Slymeat
Move on, nothing to see.
I received a research paper for Weebit Nano that mentions Brainchip and states that ReRAM may be a good fit for neuromorphic SOC implementations.
Is that seed I sewed finally picking up some traction? How wonderful would it be to see (sing the following to Stevie Wonder and Paul McCartney's Ebony and Ivory—thanks for pointing out the correction @JDelekto )
“Weebit ReRAM and Brainchip Akida IP,
living together in perfect harmony,
providing essential functionality on the same SOC”.
The report is all about Weebit Nano but contains specific reference to Brainchip and the following tidbit of how ReRAM could be useful.
In the report, specific mention of Brainchip is made on page 9. It also has a link to prior research they have done on Brainchip. Nothing new for us, but is excellent to see the flag being waved in papers dedicated to other companies.
Is that seed I sewed finally picking up some traction? How wonderful would it be to see (sing the following to Stevie Wonder and Paul McCartney's Ebony and Ivory—thanks for pointing out the correction @JDelekto )
“Weebit ReRAM and Brainchip Akida IP,
living together in perfect harmony,
providing essential functionality on the same SOC”.
The report is all about Weebit Nano but contains specific reference to Brainchip and the following tidbit of how ReRAM could be useful.
Weebit Nano has high potential in Neuromorphic Computing
Since conventional computing systems have separate processors and memory units, conventional chip architecture poses substantial challenges when it comes to hardware implementation and deployment of neuromorphic computing. To try and deal with that, several deployment techniques have been proposed, which entail cutting down on memory requirements to fit the hardware, which is actually undesirable. This has brought ReRAM into the picture.
ReRAM could offer the following advantages:
1) smaller footprint and non-volatility (compared to SRAM),
2) lower voltages and scaling below 28nm (compared to Flash),
They didn‘t even mention the lower power consumption, cheaper to manufacture, temperature resilience, radiation resilience—basically none of the cheaper, faster, stronger, lasts longer. Able to leap tall buildings in a single leap. (I made that last bit up)3) lower cost, lower area, multi-level cell (compared to MRAM).
In the report, specific mention of Brainchip is made on page 9. It also has a link to prior research they have done on Brainchip. Nothing new for us, but is excellent to see the flag being waved in papers dedicated to other companies.
Attachments
Last edited: