When are we going to see cheaper 16~32TB SSDs for the consumer market?
So, anything beyond 8TB is
very difficult to do right now in an M.2 form factor. M.2 limits power to basically 12W (11.55W or something officially, I think), with burst power going up to 25W or something. But that truly is "burst" power, meaning it's for transients and should not be sustained in any way, shape, or form.
This was one of the big problems with PCIe 5.0 drives, keeping power draw below the limits. Phison had to do some serious tuning and tweaking of their controller to get the E26 to stay below the limits, while hitting higher speeds. (See our Max14um article:
https://www.tomshardware.com/pc-com...ed-phison-e26-max14um-2tb-performance-preview)
Most of the other companies basically didn't even try to do PCIe 5.0 until they had 7nm/6nm controllers. Innogrit is the only other company to do 12nm (I think, on the IG5666) with PCIe 5.0, and it was
worse than Phison's E26 — on power as well as overall performance. (See:
https://www.tomshardware.com/pc-components/ssds/adata-legend-970-pro-ssd-review) Now we have multiple TSMC N6/N7 designs, though, and power has dropped quite a bit.
Anyway, the issue is with 8TB on M.2, you're basically capped out on the number of NAND chips already. Each controller has a maximum number of channels and CEs (chip enables), and once those are full, you actually get worse performance if you add more NAND per channel. Data center stuff has up to 16 channels, twice as many as the top consumer parts (and lower tier consumer parts are only 4-channel). So with data center, you can still get a very high performance 16-channel, 16TB SSD, and you have 25W to work with on U.2 and similar.
With consumer M.2 drives, it's theoretically possible to do 16TB with a dual-sided drive, but it would need cutting-edge QLC NAND packages to get there. And then you'd potentially have enough NAND that if it was all active at once, you'd exceed the 12W limit... and QLC would mean it's still slow.
So basically, the 8TB WD Black SN850X is as good as it gets right now for consumer SSDs in terms of capacity and performance (TLC and PCIe 4.0). I think Phison and some of its partners might do 8TB with E28, and possibly Samsung and Silicon Motion could offer that as well. But 8TB M.2 drives are very niche products, and balancing power with performance will be difficult. For anyone who seriously wants a 16TB SSD, picking up a U.2 drive (with an adapter) shouldn't be too much of an ask.
Long-term, I think with more NAND layers and process node shrinks (on the controller), we'll eventually see 16TB M.2 drives. I believe PCIe 6.0 is also looking to boost M.2 power limits to 25W when that arrives (first on data center, naturally). Basically, best-case I'd guess M.2 16TB is at least a few years off still, and when it eventually arrives, it could still end up as a $1000+ niche product.