Odd that it doesnāt specify max RAM capacity. For me more RAM and improved power/suspend would be the only way to improve over the 3588 (itās plenty fast for a mini laptop). You can never have enough RAM.
My MNT laptop isnāt the mini kind, and while itās adequate for my needs, itās lacking a bit of oomph to do certain things I do, like complex models in FreeCAD. Now, in fairness, FreeCAD is terrible at using CPU resources - let alone GPU - since it only uses one core, which is beyond frustrating. But if that single core was faster, Iād be a lot happier.
Generally-speaking, the RK3588 is kind of so-so for me. Itās more or less as quick as, or a bit slower than my old Ryzen 5 entry-level laptop of a few years ago - which isnāt fast by any measure. Itās usable, because Iām not demanding, but itās nothing to write home about and I could use a faster machine.
Good point, I always forget about the regular Reform (and the Next) because I tend to use only a mini tablet, the Pocket Reform, and an over-specced desktop workstation setup. Makes sense to want more oomph in a full size laptop, especially if itās your main system.
Ooh nice!
That was quick 
Where do I place my order?
I would love to have a RiscV. This said I believe what we see coming out this year is roughly on pair with the RK3588 in terms of raw CPU power, maybe more power efficient. Also, I mainly used riscv on embedded devices and Iām unsure yet how well supported is RiscV on Linux. But my assumption is that a lot of packages would not readily pre-compiled for the average user to install.
So I do not see it making sense for MNT to devote resources into pursuing a design based on those chips as it would overlap in terms of capability with the RK3588. The development of the Next and the move towards gnome tells me the company is pointing towards a more general public adoption. So spending resources into getting the RK3588 working flawlessly would be more mpactful (in my naive perspective).
Maybe Iām in a very specific situation that biases my perspective. I just got my RK3588 upgrade and from a performance perspective is all I need. If I were to buy a new cpu now and I had the choice I would have gone with a riscv if it were available and it had similar capabilities. Iāve got something that works for me atm, so even If I prefer a RiscV Iām not going to buy a new module just because its available. I rather avoid e-waste 
I may consider an upgrade if something came out that would replace my desktop though. But it would have to be a significant upgrade.
So would I. When itās a bit speedier though - which, like you, I hope is coming soon.
I threw it in here to see if it would stick
Iām not sure I follow. This seems to suggest that consumers (by way of conspicuous consumption and general irresponsibility) account for more e-waste than the oligopoly of vendors (planned obsolescence, lock-in, etc.) My point is lost if you take this as a reproach, so I hope you donāt.
Anyway, this only has a riscv coprocessor, apparently for SIMD operations. It seems like nearly every SBC is marketed āfor AI and Deep Learningā, but I donāt know how much learning youāll do with FP4. Half-precision was generally applied to conserve memory during inference. It also makes for impressive (entirely bogus) FLOPS/TOPS performance figures vis-a-vis outgoing models.
A proper riscv module for the reform would be awesome though.
I felt istar was expressing plain common sense: if you already have a perfectly adequate compute module there is no point getting a new one just for the sake of it.
Fair enough. Personally Iād buy one (but not this one) just to support riscv, Iād hate to see it fall by the wayside. Intel, AMD and ARM CPUs are not āopen sourceā by any reasonable interpretation, though I understand there are few alternatives at present.