I believe the DRAM industry should have a foundry side where 3rd party innovation can use DRAM process. Onur's talk illustrates the diversity of ideas out there - but it also illustrates the ridiculous barriers innovators face in trying to test those ideas. Probing OTS DRAMs with strange commands should not be the only tools available to them. The DRAM vendors can do much better! We can hope that the current market shifts start to bring a more forward looking attitude and a new willingness to empower outside ideas.
What do you do with the excess outputs on a reversible gate? I would assume that for the goal of breaking the Landauer limit you cannot simply dump the auxiliary outputs of, say, a NAND gate. That would seem to add to the complexity of wiring and more deeply to the problem of how we start with a lot of data and end up with a much smaller refined, useful data set, which seems inherently not reversible.
Fortunately the Landauer limit is far below where we are today, at least 10,000x better than the best low power CMOS today by the numbers I am familiar with. We should be able to get about a 10x improvement with CMOS at liquid nitrogen temperatures and matching materials for those temperatures and threshold voltages.
https://www.youtube.com/live/91smpJIFt-4?si=PcA8zzaXTgbJiZ5Y
EFCL Summer School - Keynote #4: Memory-Centric Computing
That is a beautiful talk, thanks for sharing!
I believe the DRAM industry should have a foundry side where 3rd party innovation can use DRAM process. Onur's talk illustrates the diversity of ideas out there - but it also illustrates the ridiculous barriers innovators face in trying to test those ideas. Probing OTS DRAMs with strange commands should not be the only tools available to them. The DRAM vendors can do much better! We can hope that the current market shifts start to bring a more forward looking attitude and a new willingness to empower outside ideas.
HotGauge: Methodology for Characterizing Advanced Hotspots in Modern and Next Generation Processors
https://youtu.be/61VJ7KJAgnM?si=XI0cxHg48lxqFInl
Oscillator for adiabatic computational circuitry (US11671054)
Granted Patent | Granted on: 2023-06-06
https://labpartnering.org/patents/US11671054
ITU-T L.1318 : Q factor: A fundamental metric expressing integrated circuit energy efficiency ; Recommendation L.1318. In force components.
https://www.itu.int/rec/T-REC-L.1318/en
https://patents.google.com/patent/US11671054B2/en?oq=Oscillator+for+adiabatic+computational+circuitry+(US11671054)+Granted+Patent+|+Granted+on:+2023-06-06
SUSTAINABLE COMPUTING
Ubiquity, Volume 2021 Issue February, February 2021 | BY Art Scott, TED G. LEWIS
https://ubiquity.acm.org/article.cfm?id=3450612
Very interesting!
What do you do with the excess outputs on a reversible gate? I would assume that for the goal of breaking the Landauer limit you cannot simply dump the auxiliary outputs of, say, a NAND gate. That would seem to add to the complexity of wiring and more deeply to the problem of how we start with a lot of data and end up with a much smaller refined, useful data set, which seems inherently not reversible.
Fortunately the Landauer limit is far below where we are today, at least 10,000x better than the best low power CMOS today by the numbers I am familiar with. We should be able to get about a 10x improvement with CMOS at liquid nitrogen temperatures and matching materials for those temperatures and threshold voltages.