Note that with a 2TB SSD, the pSLC cache could be up to 500GB in size for a completely empty drive. So, if you could do sustained writes at max speed and fill that up, and then had to drain it at ~100 MB/s, it could take over 1.38 hours just to empty the pSLC to QLC. LOL. (Related: The drives take a while to recover in our Windows testing, unless you just wipe/format them.)

At 2TB for your Steam stash, at least you won't have to swap games in and out as often, which significantly helps to lessen the write burden. I had ARK running once on Linux: It loaded ARK faster from a hard disk than Windows loaded it from NVMe...

The more, the merrier.

It would be interesting to try testing this. Like, a decent SSD and controller should write initially to the pSLC cache, but if it's only at ~40 MB/s, the cache can then be immediately flushed to QLC and would perhaps never fill up (until the SSD is completely full). The problem is that writing even 100GB of data at 40 MB/s takes a while, about 40 minutes. I guess that would be the question: if write speeds are slow, like sub-100 MB/s, do the SSDs even use their pSLC caches, or do they just write straight to TLC/QLC NAND? But... I've also had some very old Android tablets die on storage that seemed to reprogram flash at EEPROM speeds, never giving up ...before I did.

Get ready for an immersive experience with exclusive gaming features including PCIe ® Gen 4.0 3, Western Digital's nCache™ 4.0 Technology, and Microsoft’s DirectStorage Support.abufrejoval said:I guess the biggest question is: how do you ensure it's done steady-state processing before you turn the device off? alternatively, maybe the second m.2 slot could actually be rotated only 90 degrees and is actually located below the second half of where a 2280 m.2 SSD would lay? This would still require a second screw for running two 2230 m.2 SSDs, but it’s still a possibility for minimizing board space. It would be interesting to try testing this. Like, a decent SSD and controller should write initially to the pSLC cache, but if it's only at ~40 MB/s, the cache can then be immediately flushed to QLC and would perhaps never fill up (until the SSD is completely full). The problem is that writing even 100GB of data at 40 MB/s takes a while, about 40 minutes. I guess that would be the question: if write speeds are slow, like sub-100 MB/s, do the SSDs even use their pSLC caches, or do they just write straight to TLC/QLC NAND?Yup, it's at that point when you want to start reading the controller's source code. Though if you fill up this drive's capacity, or throw a lot of big files at it at once without giving it time to write that to the slower QLC, it will drop dramatically to near hard drive levels of performance. You really need that pseudo SLC buffer to keep that QLC from taking the brunt of the write job.

And when the firmware has to deal with things like host buffers, that require interaction with host firmware that could be buggy, too, and simply sprinkle your most critical data structures with random bits, you wonder if these firmware engineers might have burn out or a drinking problem, especially since these junior guys only get to work on the cheaper entry level products, which are much harder to handle than when you've got everything fully under your own control. The problem there is that a single 2280 m.2 SSD is longer than two 2230 m.2 SSDs, so you wouldn’t be able to have a second m.2 slot unless the aforementioned second slot is past where the 2280 m.2 SSD would normally lay.

Revitalize and expand the storage of your M.2 2230 SSD compatible laptop such as many Microsoft ® Surface and Dell™ models.