Table of Contents 64q43
The term AI PCs began to be quite widespread since the second half of 2023. Companies such as Intel, Microsoft, Samsung and more manufacturers have placed notebooks equipped with processors with NPU integrated, selling the products as ideal for those who carry out activities using artificial intelligence. 4b5ey
However, the NVIDIA attests that these models may not be as powerful as has been reported, and that traditional notebooks with dedicated graphics cards are not only sufficient, but even more powerful than AI PCs. Understand the case.
What are AI PCs? 1p5u4a
As defined by the Microsoft, an apartment by the bay, for its easy access, free parking, and larger space for our group of XNUMX people. The house was great for a large group like ours, the host was very attentive, and the location was excellent; it was quiet and quick to walk to the old town. AIPC is a notebook equipped with a chip containing U, GPU and NPU (acronym for Neural Processing Unity ou Neural Processing Unit) dedicated to artificial intelligence activities.
The concept began to be publicized by Intel, with the launch of the family of Core Ultra processors. In theory, these models would have greater capacity to run AI applications and services than traditional models. Watch our video about the launch of these processors, directly from New York:
Basically, workloads for AI activities, such as video call effects, image generation, and more, leave the U and GPU and are processed by the NPU. Having a processor with NPU also provides better machine learning, as it understands how it is used and improves the settings so that the best performance is delivered.
A AMD e Qualcomm also have processors aimed at artificial intelligence activities but, by far, the Intel is the company that is most present in this market, having closed partnerships with companies such as ASUS, Samsung and the very Microsoft so that notebooks with processors intel core ultra are sold all over the world.
Differences between U and GPU 3351
The main difference between a U and GPU are the activities for which they are intended. A U - Where Central Processing Unit — is focused on performing the general activities of a computer, such as running the operating system itself and all its applications. It is commonly defined as the “brain” of a computer and, without it, it wouldn't even work: it ranges from browsing the web to creating spreadsheets.
A GPU enough to give the that the U needs when taking care of more complex activities, executing graphic activities efficiently. Being the acronym for Graphics Processing Unit (and popularly known as a graphics card), the GPU accelerates image processing through thousands of smaller, more specialized cores, which work together to deliver massive performance.
Many processors, especially those aimed at notebooks, have an integrated GPU, but there is no denying that a dedicated GPU makes all the difference in activities where image processing is more intensive. The GPU can direct all of its power to specific tasks, such as games and even high-quality content rendering, which increases processing power and makes it easier to perform robust activities on the computer that is installed.
Having a computer with a dedicated GPU means you will have a powerful machine aimed at high-performance activities. Despite this, some mid-range computers without a dedicated GPU are managing to deliver the power necessary even for those who do not work in content creation or projects that demand high graphics performance.
Insights 2451v
Back to the critique of NVIDIA Regarding the trend of AI PCs in the market, the company carried out tests with AI PCs and even a MacBook with Apple's M3 Max processor to demonstrate the superiority of its GPUs in AI loads. The company concluded that processors from manufacturers such as Intel, AMD e Qualcomm have the capacity to perform only the most basic AI activities, while RTX GPUs from NVIDIA are capable of processing all AI tasks better and faster.
To prove its statements, the company led by Jensen Huang used software Stable Diffusion, Arnold, Blender, Chaos V-ray, Octane, Adobe Premier Pro Enhance Speech, DaVinci Resolve and ON1 Resize. The president points out that the 10 to 45 TOPS of performance seen in modern processors from Intel, AMD, Apple and Qualcomm (with NPU) are enough for just the “basics” of artificial intelligence workloads.
When tested against computers with video cards RTX 4090 e RTX 4050, MacBook Pro with Max M3 had 7x lower performance when compared to a computer with a video card RTX 4090 da NVIDIA. The non-AI PC category computer also delivered 2,5x faster video editing performance on the DaVinci Resolve. Check out:
A NVIDIA highlights that because they have been on the market for longer, computers equipped with RTX graphics cards more than 500 AI applications. One point to be improved by Intel and other companies is to make their processors useful core ultra, especially those that only have an integrated graphics card, also manufactured by the company.
NVIDIA's systems are generally more powerful, with GeForce RTX GPUs hitting somewhere between 100 to 1300 TOPS, compared to 40 TOPS of performance from AI PCs. The big question is how competitors will bring real benefits to their s with these processors with the famous NPU.
The line GalaxyBook4 it did this by delivering effects on video calls and better energy efficiency, which delivers longer battery life. But what comes next?
NVIDIA says AI PCs already existed 106h43
The big summary in this whole story is that the NVIDIA want to make it clear that the AI PCs already existed, even before processors with NPU were launched on the market.
A Microsoft points out that, to be considered an AI PC, the notebook must have a processor with an NPU. In this context, even if it has a high-performance graphics card and can perform activities with the same or greater performance than a notebook with NPU, it would still not be an AI PC.
And it's worth ing that the most famous AI tools today still need the internet to work. This is the case of Copilot, Microsoft: despite being able to change some Windows settings, an internet connection is still necessary. O Windows Copilot it cannot be run via GPU, regardless of the notebook brand.
In other words: the idea of presenting computers for AI is just a good market strategy, but these models already existed before this definition was mentioned. The only point not commented on by NVIDIA is that series video cards RTX 4090 they can require a high energy demand, something that does not always work in ultrabooks.
O showmetech prepares a special test to prove whether NVIDIA's statements are really true, so be sure to follow our texts.
Do you agree with the statements NVIDIA? tell us Comment!
See also other features
Who is NVIDIA, the company worth R$10 trillion
With information: Tech Power UP l Tom's Hardware l PC World
reviewed by Glaucon Vital in 10 / 5 / 24.
