This year's MemCon is on October 11, at the Santa Clara Convention Center. Last year I wrote about how MemCon got started. But since it was just my fourth blog for Breakfast Bytes, and hardly anyone knew it existed, it wasn't widely read. So I will repeat some of the story here and then provide a preview of this year's conference. A Brief History of MemCon Last year, to find out how it all started, I talked to David Lin. He was employee #3 at Denali, the first non-engineer. He started as an AE and ended up as VP marketing. Denali was founded in 1995 as a sort of rogue EDA company in a niche that didn't exist at the time. They weren't IP (at first). They were not verification. They provided models to help verify interfaces. Their focus was on models to verify interfaces to external memory. This involved a deep understanding of the specs, getting involved with JEDEC (the standards organization. Don't worry, nobody else can remember what it stands for, either). When they started, everything was DRAM and NOR Flash. Later NOR Flash became less important and NAND Flash was the only game in town. But there were lots of new interfaces such as DDR, SDRAM, SDR, Rambus, RLDRAM, and more. Customers had no idea about them and there was no forum to discuss this. Cisco, in particular, was developing a lot of ASICs and trying to figure out how to get high throughput with low latency. Up until then, DRAM had pretty much only been used for PCs, which had different requirements. JEDEC was largely driven by Intel. Networking needed something different. But they were an example of a trend. More and more designers needed to understand memory interfaces and it was not longer just the purchasing agents who needed to get involved with the DRAM companies. The technology started to become really important, not just volumes and dollars. Architects and design managers needed to get involved. The first thing Denali did was to hire Lane Mason to produce DMR, the Denali Memory Report. This covered industry trends both for Denali internally (what should we develop?) and for the users of memory (what should we use?). For example, due to the volumes of the PC, DRAM was really cheap, but specialty memory for networking was expensive. People needed help deciding which one to use. Lane and the marketing team would talk to ASIC and networking companies on the different technologies, but they would also talk to the memory companies who had a similar problem the other way around. They didn't really know what companies outside the PC industry needed. Denali decided to open it to a broader audience. So in about 2001, they held the first MemCon with memory vendors, JEDEC, and design houses. It was in the Hyatt Hotel with 300 to 400 people attending. The most popular sessions were technical sessions on how different memory technologies would benefit different markets. They had analysts come with predictions of how the market would evolve over 1 year, 5 years, 10 years. MemCon grew during those days, when there was a sort of Cambrian explosion of different technologies and interfaces. The old interfaces were slow and parallel, the new ones were fast serial synchronous ones that presented both design and test problems. MemCon peaked with 800 to 1000 people attending with two parallel tracks over two days. There were keynotes from industry experts such as Samsung on the state of the union for flash, or the chief architect of Cisco talking about future requirements, or equipment vendors like Agilent talking about testing boards and chips with memory interfaces. The focus moved gradually from DRAM to Flash, specifically NAND Flash, which had its own set of weird requirements and needed a lot more off-chip firmware to handle things like wear leveling (making sure that all the flash is not unduly written in one place, and thus burned out prematurely) and the fact that data can only be accessed a page at a time. The PC and networking industries were mature and the focus moved to mobile and smartphones with much stricter power requirements. That brings us up to date. DRAM peaked in 1995 at the height of the PC boom at $40B. The market then fell away and it took until 2010, 15 years later, before the total memory market (including flash, etc.) was back to $40B. But since then with SSD, smartphones, and a still memory-hungry PC industry, the memory market by 2015 nearly doubled to $77B. It is forecast to shrink slightly in 2016 but still well above $70B. With memory dominating almost any SoC design, it is perhaps the most important area of all in IP, as well as being a large market on its own. To me, the most interesting thing in memory at the moment is whether the divide between "memory" and "storage" will dissolve, driven by technologies like 3DXpoint and vertical NAND flash. This has the potential to disrupt how we build memory hierarchies in servers, laptops, and perhaps smartphones. MemCon 2016 This year, MemCon opens with two keynotes. The first will be from Cadence, probably by Amjad Qureshi, who is the VP of Memory Interface IP. Just a shot in the dark, but I'm guessing he'll be talking about memory interfaces. Then Stephen Pawlowski gives the second keynote. He is the vice president advanced computing solutions at Micron. The takes us up to lunch, which will be in the exhibit hall. In the afternoon, the day splits into three parallel tracks: Enabling technologes Mainstream memories Key markets The day ends with a reception in the exhibit hall. One thing to watch out for is the Cadence/Tektronix demonstration of a 3200Mbps LPDDR4 IP and measurement solution demo. Take a look at the full MemCon agenda , or register now (it's free!) And it's the only conference with its own song. (Please visit the site to view this video) Next: Linley Gwennap: Specialization Spurs Processor Innovation Previous: TSMC: Technology Update![]()
![]()
↧