ASUS ROG Strix SQ7 Specifications Finally Revealed

[TheLostSwede]

That's not true. When data is written to a NAND cell, it is written in stages. The first voltage phase sets a state in the chemistry that indicates a set bit. Then the second stage of data is written and another voltage is applied to the NAND cell to further alter it's chemical phase state. Then a third and in the case of QLC, a fourth. This is how NAND works(in VERY basic terms) and is also why NAND wears out faster as you increase the number of bits written. The chemistry of a NAND cell can only tolerate a certain number of voltage induced phase changes. Additionally, another phase change is induced to set the cell to "zero". The data in NAND cells can't just be changed on the fly. To change the data, the cell must first be zero'd out and then new data written. This why MLC is slower than SLC, TLC slower than MLC and so on, and this is also why SLC is far more durable than MLC, why MLC of far more durable that TLC and so on.

How is it not true? You're still using a part of a much smaller floating gate. An SLC floating gate is designed to only hold two states, whereas a TLC floating gate has to hold eight and QLC 16 states. The fact that you're only using two of them, doesn't mean the NAND cell itself is guaranteed to last longer or even retain data better, although with modern SSD controllers that reduces write amplification and has better ECC solutions than older controllers, it's often the case. I'm not sure what it has to do with "chemical states" as it's electric charge that's being used to store the data. I'm well aware that the NAND cells have to be zeroed out before new data can be written, but what does this have to do with using part of a TLC cell to store data? The performance issue is that when you have to store multiple charge levels in a NAND cell, it takes longer than when you store one, which is why pseudo-SLC is faster than TLC, as you're only storing one charge level instead of up to eight per cell. Yes, pseudo-SLC is touted as having much higher endurance, but it's still nowhere as good as SLC if you care about endurance.

From a purely technical perspective, forcing a NAND controller to use TLC NAND as MLC should be as trivial as using the same NAND as SLC in a caching mode.

Sure, it has already been done, as there are pseudo-MLC drives, but they cost a lot more and it's not a user controllable thing. Pseudo-MLC doesn't seem to have any performance benefit over TLC though, although the only numbers I could find are kind of old.

 

[chrcoluk]

Asus and PS5 tax on this one then.

 

[lexluthermiester]

I'm not sure what it has to do with "chemical states" as it's electric charge that's being used to store the data.

Chemistry has everything to do with how the data is stored. NAND cells, in many ways, act like a variable state batteries, which is why NAND cells wear out with use. Every time you change a NAND cell, you slightly alter it's chemistry. Such can only happen so many times before the chemistry in the cell can no longer maintain it's operational electrochemical characteristics.

I didn't chime in here to give a lesson on NAND chemistry and operational characteristics, only to share an idea based on something someone said.