Shulou(Shulou.com)12/24 Report--

The hottest topic in the tech world today is generative AI. Its emergence and popularity are gradually changing the way we work, live and play. CTOnews.com also has a profound understanding of this. For example, in the past, editors needed pictures to write manuscripts, and it was a headache to retrieve pictures on the Internet. To consider copyright, clarity, content fit, etc., it often takes a long time to find a satisfactory picture.

However, after having a generative AI model that can generate text diagrams, everything is much simpler. You only need to enter a paragraph of text, you can quickly get a more satisfactory picture, and the work efficiency has been improved too much.

This makes the editor sigh that AIGC is really a super artifact that subverts our productivity. Such a powerful technology must be deeply integrated with our commonly used production tools in order to fully release its value. When it comes to production tools, the most important thing at this stage is PC.

In other words, PC should be the main battlefield for AIGC to release the energy of productivity change.

The good news is that Intel, as a leader in the semiconductor and PC industry, has actively landed the generative AI on the PC terminal, and has made remarkable progress.

Usually when it comes to AIGC, when it comes to large language models, you may first think of all kinds of large professional GPU. That's because many AIGC large model services are deployed in the cloud and require a lot of computing power from large GPU. Even so, the generation will be slow or even fail during the peak service hours, so I believe many students have such a need: can we let the AIGC model run locally in PC, or even on the slim books that we often use when we go out to work?

Intel's answer is YSE.

At present, a variety of Intel client chips, represented by the 12th and 13th generation Intel Core processors and Intel Ruixian A-Series graphics cards, can provide strong performance to meet the high-computing needs of generative AI. At the software level, Intel further promotes the landing of emerging generative AI scenarios on personal computers through the construction of software ecology and model optimization, covering a wide range of lightweight books, omnipotent books, games and so on.

For example, for the general large language model, Intel enables the large language model with up to 16 billion parameters to run on personal computers with 16GB and above memory capacity through the BigDL-LLM framework through the acceleration of the 13th generation Intel Core processor XPU, low-bit quantification and other software optimization.

Here, CTOnews.com uses actual tests to show you the effect of the big language model running on Intel's slim books. The editor chose a slim book certified by Intel's Evo platform: Asustek dawning Air, which carries Intel 13th-generation Core i7-1355U processor and 16GB LPDDR5 memory.

Then the editor installed the big language model Demo launched by Intel on the Asustek dawning Air. This Demo integrates three major language models, including ChatGLM2, LLaMA2, and StarCoder. They are all optimized through Intel's corpus, and the main way of optimization is to quantify these large language models to reduce their demand for local hardware resources.

During the test, the editor disconnected the laptop from the network and ran it locally.

First of all, ask a more common human question under the chat assistant function, such as "my friend borrowed 10,000 yuan from me, but I am also short of money recently. How can I politely refuse?" The big model quickly gave the answer: first latency is 1490.1.ms, and the actual waiting time is only about three to five seconds. If we look at the content of the answer, apart from the problem with the second suggestion, it is quite appropriate as a whole.

Then the editor tested its copywriting ability, switched to "story creation" mode, and then asked, "the company's orientation party needs me to be the host, please help me write an opening speech of about 500 words." Sure enough, it wrote the opening remarks very quickly, and the copywriting was easy to read and appropriate. If you are thinking and editing by yourself, it will take a long time to use the AI model on PC and it will be done in minutes.

When the Intel big language model was writing, the editor took a look at the scheduling of Asustek dawn Air performance resources. Intel 13th generation Core i7-1355U processor occupancy reached 100%, memory footprint reached 9.6GB (62%), Xe core occupancy also reached 12%. It seems that the operation is indeed carried out locally. With the continuous optimization of Intel and the improvement of the computing power of the 13th generation Cooley processor, it is indeed possible to achieve the landing of AIGC on a thin and lightweight book.

The following editor makes a test in combination with many scenarios that workers often encounter, that is, to write an outline for an activity or project. For example, the company needs to hold a small product communication meeting in Zhongshan Park, with a funding of 15000 yuan, and needs to invite 50 users to attend, including product experience, communication and interaction, and ask the big model to help work out a process outline.

For this very practical problem, the big language model Demo also gives a more detailed outline, the editor looked at it, the specific feasibility is also good, according to the actual situation to change, it can be used. You know, this process is run locally in Asustek dawning Air, a lightweight local model, which is not limited by networking conditions, can be used anytime, anywhere, and is much more efficient than searching various templates online, just like the natural function of PC, which is a real increase in productivity.

In addition to this general large language model, as the editor said at the beginning, we often have the need to draw pictures in our daily work, and the Asustek Dawn Air can also run easily for some graphic models.

For example, the well-known open source image generation model Stable Diffusion, Intel enabled OpenVINO acceleration, they developed an AI framework, through a single line of code installation, can speed up the operation of the PyTorch model. With Stable Diffusion's WebUI, you can run Stable Diffusion Automatic1111 on both the Rui Torch integrated graphics card and the Arc stand-alone graphics card.

Through the actual test, we can see the performance effect of Stable Diffusion on the integrated graphics card on the Asustek dawning Air lightweight book. The powerful computing power of the 96EU version of Intel Ruiju Xe graphics card can support FP16 precision models running on Stable Diffusion software to quickly generate high-quality pictures. The editor asked it to generate a picture of "the Man on the phone". On Asustek's dawn Air, it only took 25.1 seconds to "release the film".

During the generation process, CTOnews.com also saw through the performance Explorer that the occupancy of GPU was 100%, while that of CPU was 15%, which shows that this image is indeed rendered locally using GPU.

In the past, it would be hard to imagine that lightweight could have such performance, but with the improvement in performance and power consumption ratio of the 13th generation Core processor, and the significant improvement in FP16 and FP32 floating-point performance of Ruiju Xe Graphics (96EU), as well as the addition of INT8 integer computing power, all these have greatly enhanced GPU's overall AI graphics computing power. This is an important reason why a slim book like Asustek dawning Air can run Stable Diffusion well on the local side.

It is worth mentioning that in Intel's next-generation core processor Meteor Lake, GPU kernel performance will be further improved, with 8 Xe GPU cores and 128 rendering engines, 8 hardware light tracking units, asynchronous copy of Arc graphics cards, out-of-order sampling and other functions, and DX12U will also be optimized.

Moreover, Intel has added an integrated NPU unit to Meteor Lake to achieve more efficient AI computing, which includes two neural computing engines that can better support content including generative AI, computer vision, image enhancement and collaborative AI.

At the same time, in addition to NPU,CPU and GPU can also perform AI operations, different scenarios will use different AI units to deal with, coordinate with each other, so that the overall energy consumption can be up to 8 times higher than that of the previous generation. Therefore, the performance of lightweight books with Meteor Lake processors in local AIGC creation will be even more anticipated in the future.

Through the above introduction and testing, I believe that everyone has a new understanding of running AIGC on notebooks, and at the same time, they will also have a new understanding of the productivity of PC notebooks. Under the enabling of AIGC, the nature of the future PC will even evolve, which will not only be limited to the production of "tools", but also have the same intelligence and creativity as human consciousness. From this point of view, Intel's efforts on the road of AI PC are undoubtedly more meaningful.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.