At this point, I would argue that on-device AI is more hype than a real thing. Microsoft's on PC AI known as Windows Recall is essentially an improved full-text search of PC snapshots, and local chatbots are not yet on par with online AI chatbots. I have a very capable computer and have tested several AI desktop applications such as AnythingLLM, Misty, and LM Studio. I've experimented with various 7-billion and 13-billion parameter LLMs from Hugging Face. In my experience, there's a significant difference in the 'intelligence' shown by these local models compared to online AIs like ChatGPT, Claude, or Mistral, which use models with trillions of parameters. (especially when using different human languages) On-device LLMs might be suitable for simpler tasks, but to be truly useful, they likely need some form of internet connection to access more advanced AI processing capabilities. The current limitations of local processing power and memory make it challenging for on-device AI to match the performance of cloud-based solutions. As well there was an article that new AI PCs produce 40-50 TOPs however a gaming laptop with an Nvidia RTX 40 series can produce more than 800 AI TOPs.
AI at this point is just a more hype than substance. Current AI PCs have anywhere between 10 and 40 TOPS. However, as per Nvidia, PCs need atleast 140 TOPS to fully realise AI. So, spend your money wisely!
Please review some headphones and suggest best headphones under 5000. I am confused to choose best from Sony,JBL ,boat and soundcore Please didi Rajnit From Bihar
Intel is too late to play the catchup. While they were supposed to do all this, they were busy planning how to diss Apple laptops for not having touch screens, dual displays etc LOL Even Microsoft has shifted focus on ARM PCs now. Quite well deserved IMO
Even AMD Is better than Intel by 30 percent with its AI 300 series. Still, I'd keep an eye out for Intel's Lunar Lake which will be out by fall. It looks promising. As for Snapdragon X Elite, the actual performance is a tad underwhelming. My AMD 7530u powered laptop already has an AMOLED screen and nearly all day battery life.
@@Lionel212001 I’m not a fan of Amoled screens in laptops because of possibility of burn ins due to a lot of static contact. I 5 year of Ryzen 5 Acer Swift 3 works just fine as my secondary computer and I love it for that. I’ve switched to Mac otherwise I’m never switched back. Qualcomm and Microsoft themselves made it clear on stage that their whole goal was to match and surpass the MacBook Air lol Imagine wanting to match a competitor instead of focusing on your stuff and improving it yourself instead of being a response to a competitor :/
@@vedanshchn Asus has inbuilt protection to mitigate burn in issues. It's called Asus OLED care (plus I have an extended 4 years warranty). I am not worried. As for ARM vs x86, it's a moot point as the software ecosystem and operating system is more important. The Snapdragon X Elite is overpriced, has compatibility issues still, and the GPU performance is dismal vis-a-vis the M3. Jim Keller, the CPU architecture wiz, who has worked with Intel, AMD, and APPLE designing their microchips has highlighted in his podcast that the ISA matters little. Even Intel or AMD can match ARM in efficiency if they take the system on chip approach. Intel is already doing that with Lunar lake. So is AMD.
excellent video now like how you have that website that can test that on device processing You should do that to all snapdragon 8 Gen 3 phones and see what it is
I understand that it is a sponsored video but it is a tad unfair to not mention AMD's AI 300 which is better by around 30 percent in AI tasks while consuming less power. So, is the snapdragon X Elite when it comes to AI chops. Intel is the weakest here.
Bullshit. Multiple inconsistencies. 1. At 4:17, GPU0 usage graph is clearly hovering around 50%, but your graph doesn't show it going above 30%, so that's a straight up lie on your part. 2. You also completely fail to disclose that the laptop without the NPU chip has an Intel Iris Xe GPU, whereas the one with the NPU chip has an Intel ARC GPU. The ARC GPU is ~5x strong as the Iris Xe GPU. So, what do we have here? A GPU that is 5 times stronger is still tasked to 50% of its capacity despite the help of an NPU. I conclude that the NPU doesn't really help that much. Admittedly, the computation time was nearly halved, but that would have been with 100% usage on a GPU that was twice as strong too. 50% usage on a GPU that is 5x as strong suggests that the NPU doesn't really do much.
Qualcomm Snapdragon X Ellite new arm based processor is the first AI PC launched by Microsoft and multiple other PC vendors. It is called Copilot ÷AI PC with on device AI. Waiting for India launch
NPU thing is actually a hype nothing else NPU that a 50 LAKH AI server uses and thousands rupees is different. To run AI locally you need a power GPU or . We have NVIDIA N100 etc also have something and AMD too . Goggle has TPU they are industrial workstation chipssets and machinery for them to work. Most of this AI features are cheap colour fancy named click up things that don't take much nor do much. I guarantee 30 Billion parameter model on Intel CPU laptop or desktop whatever NPU first thing I won't load. There's two ways running AI on consumer hardware use the GPU vram or use CPU tpm module and npu new stuff and ram fir storing the model and running things on CPU. There's a catch games or productivity or AI they need parallel processing. Your consumer don't have much of that. Dedicated AI servers they are deshgned different in the first place . What can we do 8 BULLION parameter language model on that consumer CPU skow response nothing i tell you they took a AI model low ehd varient put a lot of GUI on It fancy it fancy names said AI . When we create image if it's server or cloud fine but the CPU . Npu scam it will be super slow if it tries to make thing good or naje fancy funky stuff and animated crap that's the truth. Competing with a 16 GB vram GPU with 30 billion parameter model with latest technologies vs a NPU hype laptop it can never. SILICON M1 , M2, M3 are different they don't shot but look at the stable diffusion performance of 8k image resolution. The speed power compared to i9s that's performance. In AI market currently we have NVIDIA then APPLE & AND fighting for second spot 12 companies then INTEL. Becuz industries don't find good technology. AI is not about hardware itself. NVDIA has CUDA still standing then AI acceleration option. AMD RCOm Apple has it's own AI acceleration What Intel gives they are left behind in the AI race What frameworks and compatibility dies they have AMD has been bas mouthed of thier AI framework compatibility when it comes Intel they don't have anything literally 0. Industries not professional see reason for choosing a intel. Thier GPU xess need to be better thier ray tracing need to work on all games than major titles. They need newer technology and software. Rdna, RCOm beating will taje them ages at the speed they are working Catching up NVDIA seem way Hard.
Pratima didi:AI companies will take our personal data and can misuse
Also Pratima Didi:posted AI generated self portrait of her in Instagram😊
😂
At this point, I would argue that on-device AI is more hype than a real thing. Microsoft's on PC AI known as Windows Recall is essentially an improved full-text search of PC snapshots, and local chatbots are not yet on par with online AI chatbots.
I have a very capable computer and have tested several AI desktop applications such as AnythingLLM, Misty, and LM Studio. I've experimented with various 7-billion and 13-billion parameter LLMs from Hugging Face. In my experience, there's a significant difference in the 'intelligence' shown by these local models compared to online AIs like ChatGPT, Claude, or Mistral, which use models with trillions of parameters. (especially when using different human languages)
On-device LLMs might be suitable for simpler tasks, but to be truly useful, they likely need some form of internet connection to access more advanced AI processing capabilities. The current limitations of local processing power and memory make it challenging for on-device AI to match the performance of cloud-based solutions. As well there was an article that new AI PCs produce 40-50 TOPs however a gaming laptop with an Nvidia RTX 40 series can produce more than 800 AI TOPs.
Padhna jhyau layo conclusion vana
why has the video got so little views?
This is definitely a very nice video for ppl who are looking to purchase laptops in 2024.
A must needed info!
greate explanation!
how ultra core processor compared to intel 13 gen for gaming purpose?
AI at this point is just a more hype than substance. Current AI PCs have anywhere between 10 and 40 TOPS. However, as per Nvidia, PCs need atleast 140 TOPS to fully realise AI. So, spend your money wisely!
I'm not coming for watching ai laptops work, I'm coming for watching yout
Please review some headphones and suggest best headphones under 5000. I am confused to choose best from Sony,JBL ,boat and soundcore
Please didi
Rajnit
From Bihar
Kasari order garne phone haru ???
Intel is too late to play the catchup.
While they were supposed to do all this, they were busy planning how to diss Apple laptops for not having touch screens, dual displays etc LOL
Even Microsoft has shifted focus on ARM PCs now.
Quite well deserved IMO
Even AMD Is better than Intel by 30 percent with its AI 300 series. Still, I'd keep an eye out for Intel's Lunar Lake which will be out by fall. It looks promising. As for Snapdragon X Elite, the actual performance is a tad underwhelming. My AMD 7530u powered laptop already has an AMOLED screen and nearly all day battery life.
@@Lionel212001 I’m not a fan of Amoled screens in laptops because of possibility of burn ins due to a lot of static contact.
I 5 year of Ryzen 5 Acer Swift 3 works just fine as my secondary computer and I love it for that.
I’ve switched to Mac otherwise I’m never switched back. Qualcomm and Microsoft themselves made it clear on stage that their whole goal was to match and surpass the MacBook Air lol
Imagine wanting to match a competitor instead of focusing on your stuff and improving it yourself instead of being a response to a competitor :/
@@vedanshchn Asus has inbuilt protection to mitigate burn in issues. It's called Asus OLED care (plus I have an extended 4 years warranty). I am not worried.
As for ARM vs x86, it's a moot point as the software ecosystem and operating system is more important. The Snapdragon X Elite is overpriced, has compatibility issues still, and the GPU performance is dismal vis-a-vis the M3.
Jim Keller, the CPU architecture wiz, who has worked with Intel, AMD, and APPLE designing their microchips has highlighted in his podcast that the ISA matters little. Even Intel or AMD can match ARM in efficiency if they take the system on chip approach. Intel is already doing that with Lunar lake. So is AMD.
excellent video now like how you have that website that can test that on device processing You should do that to all snapdragon 8 Gen 3 phones and see what it is
good info
didi Acer ko yo model xa AN515-58-52A9
I understand that it is a sponsored video but it is a tad unfair to not mention AMD's AI 300 which is better by around 30 percent in AI tasks while consuming less power. So, is the snapdragon X Elite when it comes to AI chops. Intel is the weakest here.
when is ipad air m2 review coming
Why she speaks the ending in every sentence with a cracking voice..😅😅 she starts loud and gets to the cracking voice..
Its AI voice
Pratima would have been a great teacher to have 😁❤
My grades would drop because I would be staring at her all day!
Hell no. . . . .she can't
infinix note 40 5g review plz
Bullshit. Multiple inconsistencies.
1. At 4:17, GPU0 usage graph is clearly hovering around 50%, but your graph doesn't show it going above 30%, so that's a straight up lie on your part.
2. You also completely fail to disclose that the laptop without the NPU chip has an Intel Iris Xe GPU, whereas the one with the NPU chip has an Intel ARC GPU. The ARC GPU is ~5x strong as the Iris Xe GPU.
So, what do we have here? A GPU that is 5 times stronger is still tasked to 50% of its capacity despite the help of an NPU.
I conclude that the NPU doesn't really help that much.
Admittedly, the computation time was nearly halved, but that would have been with 100% usage on a GPU that was twice as strong too. 50% usage on a GPU that is 5x as strong suggests that the NPU doesn't really do much.
Sorry i am staring your abs ..
Wow looks like you are doing gym workouts regular. You have absolutely great abbs-abdominal muscle separations ❤.
Indeed 👍🏻
Qualcomm Snapdragon X Ellite new arm based processor is the first AI PC launched by Microsoft and multiple other PC vendors. It is called Copilot ÷AI PC with on device AI. Waiting for India launch
14 seconds ago is crazy😂
First here
Intels using gimmicks to try win back marketshare
She is Bengali..😊
NPU thing is actually a hype nothing else NPU that a 50 LAKH AI server uses and thousands rupees is different.
To run AI locally you need a power GPU or .
We have NVIDIA N100 etc also have something and AMD too .
Goggle has TPU they are industrial workstation chipssets and machinery for them to work.
Most of this AI features are cheap colour fancy named click up things that don't take much nor do much.
I guarantee 30 Billion parameter model on Intel CPU laptop or desktop whatever NPU first thing I won't load.
There's two ways running AI on consumer hardware use the GPU vram or use CPU tpm module and npu new stuff and ram fir storing the model and running things on CPU.
There's a catch games or productivity or AI they need parallel processing.
Your consumer don't have much of that.
Dedicated AI servers they are deshgned different in the first place .
What can we do 8 BULLION parameter language model on that consumer CPU skow response nothing i tell you they took a AI model low ehd varient put a lot of GUI on It fancy it fancy names said AI .
When we create image if it's server or cloud fine but the CPU .
Npu scam it will be super slow if it tries to make thing good or naje fancy funky stuff and animated crap that's the truth.
Competing with a 16 GB vram GPU with 30 billion parameter model with latest technologies vs a NPU hype laptop it can never.
SILICON M1 , M2, M3 are different they don't shot but look at the stable diffusion performance of 8k image resolution.
The speed power compared to i9s that's performance.
In AI market currently we have NVIDIA then APPLE & AND fighting for second spot 12 companies then INTEL.
Becuz industries don't find good technology.
AI is not about hardware itself.
NVDIA has CUDA still standing then AI acceleration option.
AMD RCOm
Apple has it's own AI acceleration
What Intel gives they are left behind in the AI race
What frameworks and compatibility dies they have AMD has been bas mouthed of thier AI framework compatibility when it comes Intel they don't have anything literally 0.
Industries not professional see reason for choosing a intel.
Thier GPU xess need to be better thier ray tracing need to work on all games than major titles. They need newer technology and software.
Rdna, RCOm beating will taje them ages at the speed they are working
Catching up NVDIA seem way Hard.
Intel is dying
Amd is rising✅
What the duck.
no
No
🤜🤛💛
❤
❤❤
I love you
Another repeat Madam
Hey pretty i love you so much
First
got a boyfriend??
Her reply: Got a slap?