Hot Chips Symposium Sponsor Sneak-Peak: Intel, Synopsys, Flex Logix.

4 minute read

It’s the height of summer, and in less than a week, Hot Chips 33rd Symposium is kicking off next Sunday August 22. Let’s take a closer look at some more of the sponsors of the event…

Intel is the Rhodium Sponsor of Hot Chips again this year, and has a lot going on at Hot Chips. So far 2021 has been an adventurous year for Intel as their new CEO, Pat Gelsinger just took the helm in February, and just a few weeks ago in late July Intel’s executives had a webcast to explain more about the recent announcements covering their packaging and process technologies roadmaps.

There’s also been buzz about Intel’s upcoming Innovation event in October. So we’re excited to see what Intel will disclose, and here’s a summary of what to expect from Intel at Hot Chips:

  • Deep Learning optimizations for CPUs — Given the pervasiveness of CPUs, Intel is taking part on Day 1’s Tutorials day to showcase various techniques such as low precision, sparsity, and ops fusion to reduce memory accesses. Given the industry activity in building dedicated AI accelerator silicon, it’s sure to be an interesting session in contrast with a CPU focus.
  • Advanced Packaging technologies tutorial, showcasing heterogeneous integration — Intel will talk more about chiplets in their afternoon Tutorial on Sunday afternoon, talking about compact power-efficient platforms using chiplets. Expect to hear details about leading-edge 2D and 3D technologies (as well as a case study) from this major chip manufacturer.
  • Next-gen Alder Lake CPU architecture and Next-gen Xeon Scalable processor (code named Sapphire Rapids) — On Monday, “Conference Day 1”, Intel will kick off the CPU presentations with not one but two major disclosures. First, an anticipated disclosure on Alder Lake CPU architecture. Earlier this year back in January, Intel’s outgoing CEO was reported saying that Alder Lake will be Intel’s most power-scalable SoC for desktop and mobile. There has been so little detail about Alder Lake so far, so this Hot Chips talk is sure to be interesting. A second CPU session will cover Intel’s next-gen Xeon processor dubbed Sapphire Rapids. About a month ago, Intel provided an update on Sapphire Rapids including discussing two new breakthrough technologies, including an update on Intel Advanced Matrix Extensions (AMX), which is next-generation built-in DL Boost advancement for deep learning performance and features a set of matrix multiplication instructions that will significantly advance DL inference and training. With early silicon, Intel is saying that Sapphire Rapids is easily achieving over 2X the deep learning inference and training performance compared with the current Xeon Scalable generation. This will be an exciting CPU session of talks for sure.

  • New Hyperscale-ready IPU — in Monday afternoon’s Infrastructure and Data-processing session, Intel will disclose architectural details on its next Infrastructure Processing Unit (IPU). Intel just announced its IPUs less than 2 months ago, a new product line it’s developing in collaboration with its hyperscale cloud partners.

  • Intel’s Ponte Vecchio GPU — finally, in Tuesday Afternoon’s session on GPUs, Intel will disclose details about its first Xe-HPC architecture based GPU dubbed Ponte Vecchio, first described back in March. Given the industry’s fascination with GPUs for high performance visual applications, this will be an interesting bookend to the many disclosures from Intel at this coming Hot Chips.

Our first keynote of Hot Chips on Monday afternoon features our platinum sponsor Synopsys and their CEO Aart de Geus speaking on Artificial Intelligence in chip design.

Recently Aart spoke with Anandtech about the upcoming Hot Chips keynote and said that it’ll be about the path to better chip design by using machine learning tools. For example, Aart talked about how it’s latest AI-accelerated placement tools are helping companies design better processors with accelerated time to market. But there’s an economic component to the problem, too, where companies rely on others’ progress in a mutual way. And according to Aart, there is so much more to come in terms of optimized chip layouts using AI.

Also, hot off the press this past week Wired magazine featured discussion about Synopsys using AI in design. The design tools perspective makes this a truly hot topic for Hot Chips, so this is bound to be a great opening keynote.

Finally, let’s showcase our Platinum sponsor, Flex Logix. They’re a reconfigurable computing company with an AI edge inference accelerator known for power efficiency and cost efficiency. Flex Logix say they’ve been busy staffing up as their X1 AI Edge inference accelerator moves to production in Q4. What makes Flex Logix technology interesting is the incorporation of reconfigurability — in the chip a set of dynamic tensor processors are continuously reconfigured during model execution, which allows the tensor processors to be optimized for each layer of a deep neural network, and reduces data movement on the chip. Flex Logix says they’re supporting evaluation; for evaluation purposes the X1 accelerator is available in both PCIe and M.2 form factors. For more information at Hot Chips you can look for information at the Hot Chips virtual expo.

With these three sponsors and an exciting program kicking off next week, we look forward to another great set of talks for the Hot Chips 2021 virtual conference! You can still register at hotchips.org, so we hope to see you there!

Updated: