Intel Innovation 2022 Keynote: Live Blog (9am PT, 4pm UTC)
source link: https://www.anandtech.com/show/17594/intel-innovation-2022-keynote-live-blog-9am-pt-4pm-utc
Go to the source link to view the article. You can view the picture content, updated content and better typesetting reading experience. If the link is broken, please click the button below to view the snapshot at that time.
Intel Innovation 2022 Keynote
11:29AM EDT - Welcome to our live blog of the opening keynote of Intel's Innovation 2022. Things should be kicking off very soon
11:30AM EDT - Looks like people are starting to take their seats
11:30AM EDT - And we begin
11:31AM EDT - An edit: Only Pat Gelsinger, Intel's CEO will be speaking today
11:32AM EDT - And it starts in 30 minutes, be back soon
11:57AM EDT - Ryan here, in both senses of the word. Intel is hosting this year's event as an in-person function at the San Jose Convention Center, my home away from home
11:58AM EDT - Being an in-person event, this should be a more leisurely paced event than the kind of rapid-fire keynotes we've seen during the pandemic
11:58AM EDT - But first, a spoiler:
11:59AM EDT - Newegg posted Intel 13th Gen Core (Raptor Lake) processors for sale earlier this morning
11:59AM EDT - Complete with prices
11:59AM EDT - We'll see if the prices are accurate. But at this point they seem realistic
12:00PM EDT - Meanwhile, on screenshot duty is the always-awesome Gavin Bonshor
12:00PM EDT - And this keynote should be kicking off momentarily
12:01PM EDT - Expected subjects today inclue Raptor Lake, Sapphire Rapids, Intel's enterprise GPU efforts, and more
12:01PM EDT - And here we go
12:01PM EDT - Starting with Pat Gelsinger
12:01PM EDT - "We're going to have a great event together now"
12:01PM EDT - This keynote is slated to run for 1 hour
12:02PM EDT - Pat is starting with reiterating Intel's focus on open ecosystems and a developer-first approach
12:02PM EDT - News today on hardware, software, chip design, security, and more
12:03PM EDT - (Pat's shirt is "geek" in ASCII binary)
12:03PM EDT - "We are in a new era"
12:04PM EDT - "We continue to witness the magic of technology"
12:04PM EDT - Pat is reiterating his "superpowers" analogy
12:04PM EDT - Compute, connectivity, cloud-to-edge infrastructure, AI, and sensing
12:05PM EDT - Oh? He's added a 5th superpower. Sensing is new to the list
12:05PM EDT - Pat shows off his hearing aid
12:06PM EDT - An example of a sensory device today. A device that makes his life better
12:06PM EDT - It also changes how Intel thinks about their business and chip design
12:06PM EDT - Which brings us to Intel Foundry Services
12:07PM EDT - At the heart of chip fabbing is Moore's Law
12:07PM EDT - Pat reiterates that he doesn't consider Moore's Law to be dead
12:07PM EDT - Going from 100B transistors today on a single package to 1T transistors by the end of the decade
12:07PM EDT - Making process on Intel's "5 nodes in 4 years" development plan
12:08PM EDT - Expecting first 18A test chip tape-out before the end of the year
12:08PM EDT - "We will not rest until the periodic table is exhausted"
12:08PM EDT - Intel will continue to work to grow capacity and demand
12:09PM EDT - With their previously announced Tower Semi acquisition, it gives Intel a significant bredth in capabilities
12:09PM EDT - Now talking about the idea of a Systems Foundry
12:09PM EDT - A Systems Foundry offers wafers, advanced packaging, chiplets, and software
12:10PM EDT - "Software defined, silicon enhanced"
12:11PM EDT - Investing heavily in the UCIe chiplet ecosystem
12:11PM EDT - "We're going to bring all of those pieces together"
12:11PM EDT - Now rolling a video about UCIe
12:12PM EDT - Which starts with a statement from TSMC
12:12PM EDT - Intel is emphasizing the "universal" part of UCIe
12:12PM EDT - Part of Intel's $1B IFS innovation fund is for startups and disruptive technologies
12:13PM EDT - Intel is also scaling up their university shuttle program. The goal: 100x
12:14PM EDT - Looking to develop the semiconductor talent pool of tomorrow
12:14PM EDT - Now on to graphics and accelerated computing
12:14PM EDT - High-perf computing requires vector and matrix processing
12:15PM EDT - The only thing Pat didn't do in his previous stint at Intel: high-performance graphics
12:15PM EDT - Now on to GPUs
12:15PM EDT - Announcing the Intel Data Center GPU Flex series
12:15PM EDT - Flex 170 and Flex 140
12:15PM EDT - For inference, training, AI, VDI, video encoding. Everything GPUs are used for today in servers
12:16PM EDT - Pat also has a Ponte Vecchio package in his hand
12:16PM EDT - PVC and Sapphire Rapids with HBM are the computing brains behind the Aurora supercomputer
12:16PM EDT - Intel is delivering those blades now
12:17PM EDT - But you can't talk about GPUs without talking about gaming
12:17PM EDT - Paraphrasing: the price of gaming GPUs is too damned high
12:17PM EDT - "Today we're fixing that"
12:17PM EDT - Intel Arc A770 GPU
12:17PM EDT - 65% better RT performance than the unspecified competitiomn
12:17PM EDT - Now rolling a demo video
12:18PM EDT - Available October 12th at $329
12:18PM EDT - Cards on the way to reviewers now
12:18PM EDT - So putting it up against the NVIDIA RTX 3060 and AMD RX 66xx parts
12:20PM EDT - Shifting gears, bringing on stage Intel's AI Ethics Lead Architect, Ria Cheruvu
12:20PM EDT - She's just 18
12:21PM EDT - We'll hear more about this in Greg's session tomorrow: the Intel Developer Cloud
12:21PM EDT - Which Intel uses to showcase some of their latest technologies
12:21PM EDT - The Dev Cloud is current in a limited beta, but Intel is intending to expand that
12:23PM EDT - Intel's first challenge: speeding up development cycles
12:23PM EDT - Development cycles are too slow
12:23PM EDT - Intel has a tool to help with no-code solutions
12:24PM EDT - The Neural Compressor with Neural Coder
12:24PM EDT - Converted a PyTorch model from FP32 to INT8
12:24PM EDT - Running on Sapphire Rapids
12:25PM EDT - 10.8x speed-up on SPR versus using FP32
12:25PM EDT - Now on to security
12:25PM EDT - Demoing Intel SGX
12:26PM EDT - (SGX seems to get compromised with some regularity)
12:26PM EDT - Working to enable a holistic view of confidential computing
12:27PM EDT - Breaking news: Raptor Lake launches on October 20th. i9-13900K is $589
12:28PM EDT - Now on to AI model development
12:29PM EDT - An enterprise platform for collaborating on AI development
12:29PM EDT - Showcasing how to build a CV model to maximize coffee bean yield
12:30PM EDT - Using active learning to cut down on the number of sample images required
12:31PM EDT - Showing how the model performs so far, and how it's flaws can be corrected
12:31PM EDT - Model can then be optimized with OpenVINO and deployed across datacenters
12:32PM EDT - Launching today: Intel Geti
12:32PM EDT - "Powerful AI for everyone"
12:32PM EDT - Commercially available in Q4 of this year
12:33PM EDT - (Just to be clear, Geti is a computer vision training system)
12:33PM EDT - Continuing the subject of building AI models. Now with a guest speaker from PreciTaste
12:34PM EDT - Using OpenVINO for its flexibility in being able to run against GPUs or CPUs (rather than being limited to just GPUs)
12:35PM EDT - 30% higher FPS in one afternoon
12:35PM EDT - Using Intel's RealSense cameras to watch pans of food. OpenVINO used to classify things and how much there is
12:36PM EDT - Shuffle pans; AI picks up on the change
12:38PM EDT - Chipotle will be deploying this Intel tech in a large market in the "coming months"
12:38PM EDT - Developer challenge 3: game development is complex and slow
12:39PM EDT - Discussing debugging of networked clients
12:40PM EDT - The game is Nightengale, a UE5 game
12:41PM EDT - High perf requirements mean that normally, only one or two clients could be run on a single workstation
12:41PM EDT - (Technical error: keyboard is not working)
12:42PM EDT - Using an unclear tech to allow a system to run up to 8 instances
12:42PM EDT - Ahh, this is a demo of having so many cores on an Intel CPU
12:43PM EDT - Particularly the E-cores
12:44PM EDT - Now rolling demo footage of the game itself
12:44PM EDT - Pat is hiding something behind his back
12:44PM EDT - Poorly
12:45PM EDT - Today Intel is announcing the 13th Gen Intel Core processor family
12:45PM EDT - Codename: Raptor Lake
12:45PM EDT - Flagship i9-13900K
12:45PM EDT - 8 P cores plus 16 E cores (twice as many E cores as before)
12:46PM EDT - Best gaming/streaming experience ever
12:46PM EDT - DDR5 speeds up to DDR5-5600
12:46PM EDT - "World's best overclocking experience as well"
12:46PM EDT - Will be releasing a "limited volume" 6.0GHz SKU next year
12:47PM EDT - So 13900KS?
12:47PM EDT - Over 50 processors planned as part of the 13th Gen Core family
12:48PM EDT - Developer challenge 4: creative development is time-consuming and limited
12:48PM EDT - (Someone needs to do a Stable Diffusion or similar image having Pat Gelsinger show off Ryzen CPUs)
12:49PM EDT - Showing off Stable Diffusion here running on Intel's Gaudi 2 hardware
12:50PM EDT - Showing off the "incredible power" of the next-gen Gaudi 2 accelerator
12:51PM EDT - Developer challenge 5: User experience needs to drive product design
12:52PM EDT - User experience needs to drive product design
12:52PM EDT - Samsung Display's CEO, JS Choi, is now on stage
12:52PM EDT - Choi is holding a rollable OLED display
12:52PM EDT - Expanding and shrinking it
12:53PM EDT - Announcing world's first 17-inch slidable display
12:55PM EDT - Now on to showcasing Intel Unison
12:55PM EDT - Unison promisses to "bring all of our devices together"
12:56PM EDT - Extending his laptop's display to the Samsung screen
12:57PM EDT - Unison can also extend iOS and Android devices
12:57PM EDT - iOS texts on a Windows PC. That's a pretty killer feature
12:57PM EDT - Unison coming to new laptops this holiday season
12:57PM EDT - On select Intel Evo devices
12:58PM EDT - 6th and final challenge solve high-bandwidth low-power connectivity
12:58PM EDT - Silicon Photonics
12:58PM EDT - Showing an old video of a much younger Pat talking about silicon photonics
12:59PM EDT - Silicon photonics has taken longer than expected
12:59PM EDT - But it's finally coming. And is an example of how Intel is investing in the long term
01:00PM EDT - Detachable optical in-package connector
01:00PM EDT - A live demo being streamed out of Intel's labs in Scotland
01:01PM EDT - Showing a demo measuring the light coming out of the silicon photonics chip
01:01PM EDT - Let there be light
01:01PM EDT - Now detaching and reattaching the connector
01:02PM EDT - "This is exactly how I dreamed of it decades ago"
01:02PM EDT - Now wrapping things up
01:03PM EDT - Signing one of the posters they created with Stable Diffusion earlier
01:03PM EDT - To close things out, Pat is bringing up Linus Torvalds
01:06PM EDT - Linus is recounting the creation of Linux. He was poor and could not afford a Unix
01:07PM EDT - Extolling the virtues of open source
01:08PM EDT - Linus Torvalds, Pat's favorite plodding engineer
01:09PM EDT - Linus is the first recipient of a new award. The Intel Innovation Award
01:09PM EDT - "That is very nice and classy"
01:09PM EDT - Lifetime Achievement award
01:10PM EDT - And that's Linus
01:10PM EDT - Pat's now quickly hyping up the crowd about the next 2 days
01:11PM EDT - And that's a wrap
01:11PM EDT - Thanks for joining us for another live blog. Now off to check out some other sessions here at Innovation
Recommend
About Joyk
Aggregate valuable and interesting links.
Joyk means Joy of geeK