No video

Meteor Lake: AI Acceleration and NPU Explained | Talking Tech | Intel Technology

  Рет қаралды 16,030

Intel Technology

Intel Technology

Күн бұрын

Пікірлер: 40
@typingcat
@typingcat 8 ай бұрын
The reason most A.I. is done in cloud is lack of VRAM. For years, GPU companies have been milking on VRAM, just like Apple charging a fortune for memory upgrades. I do hope CPU's take over GPU in regard to A.I. processing so we average consumers don't have to spend thousands of dollars just to get 32GB of VRAM.
@octaviusp
@octaviusp 2 ай бұрын
i think that would happen the same as cores in cpu. Intel monopoly followed this rule of static number cores until a new competition (AMD) moves towards a new approach of more cores and unleash that every intel cpu came with more cores than before. I think that now that AMD, Intel are invading the NVIDIA Market, nvidia will create new gpus with highest VRAM for lower prices.
@Luckyn00bOC
@Luckyn00bOC 8 ай бұрын
Question : why the Integrated Arc GPU on Meteor Lake not having XMX unit for AI Acceleration? Is it because the Meteor Lake already have an NPU? Or is there some other concerns like power usage? Thanks!
@__aceofspades
@__aceofspades 8 ай бұрын
Really excited for Meteor Lakes NPU. Intel showed off a lot of partners, including big names like Adobe that are working directly with Intel on AI software integration. Intel is the only one actually bring AI to everyone. Nvidia GPUs are far too expensive for the average laptop and AMD has zero software support for their NPU. Meanwhile Intel will ship 100 million AI CPUs in 2024.
@alphaomega1969
@alphaomega1969 8 ай бұрын
Lol how do software engineer make app for AMD NPU then? 🤣😂💩
@bobpritham2660
@bobpritham2660 8 ай бұрын
Intel with this looks promising. I think they purposely made their desktop processors boring to highlight their laptop processors more.
@HDRPC
@HDRPC 8 ай бұрын
Arrow Lake Desktop with next Gen CPU cores in 2024.
@BUY.YT.VIEWS.
@BUY.YT.VIEWS. 8 ай бұрын
I'm sharing this with all my friends!
@R3endevous
@R3endevous 8 ай бұрын
Appreciate the top-down explanation. What would be the potentional gaming applications? I know that XeSS is accelerated through the XMX hardware on the arc GPUs, but a competitor has also been working on NN-based denoisers for Ray/Path tracing. Is Intel going to use the NPU + XMX for such capabilities? General applications are more relevant, of course.
@IntelTechnology
@IntelTechnology 8 ай бұрын
Thank you for those kind words! We suggest you to visit our newsroom for upcoming updates via intel.ly/3TwL66D Do not hesitate to DM us if you have any further queries.
@KellyClowers
@KellyClowers 5 ай бұрын
Thanks, I have been trying to find something that would explain a lot of the guts how NPUs work, at a technical but not field expert level... I was hoping for an article, but I haven't found any, so a video will do.
@KellyClowers
@KellyClowers 5 ай бұрын
In particular the 4th section, "NPU compute engines explained" is what I was hoping for. An expanded version of that could be even better
@SaschaRobitzki
@SaschaRobitzki 8 ай бұрын
I like to see that in your desktop and workstation CPUs.
@novantha1
@novantha1 8 ай бұрын
Any hope of getting a tutorial on getting Meteor Lake set up for running AI through Python and a couple of common libraries for developers? I talk with a lot of people that develop AI apps at the moment (and make a few contributions myself) and I'd like to look at supporting Intel's AI engine, but I don't have one personally, and I'm not sure how well supported it is on Linux (where I do the bulk of my development). A look at the process of getting it setup for development would make a big difference!
@alphaomega1969
@alphaomega1969 8 ай бұрын
Can NPU be used for video game enemy npc AI?
@isekaidev7803
@isekaidev7803 2 ай бұрын
It's up to the devs to implement AI. But currently games don't use NPU kinda AI. They use Behavior tress and state machines for AI which are run on CPU. It's hard coded AI rather than random trained AI. And as for LLMs current NPUs are too weak to run good LLMs.
@wallacetran2926
@wallacetran2926 8 ай бұрын
Can the NPU run LLMs locally?
@alejandrohoyos8555
@alejandrohoyos8555 8 ай бұрын
Yes it can. It will depend on the "size" of LLM
@GameHEADtime
@GameHEADtime 8 ай бұрын
Just waiting to buy it for a desktop pc! Integrate it with peripherals!
@albiss1164
@albiss1164 6 ай бұрын
Cool new pc era incoming =)
@DukeAbbaddon
@DukeAbbaddon 8 ай бұрын
I feel that UltraCore is a mega BASS kind of positivity, But that aside i think maybe intel does adders++ on the internal integer instructions? makes sense to me. RS
@And-vx6ry
@And-vx6ry 8 ай бұрын
Intel please do arc a990 with 4 media encoder decoder, 1440p120fps with higher wattage. Will the upcoming cpus have 24 cores ? Also Music at the end ?
@AlexSchendel
@AlexSchendel 8 ай бұрын
Alchemist has had its full stack launched. Intel has limited resources to devote to graphics, so Intel is taking a measured approach to its bet on the complexity of components (lower performance, smaller dies are cheaper to design and produce). As long as Intel continues iterating with Battlemage and Celestial, they'll get there eventually. Larger parts with more Xe Cores will come and the Xe cores will get much more performant very quickly as AXG learns through trial and error where the bottlenecks of their architecture lies. As for CPUs... 24 core CPUs have been out for over a year with the 13900K. The 14900K also has 24 cores. If you're looking for all P-cores, then there's the Xeon w9-3975X as an unlocked, in-box workstation CPU with 36 P-cores.
@rebootcomputa
@rebootcomputa 2 ай бұрын
So is any of this NPU going to help with tasks outside of AI operations?? like in normal software that most people that are going to buy a portable laptop? most consumers of laptops arent going to use AI software locally, if you want high end AI computation its a cool idea to be able to do it on a laptop, but I feel like this isnt going to help performance for most consumers of laptops... unless they can accelerate software for video or image rendering.. this feels a lot like physic accelerated chips that nvidia put on their cards, Physx was pretty much abandon years later because there wasnt an incentive for devs to use it...
@garywicks5927
@garywicks5927 2 ай бұрын
NPU doesnt accelerate graphics.. not to say a dev cant use it in graphics design software for an AI add in, but currently NPU does nothing for Adobe products for example
@misterPAINMAKER
@misterPAINMAKER 8 ай бұрын
Imagine if in a few years videogames require NPU to run.
@Zreknarf
@Zreknarf 7 ай бұрын
I read a rumor windows 12 coming in June will require an NPU to run, and the next xbox will have an NPU
@misterPAINMAKER
@misterPAINMAKER 7 ай бұрын
@@Zreknarf i hope not
@Zreknarf
@Zreknarf 7 ай бұрын
@misterPAINMAKER it looks like GPUs can already do anything NPUs can do, it's just ~50x less efficient. so I imagine you'll be able to run any NPU optimized game on a GPU, just won't be power (or cost) efficient.
@garywicks5927
@garywicks5927 2 ай бұрын
@@Zreknarf FYI an nvidia GPU has massively more AI acceleration capability than any NPU - up to thousands of tiems more depnding on the GPU
@Zreknarf
@Zreknarf 2 ай бұрын
@garywicks5927 any CPU integrated NPU sure, I'd argue that's not true if you consider discrete NPUs like the $30,000 nvidia H100. and still the CPU integrated NPU will be 50x more efficient in those tasks than a GPU, which is nothing to sneeze at
@FluidRotor
@FluidRotor 9 күн бұрын
Perez Eric White Robert Martinez Jessica
@FluidRotor
@FluidRotor 5 күн бұрын
Hall Nancy Smith Larry Hall Brian
@AjidaHasan
@AjidaHasan 12 сағат бұрын
Allen Elizabeth Martin Laura Robinson Karen
@Havagah
@Havagah 8 күн бұрын
Taylor William Anderson Mary Miller Barbara
@aeromotive2
@aeromotive2 8 ай бұрын
shintel baby
@Metalmassacre07
@Metalmassacre07 8 ай бұрын
Snakeoil Lake
Linus Torvalds: Speaks on Hype and the Future of AI
9:02
SavvyNik
Рет қаралды 179 М.
WILL IT BURST?
00:31
Natan por Aí
Рет қаралды 30 МЛН
The Giant sleep in the town 👹🛏️🏡
00:24
Construction Site
Рет қаралды 21 МЛН
Unveiling my winning secret to defeating Maxim!😎| Free Fire Official
00:14
Garena Free Fire Global
Рет қаралды 11 МЛН
Do we really need NPUs now?
15:30
TechAltar
Рет қаралды 483 М.
The Coming AI Chip Boom
15:41
Asianometry
Рет қаралды 346 М.
Stack Overflow stopped caring about developers a long time ago
22:33
Coding with Dee
Рет қаралды 65 М.
CPU vs GPU vs TPU vs DPU vs QPU
8:25
Fireship
Рет қаралды 1,7 МЛН
Why Lunar Lake changes (almost) everything
19:09
High Yield
Рет қаралды 109 М.
WILL IT BURST?
00:31
Natan por Aí
Рет қаралды 30 МЛН