Not sure if it’s BS, but new ferro process “IronLattice” promises millions x lower energy and million x faster than NAND. Claims to be a hybrid memory/compute. Strange stuff.
A YT is up under the COSM 2025 conference by a Dr. James Tour.
- Density: Current FeRAM cells occupy ≈ 2 F² (where F is the feature size), slightly larger than DRAM (≈ 1 F²) but improving with HfO₂ scaling.
Ya that is going to need some massive improvement to be viable for storage to replace nand. Remember we are using 8 memory chips just to reach 256GB on a single stick at best right now (with MASSIVE cost for that), meanwhile nand is at something like 1-2TB on a single chip of the same size for use on m.2 drives. So currently for this tech you would be looking at around 24-32GB per chip of equivalent size for drive use.
The base write time of XPoint/Optane is in the mid nanosecond range, and once it gets into an actual drive you end up with around 5-10 microsecond write latency. So this new ferroelectric memory being 10-50 nanoseconds we can expect likely around a half nanosecond in actual drive use. Which is very good, that is orders of magnitude better than Optane which is already crazy fast.
But as it seems like with every tech, the faster you go the less dense you are. So this is yet again trading off speed for density and ending up with something far, far less dense than even Optane was. Not a tradeoff that is going to win since we already saw the density and price killing off Optane compared to just using nand flash, and current RAM already offers the same write speed this does while achieving slightly better density right now. Its one advantage compared to RAM seems to be power draw. So at best this ferroelectric memory will likely be a RAM replacement rather than storage drive replacement, as long as it can maintain similar costs or better to normal RAM.
This is provided that ram prices have not gone to even more obscene pricing due to artificial scarcity. The ram producers have decided to not ramp up production to meet demand resulting in ever increasing prices.
Samsung wont even ramp up production to feed their own smartphone/tablet lines.
That is not really true, fabs are already running at capacity and what Samsung said is that it will not start construction on a new fab to increase capacity in reaction to the current demand, which is smart of them. Fabs cost billions and take years to build, which means the demand usually isn’t there anymore for the specific product type it was built in reaction to once the fab actually becomes operational. Samsung already has a new memory fab they started construction on years ago finally coming online in 2027, as does Micron. To start planning construction now on a new fab to build more memory would mean it wouldn’t be operational until around 2029-2030.
1 Like
This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.