 should be obsolete? We should be able to keep them?

1 Like

Take a look at that chart in the lower right corner - Grafana. Currently, 97% of plots are k32 (and decline is rather very slow), so all those reports about the sky falling soon, as just reports about sky falling soon - not worth to spend time on.

1 Like