Profile cover photo
Profile photo
R. Colin Johnson
About
R. Colin's posts

Post has attachment
By adding fields of gallium arsenide (GaAs) nanowires, each a PN-junction solar harvester, silicon panels can boost their output by 50 percent due to different part of spectrum to which GaAs is sensitive.

Post has attachment
Adding a third layer to the cortex -- a brain cap -- is deemed fanciful by critics of Elon Musk's latest venture, Neuralink. However, the company may surprise us by first tackling medical conditions like Alzheimer's which could turn it into a "medical" Tesla / Space X.

Post has attachment
Cray supercomputers can now be run in-the-cloud using Markely Group's million square foot facility set aside from the whole Cray line. Customers can also buy a Cray and host it at Markely's Boston datacenter, bypassing the need to build-out their own facility.

Post has attachment
Cray's specializes in neural networks of human brain in a new set of clustered supercomputers specifically for deep learning using thousands of Nvidia GPUs to crunch out all the recognition tests simultaneously.

Post has attachment
NIST claims another first in quantum computing, this time at the molecular level, moving up from its previously announce atom-level control. By building up to larger molecules NIST hopes to build quantum computer from the ground up.

Post has attachment
Everybody designing MEMS chips uses ConventorWare and MEMS+ for design and simulation, and with the new combined CoventorMP tool, customized for each fab, it ill be as easy to design a MEMS chip as using a fab PDK for a semiconductor chip.

Post has attachment
3-D graphene is called topological by the researchers, making it a modern alternative to superconductors and a candidate material to replace silicon at the end of its road. These researchers revealed some new properties of topological materials (also being investigated by Microsoft for quantum computers).

Post has attachment
Electronic brains are already here, but not in large-scale commercial usage, because the most popular model are software (like Watson) and thus very slow. A few big players have moved to hardware (such as Intel's Nirvana chip) but French researchers at the University of Arkansas believe they have a better way that uses ferroelectric artificial synapses to learn in realtime.

Post has attachment
Intel is redefining its Xeon family of datacenter and supercomputer processors as bronze, silver, gold and platinum (instead of the mis-mash of SKUs today) but its not all hype. The new SoCs will have on-chip interfaces to FPGAs, Nervana's deep-learning neural-networks, multi-processor Xeon Phi and silicon photonics. This preview announcement will be consummated by mid-year.

Post has attachment
The one-nanometer (1-nm) node is achievable according to Department of Energy (DoE) researchers at their Center for Functional Nanomaterials (CFN), a DoE Office of Science User Facility at Brookhaven National Laboratories. Their demo focuses an e-beam down to one nanometer using a e-beam from an aberration-corrected scanning tunneling electron microscope.
Wait while more posts are being loaded