Intel hits out against IBM terabit chip claims

Intel hits out against IBM terabit chip claims

Summary: IBM's just-announced manufacturing process for making chips with terabit transfer speeds is not the only one around, nor is it the most advanced, Intel has argued.Both Intel and IBM are in the process of developing technology for making chips that can use pulses of light, rather than electricity, to transfer data.

TOPICS: Storage

IBM's just-announced manufacturing process for making chips with terabit transfer speeds is not the only one around, nor is it the most advanced, Intel has argued.

Both Intel and IBM are in the process of developing technology for making chips that can use pulses of light, rather than electricity, to transfer data. But the two companies have different manufacturing techniques in place.

On Wednesday, IBM announced its CMOS Integrated Silicon Nanophotonics technology. The company said its method of manufacture, which places all the technology necessary for photonic transfer onto the same CMOS chip, is more effective than Intel's.

Intel's technology keeps the photonics and the chip separate, so it does not have the benefit of the economies of scale that IBM's single-chip manufacturing process does.

On Thursday, Intel told ZDNet UK that it believes its own light-based data transfer technology is as effective as IBM's. Its own technology has an on-chip laser, unlike IBM's.

"This [IBM] research is another example of others also validating that silicon photonics is the path to high bandwidth, low-cost optical communications. While this research is interesting, there are still many challenges to commercialise this approach such as integration of lasers and integration with advanced future transistor processes," Intel's global communications manager Nick Knupffer told ZDNet UK on Thursday.

Knupffer pointed out that Intel's own light-based silicon photonic technology — demonstrated at the Intel Developer Forum this year — is competitive with IBM's technology for reasons of energy efficiency.

"Keeping the CMOS and photonics separate will allow us to use the most energy-efficient, leading edge manufacturing process for electronics as we scale link speeds. For exascale systems, energy efficiency is a major concern," Knupffer said.

He defended Intel's manufacturing process and said that "although it does require the bonding of indium phosphide to the silicon, this is done at the chip or wafer level, allowing us to create all the lasers we need in one bonding step."

Because Intel's in-development technology has an inbuilt light source, Knupffer feels that it presents advantages when compared with IBM's technique. Knupffer pointed out that unless IBM develops a technique for putting laser sources onto its chips it will have to go through a process of "buying, assembling and fibre-coupling a multitude of laser sources to silicon photonics devices that do not incorporate hybrid silicon lasers."

IBM's technology has been in development for 10 years, Intel's for six. Both companies are hoping to produce chips using the photonics technology that are capable of a terabit of transfer speeds.

In March, Intel demonstrated a chip that could transmit data at the rate of 200Gbps on eight channels of 25Gbps each.

Topic: Storage

Jack Clark

About Jack Clark

Currently a reporter for ZDNet UK, I previously worked as a technology researcher and reporter for a London-based news agency.

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • If the optics are not a size that can made commercial in few years, what is the point? From what I read, the optics from IBM are magnitudes smaller than intel architecture. It is clear, intel's motive is to move silicon and IBM is petabits so I expect IBM to win this race since intel is hard wired in to the wrong design. www.

    At some point, the rubber will hit the road. Intel is after all, the company that owes most of its leadership to Marketing and not technology. For those of us that can recall, NEC of Japan in the early 1980s was the world's largest IC company and intel sued them for stealing the "micro code" inside of their NMOS 8080 and 8086 MCUs. At that time, intel was less than $500 Million company I believe (may be wrong here). intel's brilliant lawyer or marketing executives and also sales people sold the world that if they won the lawsuit, they would demand end customers return their systems to whoever they bought them from (read: IBM) and have them retrofitted. Natually, IBM elected not to populate these IBM PCs with NEC MCUs and instead intel had 2-3 year open, free playing field as intel spun, appealed and eventually LOST the LAWSUIT to NEC but WON THE WAR!

    Intel, a company embracing "constructive confrontational behavior" (read: lot of brow beating and chest pounding) won the MCU not with superior technnology, but with superior sales, marketing, legal guidance and the facts remain this forever:

    Intel 8080 and 8086 were micro code
    NEC V30 and V40 had hard wired logic = faster

    Intel MCUs: NMOS
    NEC V series: CMOS = better

    Intel MCUs: > $10
    NEC MCUs: < $4 = better

    NEC 100% plug compatible with Intel footprint = drop in replacement and VERY GOOD!

    Intel is pulling the same nonsense with their optical platform and hopefully they will end up losing. intel is a company that is predatore to sell to, rips your arm off to buy from (hence AMD's success) but has slowly fogged the world with this fear of failure that IBM invented early in computing days.

    In the photonics race for silicon photonics, IBM's going to win because they were not afraid to invest in a new glass fiber interconnect technology that enables much smaller geometries the world can use and it is not just about speed it is about speed per mm and i would love to see the comparison made.
  • I was working for Sinclair and Amstrad when we were using lots of Z80 and 8086 chips, most certainly including NEC's designs. I don't recall the lawsuit coming up as a factor - but perhaps I just wasn't party to those discussions.

    As for the photonics - well, Intel says that its physical interconnect is going to be very cheap and easy to use. Whether IBM's will end up better, I don't know, but I'm glad that both are in the race. The best thing about competition is that it keeps people honest - and the worst thing about no-holds barred competition is that it uses legals to hold the other lot down regardless of merit.