Wednesday, September 20th 2023
Intel Demoes Core "Lunar Lake" Processor from Two Generations Ahead
Intel at the 2023 InnovatiON event surprised audiences with a live demo of a reference notebook powered by a Core "Lunar Lake" processor. What's surprising about this is that "Lunar Lake" won't come out until 2025 (at least), and succeeds not just the upcoming "Meteor Lake" architecture, but also its succeeding "Arrow Lake," which debuts in 2024. Intel is expected to debut "Meteor Lake" some time later this year. What's also surprising is that Intel has proven that the Intel 18A foundry node works. The Compute tile of "Lunar Lake" is expected to be based on Intel 18A, which is four generations ahead of the current Intel 7, which will be succeeded by Intel 4, Intel 3, and Intel 20A along the way.
The demo focused on the generative AI capabilities of Intel's third generation NPU, the hardware backend of AI Boost. Using a local session of a tool similar to Stable Diffusion, the processor was made to generate the image of a giraffe wearing a hat; and a GPT program was made to pen the lyrics of a song in the genre of Taylor Swift from scratch. Both tasks were completed on stage using the chip's NPU, and in timeframes you'd normally expect from discrete AI accelerators or cloud-based services.
Source:
HotHardware
The demo focused on the generative AI capabilities of Intel's third generation NPU, the hardware backend of AI Boost. Using a local session of a tool similar to Stable Diffusion, the processor was made to generate the image of a giraffe wearing a hat; and a GPT program was made to pen the lyrics of a song in the genre of Taylor Swift from scratch. Both tasks were completed on stage using the chip's NPU, and in timeframes you'd normally expect from discrete AI accelerators or cloud-based services.
62 Comments on Intel Demoes Core "Lunar Lake" Processor from Two Generations Ahead
Pretty weird to look at well paid, educated person who wears a hat indoors. If he doesn't care about etiquette and many do also, he is publicly on stage with many persons who might find it offensive as those really are basic manners and showing respect to others.
This person may have well-founded reasons (hair/skin problems, etc.) for not revealing his own head.
What worries me about AMD is they don't really have an overall good product. 3D chips are good for gaming, non-3D chips are good for applications. Intel has overall better performance across the board. Might loose slightly in gaming (compared to 78003D) but even 13600K destroys 7800X3D outside of gaming is my point)
I really really hope that AMD will do 3D cache on all cores for 8000 3D chips. Not only one CCD like 7900X3D and 7950X3D and hopefully they can bump up clockspeeds as well.
7800X3D is great for gaming but really not that great outside of gaming, meaning 7700X performance or even less. 7800X3D can beat 7950X3D in some games, showing why you don't want a half solution.
Give me 8950X3D with 3D cache on all 16 cores and I will probably buy that, unless Arrow Lake delivers a big step forward (which it should).
If Intel truly has 20A/18A ready in 1-2 years, then AMD might be in trouble. Ryzen mostly has been a success because AMD had a node advantage (Ryzen 1000 and 2000 series on 12nm GloFo kinda sucked) and AMD is 100% reliant on TSMC to deliver improvements (sadly for AMD, TSMC prioritizes Apple) but without Apples money, TSMC would not be where they are today.
Meteor Lake which is out later this year is mobile only, then Arrow lake is coming 2nd half of next year but (for all intents and purposes) is only on desktop, then in 2025 (hopefully early) Lunar Lake is coming out which is again exclusive to mobile. So it's not that strange really, Meteor Lake is totally complete and coming out soon, and Intel are just showing that the chip that will replace Meteor Lake is up and running already "so please don't give up on us"!
As for the manufacturing nodes, Intel 7 has been around for ages, Intel 4 and Intel 3 are basically the same node (Intel 4 is what Intel is exclusively using to iron out the bugs for Intel 3 which has very small tweaks to it's libraries and Intel are opening this up to all it's partners), and it's the exact same thing for Intel 20A & 18A. So in reality, Intel only really have 2 new node since Intel 7 which came out over a year and half ago.
If he was in a wheelchair, would he have to prove to you that he cant walk?
Are you one of those that says Jamie Fox is just faking it?
Or can we leave people with a bit of dignity and assume its for a reason.
Heck what do basic manners say about some sort of disgusting wound or malady on the head? expose it to all so they know there is a reason to cover it up....and then to cover it up?
or hell maybe its because of a general lack of basic manners from others that made him self conscious enough to keep the hat on, like something simple as getting bald.
anywho, yeah, idk man but you have an odd point of view, but that is just my opinion, equally valuable/meaningless.
Moreover, and not directed at you, but isn’t this guy presumably in some authoritative position where they get to determine how Intel is represented and, perhaps, even maybe, given some authority of the dress code? Isn’t this the whole point of determining what is and professional and what isn’t — making the rules about professionalism?
This a message that Intel isn't capable of picking right people for right tasks again. Not mentioning again this whole show is about nothing.
Welcome to the world, it ain't a small sandbox of your values, you have to consider others, not only you. If something is a miss, maybe pass the torch to someone else for public stunts.
Also running of low power NPU tile is not a proof for working node. Cannon lake released in 2018 was working, yes, with performance level of mobile celeron, but it worked
I'm more afraid of the image classification part. I like doing that myself with folders - I don't see why my PC or phone should do it for me. I mean, if my devices are so efficient at classifying my stuff, then how do I know that they're not equally efficient at sending some info about it to Google/Microsoft without my knowledge?