The walk towards innovation
Date: Thursday , August 31, 2006
The AMD India Engineering Center (IEC) commenced operations in 2004, six years after its business rival Intel did. Despite a late entrant into the Indian market, the IEC center has aggressively ramped up its development activities.
Designing next generation microprocessors—chips that drive computers and high-end networks for Internet-based applications—would require AMD to attract engineers who have ample years of experience in the field. Attracting talent from competition wasn’t easy, especially in the beginning. AMD relied on the engineering talent that came back from the U.S [returnees] and fresh graduates from the elite schools.
Raghuram Tupuri, General Manager, Micro-processor Design Engineering Center, AMD India, often visits IITs or RECs. Either he is seen picking up the brightest of the talent or talking about new technologies in microprocessor design to the young minds.
The new recruits at AMD are put through the Compressed Design Cycle, a program that exposes them to the entire microprocessor design cycle in fast motion. It covers various aspects of chip making like meeting the electrical specifications, architectural speed et al, and concludes with a feed back review session where the problems are put under post mortem and discussed. This helps the new engineer envisage the end result before he even begins. “The foresight of knowing your final product and the obstacles likely to appear will eventually help you design more effectively,” asserts Tupuri.
The program is indeed quick and effective in helping these new designers get a feel of the entire process. This can be gauged by looking at those who have just been through the exercise. Almost all are excited and apply their research skills. Many a times they figure out the root of the problem before the design gets to its end stage! It is this enthusiasm among the young breed that strengthens the confidence in Tupuri, that things could be designed out of India.
Unlike ASICs, microprocessors can be designed without keeping only functionality in mind. The complexity arises from the various timing, process constraints and features. The technology is being tested in every direction and there is no clear picture presented. Everything is designed at the same time - the chip, the application and the process technology. To put in simpler terms, “You are trying to get 110 percent of a technology that is not even there,” says Tupuri.
With new technologies emerging by the day, the product design is susceptible to change from the original plan. The trick is to be prepared for these surprises - an art that comes through experience. With India lacking in enough experienced seniors in the field of microprocessor designing, engineers worked with AMD’s global teams and jointly worked on projects. “The team acted as a global center and problems were treated as an India centric problem or an Austin centric problem, but a problem that had to be addressed,” says Tupuri. Within a short span of 12 months, the India Center has gained considerable expertise and exposure to every element of the microprocessor design cycle.
Today, nearly 100 engineers at the ICE are working on the 65 and 45-nanometer technology. Starting with performance modeling, architectural verification, implementation of logical designs and GDS the team is making its fair share of contribution towards designing a chip.
While the team is busy developing next generation server technologies, it has already contributed to the design of its latest native Quad core design project that is perched on the anvil. The Bangalore team will be working on the performance implementation of the Quad core design with its expandable-shared L3 cache. Tupuri is unwilling to disclose other features of this much-awaited processor. However, he quickly assures that it is ‘definitely beyond and above known boundaries.’
Designing microprocessors is always going to have a fresh set of problems and this is what excites his team the most. “A problem that arises in architecture could have a simple solution in transistor level design and without the wholesome knowledge you could be wasting a lot of time and decreasing efficiency,” emphasizes Tupuri.
To avoid the fixed mindset and encourage innovation rather than force it, AMD IEC, practices the brainstorming session. Here all the ideas are out on the table, dissected, making sure that everyone is aware of the ongoing hitch in the chip design and the floor is then laid open to suggestions and ideas. This creates an open environment of interaction where techies can work together. Here good practices of old designs can be implemented while the bad practices are avoided.
With emerging innovations in technologies and the transistors getting smaller in size, solutions today can be found in places least expected. And the best way to achieve this was running through the existing data and re-applying every possible method on new designs irrespective of its previous outcomes on other designs. “The techniques that did not work in the past cannot be thrown out without getting re-evaluated in the new technology. And this signifies the relevance of being updated with technology’s ever evolving nature,” says Tupuri.
The TIPS (Technical Interchange Presentation Series) is one such programme that helps the techies update themselves on the latest developments. Orchestrated completely by internal folks, the programme allows them to present topics of interest in the tech field. This not only helps innovation, but also subtly increases productivity and performance.
The AMD IEC also has visiting faculty to share their latest research papers and technological updates. An active collaboration with UT (University of Texas) Austin permits the India center to have distinguished faculty visit their team, whenever they stop by India. Promoting the same concept, AMD IEC is in talks with various elite schools in India that would help have a collaboration with Indian faculty as well as help engineers at the Bangalore center pursue their higher studies. “We want our engineers to grow and have the option to pursue their higher studies while still working at AMD,” says Tupuri.
The fact is AMD faces relentless pressure from Intel in the race for more speed and cheaper designs. For every dollar AMD spent on research and development last year, Intel spent $4.52.
As Hector Ruiz, AMD’s President, Chairman and Chief Executive Officer puts it, “We believe the purpose of our company is to reinvent the dynamics of the microprocessor industry.” A company where talent enjoys precedence over cost, somehow.