Processing .....please wait.. 
The article has been forwarded.... 
Indian Genius Makes World's First 100 Core chip
Bangalore: You might have seen chips with eight, sixteen and even 32 cores. But Anant Agarwal, an MIT genius is manufacturing the world’s first chip with 100 cores in his commercial venture Tilera.
Anant is currently the Director of Massachusetts Institute of Technology’s Computer Science and Artificial Intelligence Lab (CSAIL). Tilera, the many-core chip is a product of his focus on making efficient future servers with power saving and high performance chips.
The new chip is made by assembling multiple cores into a single one. Currently, the high end chips from manufacturers have a maximum of 16 cores whereas Tilera offers chips with 100 cores which are low in power consumption and high in performance and memory support. The traditional “connecting bus” is replaced by a “switch” in new processor. According to Anant, “every processor has a switch and they all talk to each other like in a peer to peer network.”
Tilera’s performance claims were validated by Facebook engineers in Memcached, a high-performance database memory system for web applications. They compared Tilera’s second generation 64-core processor to Intel’s Xeon and AMD’s Opteron. According to the engineers, a tuned version of Memcached on the 64-core Tilera yielded at least 67 percent higher throughputs than low-power x86 servers.
Currently Tilera manufactures chips with 16, 32 to 64 cores and it plans on shipping the 100 core chips later this year. Anant targets network and videoconference equipment makers with Tilera.
Tilera’s future plans code named “Stratton” focus on improving memory, interface, I/O and instruction set. It will produce chips with cores ranging from 4 to 200 by 2013. Anant also leads a new MIT project code-named Angstrom, aimed at building exascale supercomputers.
Post your Comment
All form fields are required.
In India a scientist and their innovation have to fight to survive by their Higher up Managers at all levels. Forget the Govt but also in the private sector. It is cultural like it or not. That has to change.
Why is that a many "Indian Americans" in the U.S. have shown such creativity and innovativeness while working for American institutions? Why is it that we do not see the same creativity and innovativeness from Indians based in India, such as those from the famous IITs and IISC? What makes this difference?
Let us do some emotion-free fact finding and discuss here. Thank you.
Being in an English language environment, or even being ruled by the English is a wonderful situation, the like of which is not there in any Asian or African nation.
I experience this in my personal life day-in-day out. I am in an industry where all the headache of setting up a business is pure madness in India, but my experience in NZ was simple and neat. My creative ability was at an all time high in NZ, but in India, my creativity has been consumed by bullheaded bureaucrats and babus and eventually i could not even give 25% of my creativity here. Screw them, i feel now like i should have appointed someone to handle these leeches in positions that dont do talking without bribery.
The obvious ones that come to my mind are,
1. Lack of a semiconductor fab, even an older technology node capable one, NOT necessarily the state of the art like 21nm. This is an essential piece to try out designs, characterize and calibrate chip designs.
2. Lack of backend, physical design and packaging expertise within the Indian semiconductor ecosystem. More than 90% have had expertise in chip designs but have never been to a lab, wouldn't know how to use any characterizing equipment. Just programming in VHDL does not make you an chip design expert -one needs to develop an understanding and appreciation of the end-to-end process in chip design.
3. As stated in 2 above, many-core processors are a challenge in physical design, packaging and interconnects, power management and high speed on-chip communication and management using a hardware-level hypervisor. These require analog-mixed signal expertise in addition to digital design. I am not convinced that we adequate professional expertise let alone academic exposure in this area.
These are some of my thoughts for now, I welcome comments.
 |
© 2012 SiliconIndia all rights reserved