Shulou(Shulou.com)11/24 Report--

Since ChatGPT came out of the circle, the generative AI model has set off a new wave of technology all over the world. As the most important underlying technology in the digital future, AI will inevitably bring subversive changes to the life and production of human society.

But at this stage, AI still has a long way to go to really change the world, and there are still many challenges to promote the large-scale expansion and application of generative AI.

For example, today's generative AI large model products are mainly deployed in the cloud. Traditionally, the cloud can provide far more computing power and storage than the terminal, but the reality is not absolute. We often encounter problems such as slow response and failed generation when using these cloud large model products, because even cloud servers cannot withstand extreme computing requirements during peak usage periods.

Moreover, for service providers, the cost of each search query of spanning AI is 10 times that of traditional search methods. At present, more than 10 billion search queries are generated every day, so the scale of the load and cost of cloud computing is unimaginable.

In this case, the development of generative AI to the terminal side is particularly important.

Although the computing power provided by a single terminal is obviously not comparable to that of a cloud server, if you invoke the computing power of billions or even tens of billions of terminals around the world, it can undoubtedly greatly divert the pressure from the cloud. This is the basic logic that AI wants to develop to the terminal side.

More importantly, with the efforts of the semiconductor industry, the AI performance and computing power of terminals have also improved by leaps and bounds in recent years. Take the productivity of PC as an example, the industry leader Intel has made outstanding contributions to the landing of generative AI on PC terminals.

For example, in our traditional understanding, to run a multimodal AI large model, it is necessary to have a professional graphics card with super large graphics memory to complete a large number of AI parallel operations, so can it also support the smooth operation of the AI large model for lightweight notebooks or consumer desktops? This is actually one of the directions of Intel's AIGC efforts on the terminal side.

At present, Intel's 12th and 13th generation cool processors and Intel sharp graphics cards can meet the high-speed computing needs of AIGC on the local side of PC.

For sharp graphics cards, first of all, Intel continues to enhance its own performance experience. Since its launch, Intel Ruixuan graphics cards have released more than 20 driver updates, and earlier this year, Intel also improved the performance of Ruixuan graphics cards when running a series of popular DX11 games through the release of Game On drivers, giving the game frame rate an average frame rate boost of about 19% and an average 99th Percentile frame rate fluency of about 20%.

In May this year, Intel also showed an example of using generative AI to speed up the creation of Vincent diagrams. Based on the Intel OpenVINO,AI drawing open source model, Stable Diffusion can use the open source image editing software GIMP to run smoothly on Intel Ruixuan A750, A770 and other graphics cards. You only need to enter simple text to intelligently realize creative drawing, which is very practical for picture creators.

In August, Intel showed a solution based on the OpenVINO PyTorch back-end, using Pytorch API to make the community open source model work well on Intel's client processors, integrated graphics cards, stand-alone graphics cards and dedicated AI engines.

For example, for the open source image generation model Stable Diffusion, Intel enabled OpenVINO acceleration, and they developed an AI framework that can speed up the PyTorch model through a single line of code installation. 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.

Among them, it can be said that it is of great significance to let Stable Diffusion run on a slim book with integrated graphics cards.

Here, for example, CTOnews.com chose a slim version to test. This product is the Asustek dawning Air certified by Intel Evo platform, which is equipped with Intel 13th-generation Core i7-1355U processor, Ruiju Xe integrated graphics card and 16GB LPDDR5 memory.

You can see the performance effect of Stable Diffusion on Asustek dawning Air integrated graphics card. 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 Stable Diffusion to generate a "puppy with black ears". It only took about ten seconds for Asustek to generate Air at dawn. This is a picture of 512 × 512. If you want to draw better, you can adjust the parameters yourself.

Another example is to ask Stable Diffusion to generate a "photo of food with a bunch of pancakes", which can also be easily generated on Asustek Dawn Air and is being implemented.

We usually think that the lightweight book is not suitable for AI work, because its configuration is relatively low. But through the demonstration of Stable Diffusion above, I believe you can see that Evo high-performance lightweight book can indeed be competent for these simple AIGC image creation work.

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.

In addition, if you pursue better performance, you can also choose Intel Ruixuan Arc unique equipment, in the Arc alone on the Stable Diffusion, the speed will be much faster. For example, earlier this year Intel also demonstrated the effect of running Stable Diffusion "picture" and "3D digital reconstruction of character movements" on a machine with i7-13700K CPU + Arc A770, which is very fast.

In a word, for PC in the future, the so-called performance will not only be limited to the traditional parameters such as the number of cores, threads and main frequency of the processor, but also depend on whether the computing and creative ability of AI is strong. In other words, the era of AI defining chips is coming, and AI PC will really help us achieve the liberation of productivity. Therefore, Intel's efforts to achieve terminal-side AIGC are undoubtedly of great significance, they provide users with a more intelligent and efficient mobile computing experience, and promote the development and application of artificial intelligence technology to a new stage of terminal and cloud collaboration.

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.