Not physics, but economics limits Moore's Law'
‘Pentium is dead’ screamed the headline. It had a rather ominous ring to it. For over ten years, Pentium (standing for fifth, penta = five) and ‘Intel Inside’ were symbolic of computational power or rather man’s technological prowess.
The name Pentium stood for trust, people were ready to pay hideous sums of money just for the latest version. And with the launch of every newer version, the previous one met a vain end. Thus, Pentium II killed I; III killed II; and IV killed III. So, it was but natural, to expect Intel to launch a Pentium V, after all it had been over five years since the launch of Pentium IV.
But, that was not to be. Paul Ottelini, CEO, Intel, deemed otherwise. The company last month launched Intel Dual Core brand of processors, signaling the end of the trusted Pentium.
Circa 1975, Vinod Dham, a graduate from Delhi College of Engineering, arrived in Cincinnati on a scholarship, he had a few dollars in pockets but was high on ambitions. After completing his MS, Dham joined NCR and shortly thereafter he was at Intel.
Beginning at the lower rungs, Dham quickly scaled up the ladder and in January 1990 was made in charge of the 586 (later renamed as Pentium) development program. It was due to this, he earned the sobriquet, Father of the Pentium.
Yet, there is more to Dham than Pentium. At the height of his fame in 1995, he quit Intel and joined a startup named Nexgen, which was acquired by AMD. He was briefly at AMD, before the entrepreneurial bug hit again and he joined another startup Silicon Spice. Broadcom acquired the company and Dham moved on.
Currently, he is donning the cap of a venture capitalist; he is the cofounder of New Path Ventures. He has invested in a host of new ventures. In an exhaustive interaction with Shashwat Chaturvedi from CyberMedia News, Dham spoke at length about his views on the latest trends in the semiconductor industry, the roadmap for India and of course, on the death of Pentium. Excerpts:
Your take on the story, Pentium is dead.
I do not know in which context this was said, for 15 years we were trying to improve on the performance of ever-faster processor, if the story spoke about the end of that idea, then they are right about that.
From 8086 to Pentium IV, Intel was always striving for more and more processing power. In the eighties and through the nineties, due to the limitation of the processors, things like surfing on the Internet, or working on the spreadsheet were quite an onerous activity. The only solution was a processor that could do all this and more at a faster speed. But in the last few years, the balance had been achieved. The software that ran the CPU and the processor were finally evenly matched. Thus the imbalance that was fuelling the race for ever-faster processor is over. Pentium was a representative of that idea, that notion. In a way, you could say that Pentium is no more.
Your views on the latest Intel Dual Core processor. Have they got it right with two cores against one?
The idea is not necessarily two against one, the objective is to provide higher performance but with lower power consumption. This requirement has its roots in the laptops, where battery life is a major concern. But now mobile machines are driving this initiative.
Take the case of a cell phone. Normally, most of us charge the battery in the night before and then the whole of next day we do not bother about it. No need to tag along extra battery or charger, and things like that. That’s a good sign of mobility. Even desktops need to adhere to this criterion. With the global energy crisis, power consumption is a big issue and electricity is a part of it. Currently, desktop machines consume too much power. Especially, when you put up a data center like a server farm, the amount of electricity consumed is mind-boggling. We could not afford to continue in this vein.
One of the solution was to keep the performance the same by using multiple cores, each one cooler than the big heavy core. By stringing these cores together, one can get higher performance, at lower power. Dual Core is the first step in that direction. In future, you would multiple cores and more of them. It is the beginning, probably the best solution that one could have.
Does that mean that the focus shifts to power and performance takes a backseat?
Indeed, the focus has shifted to power and as I said earlier mobility is the key driver. Power is an important issue for laptops and more so for the server, for things like data center, grid computing. The next decade, in semiconductor terms, belongs to power, the last two were dedicated to performance.
What about the famed rivalry between AMD and Intel? Who has got it right?
It is not a question about who has got it right, or who has it wrong. AMD has a big leg up on Intel, both in terms of introduction of solution in the market, as well as creating a multi-core product ahead of Intel. Finally, Intel seems to be catching up. But I feel somehow, Intel’s solution isn’t the most elegant one today. Though I am sure, with time they will modify it, refine it and get it right.
Is the semiconductor industry becoming too consumer centric?
The semi-conductor market has evolved over the last four decades or so and has run through its course. Were we to plot a graph, we would be at what one traditionally refers to as the bit curve. First there is the innovation, followed by adoption and then saturation. I think we are at the top of the bit curve, a saturated market, where the cost of semiconductor is very cheap and the performance required is no longer an issue.
Semiconductor has become an integral part of our everyday existence; it is present in cameras, cell phones, DVD players, and others. Every aspect of home is getting the full benefit of this semiconductor revolution and the main beneficiaries are the consumers. Companies have become conscious of this demand, and the industry is re-structuring itself right now.
In the past, you had spoken quite vociferously about the telecom processor, what is the latest on that?
Back in the nineties, I had realized that the race for higher and higher performance would come to an end. Connectivity would be the key in the future. The idea was to stay connected on high-speed bandwidth. But, with Internet becoming more pervasive, there was a slowing in the flow of information. In that context, I had coined the word telecom processor to put the discussion, that we need a chip that allows us to have that connectivity, in center stage. At Silicon Spice, we had created a prototype of the telecom processor that enabled Voice Over Internet Protocol (VOIP) and even allowed multiple VOIP conversations. The idea has now taken off in a big way, globally.
What about the race to Moores Law (according to which the transistor density of integrated circuits doubles every 18 months)?
I think that Moores Law is reaching a point where it is a getting more limited by economics rather than by physics. Doubling the density is getting very complicated by the day and it will take lot more effort and much more expense to reach the same point, every time. Thus very few companies would be able to reach the same point (doubling the capacity). The primary reason for reaching that point was to get higher density, which translated in lower cost. This was the driving force for most of the microprocessor industry to for last 20 years. But now affordability will be a big constraint, thus doubling the density will not be the main objective anymore.
There has been a lot of debate on the road India should take, should we go for manufacturing or design?
India is a great destination for chip design. Silicon Spice, a company I funded back in 2002, is a testimonial to that. I think there is absolutely no reason why Indian engineers who have been doing so well in software, should not move into the new space. India could easily extend its software expertise to chip design.
But as far as manufacturing is concerned, one has to be very careful. There has to be in-depth analysis on the course of action to be taken. For instance, what is the real competitive cost that India can offer over the Chinese manufacturers? How can Indian players compete in against these well-entrenched companies? Before we invest billions of dollars in the country, one has to make a partnership with potential customers so facilities are not idle.
It is like buying a Jumbo 707 and not getting the permission to fly that plane, it does no body any good. One might take the high ground of owning a big plane but you will be losing millions everyday. Hence, I caution everyone against having a fab in India, one has to careful about how to go about such an enterprise.
But then India is losing out to even smaller countries, for instance Intel chose Vietnam over us?
My gut tells me that Intel would have very much liked to do this plant in India. It was in reality an assembling and packaging plant, not exactly a fab. But it was the perfect way to start in India. India should have been very firm on the commitment. But for some reasons the Government of India was not willing to offer the same concessions that Vietnam or prior to that the Chinese and Malaysian governments have offered to Intel. At the end, it is all about business and Intel chose the location that gave them the highest returns. According to me, it was India’s loss.
Indian government should have been more accommodative to Intel because not only would this plant have got more jobs but also the technology. Sadly, the Government of India was not willing to go to the extent where other governments are willing to go, in order to bring these business companies in to the country
What is the latest on you avatar as a venture capitalist?
At the start of my VC days, I had focused on hardware companies, builders of systems, semi-conductors and embedded software, etc. I had invested in companies like Nevis Networks, and others.
Going forward, I am doing a new fund for Indo-US ventures, for which we will be addressing the market in terms of projects in the mobility spaced and things surrounding service infrastructure, healthcare and other areas
A word on Indian innovation and Indians, at large.
Indians are doing an outstanding job, across the board. I am proud of people like Ram Krishnamurthy (at Intel) and others for the wonderful work they have done.
I think it is going to become more prominent then it has been because in 1975, when I came here there were very few Indians who were given the opportunity to lead and work on these kind of programs at big and high profile companies but now a host of Indians across that are doing more.