Kicking off this morning is Intel's annual technology conference, Innovation. The second year of the revived show once again has Intel's CEO Pat Gelsinger leading things off, with what's scheduled to be a 90 minute keynote.

Intel has four major themes for this year's show:

  • Accelerating the AI Era
  • Transformative Innovation for the Future of Computing
  • Hype vs. Reality: Bringing Modern Applications to Market Faster (Edge to Cloud)
  • Building and Scaling Industry-Leading Next-Generation Systems and Platforms

Expect to see AI everywhere - both in regards to silicon and promotional efforts. AI is the golden goose of the tech industry at the moment, and everyone is either making massive profits off of AI harfdware (NVIDIA), or wants to be (more or less everyone else, including Intel).

Meanwhile, on the product front, this is the year that Intel's Meteor Lake client platform is scheduled to launch. The first Intel CPU built on their EUV-enabled Intel 4 process, Meteor Lake is also Intel's first disaggregated CPU, breaking up what would be a monolithic CPU into several chiplets. Coupled with a new VPU/NPU for AI, Meteor Lake will be a chip of several firsts for Intel when it arrives.


11:28AM EDT - And things should be kicking off here in a couple of minutes

11:28AM EDT - The keynote is scheduled to begin at the bottom of the hour, but people are still coming in

11:29AM EDT - Once again, Intel is at the San Jose Convention Center for this year's conference, which is Intel's front yard, relatively speaking

11:30AM EDT - Also joining me on live blog duty today is a very tired Gavin Bonshor

11:31AM EDT - And here we go

11:31AM EDT - Opening wth a trailer

11:31AM EDT - Pat Gelsinger is writing on a whiteboard

11:32AM EDT - And now it's a Gelsinger training montage for the Innovation keynote

11:33AM EDT - And here's Pat

11:34AM EDT - He's always very energetic. This year is no different

11:34AM EDT - "We have exciting achievements to share with you today"

11:34AM EDT - "We have a lot to cover today"

11:35AM EDT - Now on stage, Rich Felton, the Director of Sporst Science and COO at ai.io

11:35AM EDT - Also, the embargo on something called "Meteor Lake" has just dropped: https://www.anandtech.com/show/20046/intel-unveils-meteor-lake-architecture-intel-4-heralds-the-disaggregated-future-of-mobile-cpus

11:36AM EDT - All about using Intel's technologies within their product stack

11:37AM EDT - ai.io benchmarked Gelsinger in relation to soccer players

11:38AM EDT - Report date: 27 May 2022?

11:38AM EDT - Cognative metrics are quite high, however

11:40AM EDT - Now on to the subject of the economy and the size of the silicon industry

11:40AM EDT - $574B trillion silicon industry

11:41AM EDT - The "siliconomy"

11:41AM EDT - 4x increase in managed devices over the past 5 years

11:42AM EDT - Computers in everything

11:42AM EDT - AI is representing a generational shift in how computing is used

11:42AM EDT - A simple rule: Developers rule

11:44AM EDT - Discussing the Intel Dev Cloud

11:44AM EDT - Intel's cloud hardware service for software development

11:44AM EDT - Now in general availability

11:45AM EDT - 3 tiers of service: freemium, commercial premium, and enterprise

11:45AM EDT - (Freemium isn't free)

11:46AM EDT - "You're not my best friends"

11:46AM EDT - Giving everyone here a week of free access to the dev cloud

11:46AM EDT - Now Intel is going to have a bit of a "Shark Tank" event

11:47AM EDT - 3 companies. They will pick a winner at the end of the morning

11:47AM EDT - Bachelor #1: Deep Render

11:47AM EDT - AI-based compression for video

11:48AM EDT - Now rolling a trailer

11:48AM EDT - Deep Render wants to do an AI-only compression pipeline

11:49AM EDT - At 5x smaller file sizes right now (not sure compared to what)

11:49AM EDT - Now on to Intel's Gaudi AI accelerators

11:50AM EDT - "The last 10 years of AI have been incredible"

11:51AM EDT - Intel has formed a new partnership with Stability.ai. They'll be building a large AI supercomputer, expected to be one of the top 15 AI supercomputers in the world

11:51AM EDT - Dell is now delivering systems with Gaudi accelerators as well

11:52AM EDT - Now on to Xeon

11:52AM EDT - Recapping 4th generation Xeon (Sapphire Rapids) progress

11:53AM EDT - Sapphire Rapids was released in volume earlier this year

11:53AM EDT - Now for another partner spot: Alibaba, who is using Xeon for their AI workloads

11:53AM EDT - Rolling trailer

11:55AM EDT - (Keep in mind that SPR introduced Intel's AMX matrix units, so they're not doing AI solely on traditional serial cores)

11:55AM EDT - Showing the latest Gaudi roadmap

11:55AM EDT - Gaudi 3 in the works on 5nm (presumably TSMC)

11:55AM EDT - Silicon is just out of fab

11:56AM EDT - Falcon Shores in 2025

11:56AM EDT - Gelsinger: Moore's Law is Alive and Well

11:57AM EDT - Bachelor #2: Scala Biodesign

11:57AM EDT - Rolling trailer

11:58AM EDT - Computational biology

12:00PM EDT - Now back to hardware with AI in the PC

12:00PM EDT - "What if we put it in the hands of every human on Earth?"

12:02PM EDT - "We see the AI PC as a sea change moment"

12:02PM EDT - What's the killer app? Intel needs developers to create it

12:03PM EDT - Demo time

12:03PM EDT - Taking a look at a couple of new AI PC applications

12:03PM EDT - Running Audacity with a plugin for Riffusion

12:04PM EDT - Looks like it's going to be a music generation demo

12:04PM EDT - A song in the style of Taylor Swift

12:05PM EDT - Meanwhile, another machine with GIMP and an OpenVINO plugin for Stable Diffusion

12:05PM EDT - Text to image demo

12:06PM EDT - Now playing the generated song. It's not bad, it's not great. But it's a lot better than what could be done on a local system 5 years ago

12:06PM EDT - As for the Stable Diffusion demo, a giraffe in a cowboy hat

12:06PM EDT - World's first showing of a Lunar Lake system

12:07PM EDT - Stable enough and running to showcase AI demos

12:07PM EDT - Lunar Lake is Intel's next next gen platform

12:08PM EDT - The successor to Arrow Lake (2024)

12:08PM EDT - Now on to the immediate future: Meteor Lake

12:08PM EDT - Launching December 14th

12:08PM EDT - (So just in time to count for 2023)

12:08PM EDT - Carrying the Core Ultra branding

12:09PM EDT - Gelsinger is discussing Microsoft's Co-Pilot software and how it runs on Meteor Lake

12:09PM EDT - Meteor Lake is a whole bunch of firsts for Intel. Intel 4 process (with EUV), chiplet client CPU, using Foveros packaging, first CPU with a high performance NPU block

12:10PM EDT - Now bringing out some ecosystem partners to talk about upcoming systems

12:10PM EDT - Starting with Jerry Kao of Acer

12:11PM EDT - (If Lunar Lake is working, that means Intel has 18A process chips working)

12:12PM EDT - Jerry is running a Meteor Lake laptop demo

12:12PM EDT - Another Stable Diffusion demo

12:13PM EDT - "Please press the button"

12:13PM EDT - And done. Quite quickly

12:13PM EDT - Acer's demo includes a parallax view ability

12:14PM EDT - And that was Acer's upcoming Swift laptop

12:14PM EDT - "But this is just the beginning"

12:14PM EDT - Wafers!

12:15PM EDT - Arrow Lake wafers. Intel's 20A process node

12:15PM EDT - RibbonFETs and PowerVias

12:15PM EDT - 2025 offering: Panther Lake

12:15PM EDT - Sending it to fabs in Q1 of 2024

12:15PM EDT - Using Intel 18A process

12:16PM EDT - So Meteor -> Arrow -> Lunar -> Panther by the end of 2025

12:16PM EDT - Now back to Xeons

12:16PM EDT - Emerald Rapids is finally confirmed as the 5th generation Xeon

12:17PM EDT - Also launching December 14th

12:17PM EDT - One of the last Intel 7 products

12:17PM EDT - EMR should bring "huge PPW improvements"

12:17PM EDT - Same platform as Sapphire Rapids, so allowing rapid adoption

12:18PM EDT - Meanwhile, the 2024 Xeons will be a next-gen platform with Granite Rapids and Sierra Forest

12:18PM EDT - P-cores and E-cores respectively. Using a shared platform, and the same I/O die

12:18PM EDT - Built on Intel 3 process technology

12:19PM EDT - Pat has a wafer of this as well

12:19PM EDT - Platform name: Birch Stream

12:19PM EDT - Following that will be Clearwater Forest in 2025, a second-gen E-core part using the same platform

12:20PM EDT - Sierra Forest will offer up to 288 cores on a single chip. using 2 dies

12:20PM EDT - (A core for every business day of the year)

12:20PM EDT - Now pivoting to the subject of security

12:22PM EDT - There will be a second keynote tomorrow, with a focus on software. So more details to come then

12:24PM EDT - Now another partner: Rewind AI

12:24PM EDT - Talking about hearing aids

12:24PM EDT - Rewind records you screen and keeps a record of it. Then you can ask Rewind about anything you've seen

12:24PM EDT - Demo time as well

12:25PM EDT - Ahh, anything seen, said, or heard

12:26PM EDT - Asking Rewind about Innovation sessions that were previously viewed

12:26PM EDT - That was apparently server-hosted GPT-4

12:27PM EDT - Now switching to a local demo, Wi-Fi off

12:27PM EDT - Running on a Core Ultra

12:27PM EDT - It knows that Pat's favorite sound is his granddaughter calling him papa (Pat had mentioned this off-hand earlier in the keynote)

12:29PM EDT - Now on to hearing aids again. How AI can improve the hearing aid experience

12:29PM EDT - Live demo with Starkey labs

12:31PM EDT - Head tracking AI, switching the mode of the hearing aids between ambient and focus modes

12:31PM EDT - Running on the NPU in the Core Ultra

12:32PM EDT - Pat's chat partner was talking in French while Pat was off doing something else

12:32PM EDT - The demo app translated it to English

12:33PM EDT - Showing off an AR visor as well

12:33PM EDT - A gimpse into sensing in the future

12:33PM EDT - Bachelor #3: Antaris (Software for Space)

12:34PM EDT - Antaris staff talking about their company and their vision

12:35PM EDT - Software-driven engineering for satellites and other space hardware

12:35PM EDT - Make all your mistakes in the cloud before you ever make a real satellite

12:36PM EDT - Making it easier to get more satellites in orbit providing services

12:37PM EDT - Now on to Intel + Arm

12:37PM EDT - Arm is supporting the OpenVino platform

12:38PM EDT - Models are going to run on everything from CPUs to NPUs and FPGAs

12:39PM EDT - Hybrid AI

12:39PM EDT - Intel will be releasing a hybrid AI SDK early next year

12:39PM EDT - (Where's oneAPI?)

12:40PM EDT - Next decade or two of development will not be cloud-native; it will be edge-native

12:41PM EDT - A lot of work to make the edge more accessible and supportable. Enter Project Strata

12:41PM EDT - Launching in early 2024

12:42PM EDT - Now for another partner demo. Fit Match.ai

12:43PM EDT - Fit Match concierge solution

12:44PM EDT - Tools to help with the issue in retail of clothes that don't fit

12:44PM EDT - Give customers a full body scan and then show a curated selection of clothing that will fit their shape

12:45PM EDT - Using a scan of Pat to show a selection of activewear

12:46PM EDT - (There's a Windows activation message in the background)

12:46PM EDT - And now back to the Dating Game (Intel Ignite)

12:47PM EDT - "They're all amazing"

12:47PM EDT - 2023 Ignite award winner: Deep Render

12:47PM EDT - 2023 Intel Startup Innovator award

12:48PM EDT - Pat previously worked on video conferencing (there's a CNET Central clip of him showing it off somewhere in my archives...)

12:48PM EDT - So this is near and dear to his heart

12:49PM EDT - Now for a live demo of Deep Render technology

12:49PM EDT - Using a Core Ultra demo laptop

12:49PM EDT - Using the NPU

12:49PM EDT - Deep Render can now bring their tech to hundreds of millions of users thanks to Core Ultra's NPU

12:50PM EDT - Looks like they're doing AI upscaling and denoising?

12:52PM EDT - Now on to fab stuff

12:52PM EDT - A look at Intel's current fab roadmap and what comes next

12:52PM EDT - Meteor Lake ramping in Oregon, working on getting it transfered to Ireland

12:52PM EDT - Looking beyond 5 nodes in 4 years

12:52PM EDT - Intel 18A: 5th node in roadmap

12:53PM EDT - Almost done with 0.9 PDK

12:53PM EDT - 18A silicon on track to go into the fab in Q1'2024

12:53PM EDT - High-NA EUV update: development with 18A. Production will now be with post-18A "Intel Next"

12:54PM EDT - (Does this mean Lunar Lake is in fact 20A?)

12:54PM EDT - Working on the next version of PowerVia as well

12:54PM EDT - Dr. Ann Kelleher's Christmas present: first High-NA EUV machines

12:56PM EDT - Recapping yesterday's glass core substrate anouncement

12:56PM EDT - Glass core substrates in HPC chips by the end of the decade

12:57PM EDT - Now switching to chiplets and the UCIe standard for an open chiplet connectivity standard

12:57PM EDT - Still early in UCIe progression

12:57PM EDT - But Intel has their first test chips

12:58PM EDT - Intel 3 die combined with EMIB and a TSMC N3E die

12:58PM EDT - Intel is looking to push the envelope on 3D silicon

12:59PM EDT - Now a word on neuromorphic computing

12:59PM EDT - Loihi 2 on Intel 4, and the stackable 8 chip Kapoho Point

01:00PM EDT - And Intel is working on quantum computing as well

01:00PM EDT - Probably post-2030 for quantum supremacy

01:01PM EDT - Intel is doing silicon qbits

01:01PM EDT - "If we get this working, we can do it at scale"

01:01PM EDT - Intel Tunnel FAlls, a 12 qubit device, 50x50nm. 95% yield rate across the wafer

01:02PM EDT - And a Quantum SDK stack is in development, as well

01:02PM EDT - And now for the Intel Innovation Lifetime Achievement Award

01:02PM EDT - Last year Linus Torvalds won it

01:03PM EDT - For 2023: Dr. Fei-Fei Li, Inaugural Sequoia Profession in the CompSci department at Standford University

01:03PM EDT - And the co-director for Stanford's Human-Centered AI

01:04PM EDT - (Who, coincidentally, is giving an industry luminary session coming up next)

01:04PM EDT - Pat is now recapping and wrapping things up

01:07PM EDT - And that wraps up the Intel Innovation 2023 opening keynote with Pat Gelsinger, Intel's CEO.

01:08PM EDT - As always, thanks for joining us Don't forget, our piece detailing Intel's Meteor Lake disclosures can be found here - https://www.anandtech.com/show/20046/intel-unveils-meteor-lake-architecture-intel-4-heralds-the-disaggregated-future-of-mobile-cpus

 

POST A COMMENT

7 Comments

View All Comments

  • m53 - Tuesday, September 19, 2023 - link

    "Sierra Forest will offer up to 288 cores on a single chip. using 2 dies"... Wow, that's new. I was expecting 144 cores only. Reply
  • mode_13h - Monday, October 2, 2023 - link

    It's cool that they can do it. I'm sure it's needed to beat AMD's 128-core/256-thread Bergamo. The key question is how they'll compare on perf/$ and perf/W. Reply
  • Roy2002 - Tuesday, September 19, 2023 - link

    Meteor lake is finally here. Reply
  • III-V - Tuesday, September 19, 2023 - link

    It's out in December Reply
  • name99 - Tuesday, September 19, 2023 - link

    I don't want to kick Intel when they are down, but for an "Innovation" keynote didn't that feel like pretty much every segment was a RESPONSE to some other company?

    Most of it, of course, felt like "we too are kinda sorta doing nVidia stuff" [lagging by three years], but the server stuff felt like "we too are gonna compete with AMD, just you wait", the visor/hearing aid stuff felt like "we too are gonna compete with Apple, just you see"; even the glass and foundry stuff felt like "hey hey, we're still relevant, TSMC isn't the only game in town".

    The neuromorphic and quantum computing stuff was the only part that didn't feel "me too" (and honestly also the stuff that felt most like "is this ever actually going to pan out?"; it's slightly disturbing that Intel can't talk about this except in terms of hitting a few checkboxes; they can't actually point to even preliminary use cases where these two schemes are achieving something dramatically different from what we can do today via traditional means).
    Reply
  • mode_13h - Monday, October 2, 2023 - link

    I think Meteor Lake is pretty innovative. It's just not news to anyone who's been following Intel and their announcements.

    APX was also rather bold, if only half of what's needed to truly address the deficiencies of x86-64. Are you not going to credit them for it, just because they didn't first announce it at this event?
    Reply
  • mode_13h - Monday, October 2, 2023 - link

    Thanks for the coverage, as always! Reply

Log in

Don't have an account? Sign up now