Intel unveils new hardware, software and services at inaugural Intel Vision 2022 event

Intel also provided updates on several important projects and other future items, which are still some distance on the horizon.

DALLAS — On Tuesday, Intel kicked off their first large-scale face to face event since the pandemic began, with Intel Vision 2022. This is a new event, which was also available online, and which the company is billing as focusing on the future of business and technology. This involved a decided focus on customers, over a broad swath of industries, explaining how they developed cutting edge capabilities through Intel silicon, software, and services.

“Today, we can get the scale and capability of the cloud through the intelligent edge,” said Intel CEO Pat Gelsinger. “Intelligence is everywhere, turning infinite data into actionable insight. We are driving this torrid pace, moving rapidly with decisiveness. We believe the world is at this strategic inflection point where things can go incredibly well or incredibly poorly, based on decisions made in a short period of time.”

Gelsinger also highlighted a theme that was a constant one during the first day – the need to be much more proactive in diversifying the semiconductor supply chain, from Intel’s own efforts to the stalled efforts of the U.S Congress around the CHIPS act, which is designed to onshore more of the industry and reduce both supply chain pressures and the impact on chip important that could be impacted by tensions in the U.S. China relationship. It has been stalled in conference between the two Houses of Congress, which is designed to reconcile the two versions of the bill.

Pat Gelsinger onstage at Vision 2022

“Chips have never been so sexy,” Gelsinger told his audience. “But we also need a diverse geographically balanced supply chain.”

Michelle Johnston Holthaus, EVP and GM of the Client Computing Group at Intel, then took the stage to announce the final release in Intel’s 12th Gen Intel Core HX processors, which Intel has termed their most significant breakthrough in x86 architecture in over a decade. The HX processors are aimed specifically at the workstation market.

“The PC remains the foundation of computing, even though over 140 million commercial devices more than four years old are still being used globally,” she said. An October 2020 TechAisle study showed that it cost $1500 a year to maintain an old commercial PC.

Johnson Holthaus noted that the 12th Gen Intel Core HX workstation processors have up to 16 cores and clock speeds up to 5 GHz.  They also provide up to 45% better performance in media, 72% in life sciences, 79% in finserv and 82% in energy, oil and gas than their predecessors.

Sandra Rivera, EVP and GM of Data Center Solutions, then took the stage to announce that the initial SKUs of 4th Gen Intel Xeon Scalable processors, formerly codenamed Sapphire Rapids, are now shipping. These are the first of many new Xeon SKUs, with more due to ramp throughout the rest of the year.

“StubHub gained 64% performance moving to cloud with these, while lowering licensing costs at the same time,” she said.

Rivera also announced that Intel unveiled its IPU roadmap extending through 2026. It features new FPGA and Intel architecture platforms (code-named Hot Springs Canyon) and the Mount Morgan (MMG) ASIC, as well as next-generation 800GB products. She also called attention to the importance of the recent Granulate acquisition, which makes real time autonomous continuous learning software for AI-based workload optimization.

Rajah Koduri, executive vice president and general manager of the Accelerated Computing Systems and Graphics (AXG) Group at Intel, provided an update on the Arctic Sound M Data Center GPUs , which will be available in the third quarter of 2022, and will  be targeted at media transcode, visual graphics and inference in the cloud.

“This is a Super Flexible Data Center GPU that runs the same software stack that already exists in Xeon,” he said. “We are rolling this all out next quarter.”

This data center GPU is the industry’s first discrete GPU with an AV1 hardware encoder, which targets 150 trillion operations per second (TOPS). It will be available in two form factors and in more than 15 system designs from partners including Dell Technologies, Supermicro, Cisco, Inspur, HPE, and H3C.

“We aren’t planning to provide games ourselves with this, but we are working with providers in the cloud gaming space,” said Jeff McVeigh, VP and GM of Intel’s Super Compute Group. “AliBaba was wired with precursors to Arctic Sound, and we are having conversations there with Arctic Sound being the focus.”

Koduri also provided an update on Intel’s Project Endgame,

“It is continual compute for the immersive web – the metaverse,” he said. “We can’t wait to show a demo. We will be beta testing continual compute later this year, and expect to see a full version for the metaverse four to five years from now.” Later in a media breakout, he fudged on the date somewhat, indicating he did not want to be tied to a date, but is optimistic they will see rapid progress over the next three, four or five years.”

While gaming is a splashy showcase for Project Endgame, Koduri also emphasized it is about more than gaming.

“We are building Endgame so it is not just cloud, and could take advantage of computing in your device,” he said. “It’s not typical cloud gaming stuff.”

Intel also discussed its plans to make high performance computing much more ubiquitous, highlighting a customer, Argonne National Laboratories, which is on track to deliver 2 exaflops of peak performance with the Aurora supercomputer running on the Intel Xeon processor, code-named Sapphire Rapids, and the Intel data center GPU, code-named Ponte Vecchio.

“We need a lot of flops,” said Rick Stevens, Associate Laboratory Director for Computing, Environment and Life Sciences at Argonne National Laboratory. “Exascale is just the beginning to solve future simulation products like around  weather and cancer treatments.

Stevens said they are just starting to build the Aurora 2 Exaflop computer.

“This will be one of the world’s most powerful systems when it comes online,” he stated. “There are major software challenges in programming a machine with so many compute elements. This uses one interface to program CPUs and GPUs. You don’t have to change code. It just works.

“We are making progress in making this available to everyone,” Stevens stated. “Aurora will be available now for reservations for industrial and scientific research, in order to have apps running from Day One.”

Rivera then returned to the stage to introduce Project Apollo, which in partnership with Accenture, will provide enterprises with open-source AI solutions kits that will make AI more accessible to customers in on-prem, cloud and edge environments.

“Project Apollo will provide enterprises with over 30 open source AI reference kits,” Rivera said. “The first kits will be released in July.”

Finally, Rivera unveiled Habana Gaudi2, the second generation of their highest end deep learning AI training processors. Habana Gaudi2 and Greco AI accelerators are built on a single software stack, Synapse AI, that easily supports different architectures, enabling end-users to take advantage of the processors’ performance and efficiency. Gaudi2 also delivers two times better AI training performance compared with current inarket A100-based offerings.

“With Gaudi, we are able to address the highest-end deep learning cases, to train increasingly large data centre and complex models,” she said. “Gaudi2 is based on our 7nm silicon and provides a major leap in deep learning performance, with up to 40% better price performance.”