This is one field where it&#x27;s extremely hard to keep up with research internal to the manufacturers - or even fully understand what they currently produce. Hard to be relevant, basically. So well thought out, published, detailed reverse engineering is super useful - and at that scale, it&#x27;s a research discipline of its own.Even then, they seem to be only looking at space overhead (the dimension of the accessory circuits - besides the memory cell).

I didn&#x27;t have the same feeling when I read the summary. It looks like they uncovered an important disconnect between academic research and real world.From my experience, some academic fields are more prone to make mistakes because they don&#x27;t get to test their ideas in a real environment. IC research is a prime example for this. You can&#x27;t tell a class or a Ph.D. student to tape out a design to TSMC (or an imaginary local fab) to test their design. You can simulate up to a point, but the design never be built in millions and tested in billion scenarios.So, simply saying that &quot;we have the wrong assumptions about how these things are done&quot; is a great correction signal which must be fired more frequently, to bring academia and industry to the same line, and maybe reduce the negative views they hold for each other....or I might be totally wrong, of course.

I am not an expert in the DRAM research, but this paper looks like a checkbox on the list to get a research grant and deliver something at the end.I just saw a video yesterday where Sabine Hossenfelder was telling that the number of researchers is increasing significantly (per capita, not just in absolute, numbers), but the output is decreasing. Even if millions of papers are published every year, the scientific value of the combined output is decreasing, so the value of each paper is lower and lower.

HiFi-DRAM: Enabling High-Fidelity DRAM Research by Uncovering Sense Amplifiers