Hobbysts and students have done single-core RISC-V that boots linux.. 400MHz on 22nm FFL is underwhelming. The real challenge is in scaling it up to multiple cores, with hierarchical caches and fabric interconnect. And all within power-budget. Not that IITM cannot do it. This is just 5% milestone.
Instead of trying to compete head-on in the highly competitive consumer microprocessor market, Indian chipmakers should try to go after the market for radiation-tolerant microprocessors for use in outer space. Indian chip designers nearly always have to outsource the actual fabrication to foreign chipmakers anyway - so they could likewise use chipmakers specializing in advanced space-age materials like Gallium Arsenide, which are more radiation tolerant than silicon due to their higher bandgap. India is trying to gain a greater share of the satellite manufacturing market anyway, to build off its strengths in the satellite launch market. I think this might be a good way to go.
Furthermore, Russia also has a lot of expertise in Diamond semiconductors, and diamond too is a robust high-performance material with good radiation tolerance, so maybe India should try to acquire technology from them.
Note that IISc's Centre for NanoScience & Engineering (CENSE) has been developing technology in Gallium Nitride, which is a very high-end bandgap material, used in cutting-edge military electronics like X-band radar.
Spintronics is supposed to be a good candidate for radiation-tolerant microprocessing. MRAM provides resistance/resilience against radiation because magnetic states are based on local spin majority, which is harder for the random stray cosmic ray/particle to overturn.
Today's MRAM is not large enough for storage (to replace SSD). So, it's trying to be a replacement for embedded SRAM.. Spintronics is thriving in the sensor market, though.
It's probably because nobody's put enough effort into developing MRAM. I'd read that nanotube memory is a contender to replace Flash memory, with a company called Nantero leading that pack. Another good candidate may be Memristor technology, which HP seems to own, because its hysteresis properties resemble that of neurons, thus making it potentially useful for AI as well.
7 comments:
Hobbysts and students have done single-core RISC-V that boots linux.. 400MHz on 22nm FFL is underwhelming. The real challenge is in scaling it up to multiple cores, with hierarchical caches and fabric interconnect. And all within power-budget. Not that IITM cannot do it. This is just 5% milestone.
Instead of trying to compete head-on in the highly competitive consumer microprocessor market, Indian chipmakers should try to go after the market for radiation-tolerant microprocessors for use in outer space. Indian chip designers nearly always have to outsource the actual fabrication to foreign chipmakers anyway - so they could likewise use chipmakers specializing in advanced space-age materials like Gallium Arsenide, which are more radiation tolerant than silicon due to their higher bandgap. India is trying to gain a greater share of the satellite manufacturing market anyway, to build off its strengths in the satellite launch market. I think this might be a good way to go.
Furthermore, Russia also has a lot of expertise in Diamond semiconductors, and diamond too is a robust high-performance material with good radiation tolerance, so maybe India should try to acquire technology from them.
https://www.wired.com/insights/2015/01/the-rise-of-diamond-technology/
https://spectrum.ieee.org/nanoclast/semiconductors/materials/diamondbased-semiconductors-take-a-step-foward
Perhaps India should try some of these angles, in order to find a strong niche as a beach-head into the very competitive microprocessor industry.
Note that IISc's Centre for NanoScience & Engineering (CENSE) has been developing technology in Gallium Nitride, which is a very high-end bandgap material, used in cutting-edge military electronics like X-band radar.
http://www.cense.iisc.ac.in/research/cvd-reactor-galium-nitride-gan
Why couldn't Indian govt try to spin off such technology to private sector partners, to let them establish a lower-cost foothold in the marketplace?
Spintronics is supposed to be a good candidate for radiation-tolerant microprocessing. MRAM provides resistance/resilience against radiation because magnetic states are based on local spin majority, which is harder for the random stray cosmic ray/particle to overturn.
https://www.titech.ac.jp/english/news/2018/042001.html
Today's MRAM is not large enough for storage (to replace SSD). So, it's trying to be a replacement for embedded SRAM.. Spintronics is thriving in the sensor market, though.
It's probably because nobody's put enough effort into developing MRAM. I'd read that nanotube memory is a contender to replace Flash memory, with a company called Nantero leading that pack. Another good candidate may be Memristor technology, which HP seems to own, because its hysteresis properties resemble that of neurons, thus making it potentially useful for AI as well.
Post a Comment