Hot Chips 33 is shaping up to be an exciting event. The annual chip conference is online only again this year, but it seems like there will be quite a few very cool presentations. Here is the advanced schedule of the event.
Since we review a lot of CPUs on STH, we are excited for these. Learning about the IBM mainframe parts with the next-generation Z(16) is always fun.
- Intel Alder Lake CPU Architectures
- AMD Next Generation “Zen 3” Core
- The >5GHz next-generation IBM Z processor chip
- Next-Gen Intel Xeon CPU – Sapphire Rapids
Two next-generation Intel parts as well as AMD Zen 3 (which we know a bit about already) make for a fun presentation.
Academic Spinout Chips
While the first group is largely the domain of major CPU providers, the academic group shows projects that are often a bit further out.
- Mozart: Designing for Software Maturity and the Next Paradigm for Chip Architectures – University of Wisconsin- Madison
- Morpheus II: A RISC-V Security Extension for Protecting Vulnerable Software and Hardware – University of Michigan
These projects are probably not what you will be buying in the next year or two, they are great showcases for the researchers and universities.
Infrastructure and Data Processors
For STH, this is always a big area. We have covered the Arm Neoverse N2 and V1 at Arm Tech Day 2021 and have started showing off the NVIDIA DPUs.
- Arm Neoverse N2: Arm’s second-generation high performance infrastructure CPUs and system products
- NVIDIA DATA Center Processing Unit (DPU) Architecture
- Intel’s Hyperscale-Ready SmartNIC for Infrastructure Processing
We have shown some of Intel’s FPGA SmartNIC C5020X combining FPGA and Intel Xeon D. Intel is resisting using the “DPU” terminology that other vendors such as NVIDIA, Fungible, Broadcom, Marvell, Pensando, and others are using. We hope Intel aligns terminology by this event.
Enabling technologies is an area where sometimes we get some really interesting presentations.
- Heterogeneous computing to enable the highest level of safety in automotive systems – Infineon
- Architecting an Open RISC-V 5G and AI SoC for Next Generation 5G Open Radio Access Network – EdgeQ
- Aquabolt-XL: Samsung HBM2-PIM with in-memory processing for machine learning accelerators – Samsung
That Aquabolt-XL may be the one to watch for STH readers.
ML Inference for the Cloud
ML/ AI inference is a big area that is growing in terms of TAM. We have three pieces here:
- Accelerating ML Recommendation with over a Thousand RISC-V/Tensor Processors on Esperanto’s ET-SoC-1 Chip
- AI Compute Chip from Enflame
- Qualcomm Cloud AI 100: 12 TOPs/W Scalable, High Performance and Low Latency Deep Learning Inference Accelerator
This is an area ripe for a company to make a major move in since it seems like we have seen a shake-out in training startups.
ML and Computation Platforms
When we discuss those training platforms, we have a number of presentations here as well.
Graphcore Colossus Mk2 IPU
The Multi-Million Core, Multi-Wafer AI Cluster – Cerebras Systems
SambaNova SN10 RDU: Accelerating Software 2.0 with Dataflow
The Anton 3 ASIC: a Fire-Breathing Monster for Molecular Dynamics Simulations
It seems like this is where one wants to get a monsterous chip name, or just call your presentation of “a Fire-Breathing Monster” to get attention. For STH, this is another key area to watch.
Graphics and Video
Of course, we review a lot of GPUs and have already covered AMD RDNA2 as well as the Google YouTube VCU for Warehouse-scale Video Acceleration.
- Intel’s Ponte Vecchio GPU Architecture
- AMD RDNA(TM) 2 Graphics Architecture
- Google’s Video Coding Unit (VCU) Accelerator
- Xilinx 7nm Edge Processors
There are going to be a lot of folks tuning in for the Ponte Vecchio GPU talk from Intel.
The new technologies segment covers some emerging areas.
- Mojo Lens – AR Contact Lenses for Real People
- World Largest Mobile Image Sensor with All Directional Phase Detection Auto Focus Function – Samsung Electronics
- New Value Creation by Nano-Tactile Sensor Chip Exceeding our Fingertip Discrimination Ability – Kagawa University
- The IonQ Trapped Ion Quantum Computer Architecture
Quantum computing and a Mojo Lens AR contact should be fun to just learn about.
There are a number of keynotes planned as well. Here is the basic outline.
- Synopsys Keynote
- Skydio Autonomy Engine: Enabling the Next Generation of Autonomous Flight
- US Department of Energy (mention of AI here)
Hopefully, we get more details on these keynotes as we get closer to the event.
Expect some major coverage of Hot Chips again from STH. We pushed out a lot of content during the event last year and we will be covering a majority of these talks when the event happens.
As a quick note, this is the program as of the time of this writing. Topics have been known to change in the past but we are excited for Hot Chips 33.