From what I understand, newer consumer level SSDs are designed to have a lower TBW (TeraBytes Written) - mostly because any reasonable consumer doesn't write terabytes of data to their SSDs in any decent length of time. Enterprise/Higher end "GaMeR" SSDs have higher TBW ratings - I believe 1TB FireCuda Model NVMe SSDs have something insane like a 1800 TBW rating when I was looking a couple weeks back.
These above mentioned consumer drives, when written to in massive amounts (I believe even the lower sized Chia plots thrash the SSD to the tune of at least a couple terabytes of write usage) cause it to degrade at a significantly higher rate.
Long story short, consumer SSDs aren't designed for super high write tasks, while enterprise and higher-end ones are.
That's incomplete, it turns out; apparently it's a weirdly capitalized form of "gamer" with sarcastic/disrespectful intent towards the referenced group.
Firecudas are terrible for Chia. TBW doesn't matter as much as people think. Some of the lower TBW SSDs will outlast the higher TBW ones (again, Firecuda is an example of the latter).
The standard plot size uses 1.2TB of writes.
You won't wear out your SSD in a few weeks with Chia...you just won't. If you use a 120GB one, maybe...but no-one is doing this.
Yes and no. As I understand it, there are two parts: you create plots which take up a certain amount of disk space, and then farm those plots indefinitely, and only the first plot creation part is extremely I/O intensive. So what a lot of people seem to be doing is creating the plots on SSDs and then moving them to cheaper spinning rust hard drives once created, and it's this that can apparently destroy consumer SSDs relatively rapidly (so long as you have much more storage used for Chia than the capacity of the SSD). I don't think just creating a plot or two in the spare space of your consumer SSD and leaving it there farming away would be a huge problem.
Yes, presumably as a measure to make it harder to scale the plotting process (to make it more "fair"), it apparently does a ton of reads and writes across the whole region being plotted, so it will quickly exhaust the total number of write cycles available on the drive. A cheaper drive will be effectively destroyed (high quality drives can survive more write cycles, so they're comparatively okay.) I would have assumed it just linearly computed the whole plot to be farmed, but presumably that would make it vulnerable to some sort of attack where you create a "winning" plot on demand.
You could theoretically plot on a RAMdisk to avoid this problem, but the necessary amount of RAM would be incredibly expensive, so it's unlikely that many people will do it.
If they are used to produce many orders of magnitude more plots than could fit on them. His point is you don't need SSDs to plot, and can reasonably make them directly on the spinning HDD on which you'll store them. I know, because I'm doing this. I make one plot in about 10 hours on the 7200rpm commodity drive in my home media server. It may take me longer to fill up my space than it would with an SSD, but compared to the amount of time in the future I'll be able to farm the plots it doesn't really matter. Cranking out plots as fast as possible probably only helped for the first week or two.
People are burning out SSDs since they're buying into the arms race and trying to get/stay ahead of the netspace growth, hoping for some "easy" XCH, but that's a choice they're making, using poor (or more likely no) ROI calculations and outdated information.
These above mentioned consumer drives, when written to in massive amounts (I believe even the lower sized Chia plots thrash the SSD to the tune of at least a couple terabytes of write usage) cause it to degrade at a significantly higher rate.
Long story short, consumer SSDs aren't designed for super high write tasks, while enterprise and higher-end ones are.