Hot Chips, where the semiconductor industry struts new stuff, focuses on the IoT, with big announcements expected from Applied Micro, Nvidia, Oracle, and interesting smaller companies.
I've been attending the Hot Chips technical symposium for many years. Over the last few years it has become a premier venue to reveal new chip architectures, replacing the venerable, but now defunct, Microprocessor Forum in that role. In fact, I value the conference so much, I spent a number of years volunteering on the organizing committee.
This year's program was released last week, and there are some significant presentations and timely tutorials and keynotes you should know about. While abstracts are not part of the program (which might get fixed soon), it's possible to infer some of the content from the titles. I also talked with Sam Naffziger of AMD, one of the Hot Chips program chairs, to gain a little insight beyond the titles.
This year's Hot Chip starts off on Sunday August 10, with two tutorials -- one on hardware support for security and the other on the Internet of Things (IoT). The two session have some synergy (yes, I used that word) because IoT is going to need security as it is widely deployed. The security session includes presentations from vendors AMD, ARM, and Intel. The IoT tutorial includes presentations from ARM, Qualcomm, and TI.
The keynotes this year take on a practical orientation (no Google self-driving car) with ARM talking about power-constrained design (something they may know something about) and a second keynote by Rob Chandhok of Qualcomm breaking down various aspects of the Internet of Everything.
Monday, August 11, kicks off the general session presentations with high-performance computing. The chips from NEC, Fujitsu, and D.E. Shaw are specifically targeting fast computing workloads and are not mainstream. But they do offer a look at some unique and purposeful designs -- it's like visiting the rare birds cages at the zoo; very unusual, but you're not taking one of them home. For example, the Anton 2 chip from D.E. Shaw is an ASIC designed to perform molecular dynamics simulation, and it's a monster chip.
The mobile processor session blurs smartphone and tablet SoCs with notebook processors. It includes Nvidia's latest Tegra K1 chip, which has design wins in Google's Project Tango tablet and in automotive. Nvidia gets a second presentation to reveal for the first time the architecture of long-awaited Project Denver, the company's custom ARMv8 CPU. After years of speculation, we get to see if the leaks about binary translation are true. While there have been a lot of revelations about AMD's Kaveri, this paper will drill deeper into its HSA implementation.
In the technology session, SK Hynix will propose a solution to increasing memory bandwidth, which is an issue for all high-performance computing. In the same session the always engaging Dave Ditzel will present ThruChip's wireless signal interconnect for 3D stacked silicon. As biomedical computing is becoming critical to future medical care, the interest in biomedical chips is increasing. Insilixa will show its solution for point-of-care molecular diagnostics.
The growth of ARM-based servers requires its own session. Facing off are AMD's Seattle (which is now sampling) and Applied Micro's second-generation ARMv8 processor, the 28nm X-Gene 2. ARM and LSI will discuss networking fabric implementation.
Next Page: FPGAs & ASICS plus Intel & IBM