After multiple tries, Cypress Semiconductor made a successful offer for Ramtron and the deal is currently in process. We talked with analyst Jim Handy of Objective Analysis to get his take on the dynamics of the deal and what it might mean for the memory market.
Kristin Lewotsky: Why do you think Cypress bought Ramtron? Jim Handy: Cypress started out as an SRAM manufacturer. SRAM is very expensive type of memory to build—it uses six transistors per cell whereas most other memory technologies use one or two, so what do you end up getting for that expense? You get high speed, very low power, but it’s still volatile memory.
Cypress appears to be looking for alternative technology that can perform like SRAM but have a lower cost structure and also offer non-volatility. That’s why they acquired Simtek [now AgigA Tech]. The Ramtron deal indicates to me that Cypress has a grand plan to be an important player in a market that is just developing for high-speed nonvolatile memory. Ramtron also has some really blue ribbon licensees like Texas Instruments and IBM.
K.L.: How will the deal affect the overall memory space? J.H.: The situation is ripe for some changes. You’ve got Everspin, which is actually making inroads in the same market that Cypress is going to be going after with Ramtron and Simtek. Today, MRAM and the Ramtron products cost more because they’re shipping in lower volumes; over the long run, they should cost less than SRAM.
K.L.: What’s the biggest challenge likely to be as far as integrating Ramtron into the larger company? J.H.: They’re going to have to find a way to make their processes work together. Ramtron’s process is a little bit difficult because it uses elements that standard CMOS or standard silicon doesn’t like very much. There’s lead in the Ramtron chemistry, so they have to put platinum barrier layers over the silicon before they put the lead on to prevent any lead from getting into the silicon.
K.L.: Can you give us an idea of the overall market size of the market for alternative RAM technology? J.H.: Very, very small—probably in the neighborhood of $100 million
K.L.: What do you think about the future of FRAM? J.H.: When I moved to Dataquest back in 1991, I found somebody’s old forecast. They had listened to the story and decided that the entire memory business would go over to ferroelectric memory by the early 1990s, and it just never happened. It ended up that it was always a more expensive technology to build than conventional memory technologies.
It seems that new memory technologies have been waiting at the altar since the 1980s. The incumbents continue to scale way past where anybody thought they would and so they continue to be the most cost-effective memory. It’s anybody’s guess when the real opportunity is going to arise for alternative memories and mainstream applications. If they don’t go into mainstream applications, then the economies of scale are going to prevent them from ever becoming cheap, although there will always be more expensive memories that have technical features that a rare few are willing to pay a lot more for.
I would definitely like to see memory technology like FRAM become more popular and cheaper. The benefits such as speed and EEPROM-like behavior such as byte-level modification but with orders of magnitude better endurance could greatly simplify applications that normally require wear-leveling and page-erase handling algorithms. Let's hope Cypress doesn't do anything to mess things up.
Discrete SRAM market is a rapidly shrinking, even Samsung & NEC left the market, only GSI and Cypress are left. Cypress as company has issues, Ramtron is not a cure for this. Problems at company are large in scale around PSUC, PSUC is not a competitive product against fixed function MCU.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.