Gordon Moore's ink on The Future of Integrated Electronics in 1965 was dry, and the prophets had already predicted the death of the so-called 'Moore's Law' Fifty years later, the law remains the dominant rule in the semiconductor industry ... to some extent.
Prediction that doubling the number of transistors on the chip every 18 months is still largely a firm belief that there have been some longstanding beliefs about how long industries can keep up with Moore's Law However, almost everyone agrees that it is becoming more and more difficult and costly to keep up with Moore's Law.
A recent report by Syed Alam and Greg Douglass at Accenture Strategy, a consultancy, explores the issue of maintaining Moore's Law and throws the question: Should we follow it blindly?
Alam, one of the report's authors and director of semiconductor business at Accenture Strategy, said in an interview with the EE Times: 'In the past, when you downsized your chips and advanced advanced process nodes, you could benefit from economies of scale. But now when we enter the next 10nm node The costs are very high.
'Our point is that if you go 45 nanometers from 45 nanometers, the cost increase is a little bit more expensive, but now you have to jump from 16 nanometers to 10 nanometers, the cost can be quite high;' he said: 'So now you have to Carefully assess your costs. '
It is also the long axis of Accenture's report, Moore's Law ... Moore ... or Less, that a model can be profitable to upgrade to a more advanced process node. As for when - ─ ─ and even if ─ ─ invest more advanced node chip design and manufacturing issues, today's chip makers need to think more long-term, more carefully.
The report notes that to go from 65nm to 10nm node, the cost of the chip design alone will increase by 1,300%; just follow the rules of the industry game and follow the footsteps of Moore's Law is the good old days to success, has gone The situation is more complicated and requires the chip makers to think carefully.
It is widely recognized that many components do not need to be upgraded to more advanced nodes. In fact, most chips do not cost much to justify the need to follow the process miniaturization rules. Currently, most chips on the market use 28 nm or The more advanced manufacturing process and the good operation will continue in the foreseeable future.
The reality today is that chipmakers need to think carefully about whether it is economically prudent to move their designs to the next generation of nodes, and as the Accenture report calls for chipmakers to look for older, yet mature and more cost-effective technologies Other opportunities.
However, the report also pointed out that not just a simple choice of a strategy for the future of technology transfer decisions should be based on a single case or product-based consideration, which allows chip makers old and new manufacturing process of investment between the wise to strike a balance; In other words, if you can successfully walk through the wire, chip makers will have a great chance of succeeding in growth and profitability in today's industry.
There are currently only a handful of well-known manufacturers such as Intel, Samsung, TSMC and Globalfoundries, which are capable of producing 10-nanometer processes; Alam said the reason these vendors are continuing their best efforts to sprint the most advanced nodes is that they After calculation, this is in line with economic benefits, and the key issue is that those second-tier chip makers and fabless IC manufacturers will do?
His advice is that before taking the next step: 'Maybe let the pioneers get the cost first, get the situation right, and make the technology stable again!'
Compile: Judith Cheng