MRAM (Magnetoresistive Random Access Memory) replacement for NAND?

[JTsyo]

Looks like it has a ways to go but looks fast. For computers to continue to get faster in the next decade we'll need new technologies like this. There's only so far smaller die sizes can take us.

 

[Cerb]

The density difference is too great, really. However, there can be good uses for an in-between storage medium: faster than flash, no write limits, and non-volatile.

Possible scenario: an MRAM write cache, organized as a transaction log, could allow for reliability improvements in SSDs, without the need for big caps. Detect that voltage is dropping (assuming the SSD can run at lower-than-spec voltages for a small fraction of a second), finish only writes physically in progress, stop everything, and then figure it all out on the next bootup. It could require a few slow caps, but it would be the difference between $0.10-0.50 and $2-5, which would be enough savings for all but the vertical market SLED server folks to make do with not flushing the whole queue.

 

[Hellhammer]

Currently the most likely "replacement" for NAND is 3D-NAND. That is more like a new iteration of NAND as it shares similar characteristics with NAND but eliminates some of the key issues found in today's planar NAND.

 

[yefi]

If this scales down to smaller processes as good as they say it does, it should become a viable option for enthusiasts a few years from now. It'll still be good as non-volatile cache in the meantime, of course.