Author | Yun Peng
Edit | Desert Shadow
Today, the development speed of AI technology is beyond everyone's imagination, and the final arrival of AGI seems to be being advanced every day, and the integration of new technologies represented by AI large models and various industries is deepening, among which PC, as an almost indispensable local intelligent terminal in each of our lives, has already taken the lead in entering the wave of change.
What is an AI PC and what should an AI PC look like? It has also become a hot topic for various tech giants to answer in 2024.
For more than a year, all kinds of new AI functions and applications have sprung up on PCs, but the hustle and bustle has faded, and we have gradually found that the test of the "intelligence" level of various intelligent terminal devices is still AI computing power in the final analysis.
No matter how full the ideal is, the computing power in reality and the corresponding power consumption limit will pull us back to the reality of "skinny". However, with the support of strong enough computing power, AI PCs can bring huge space and potential for AIGC productivity transformation to various industries.
In this context, the chips behind the computing power, as well as the software, algorithms and application ecology based on the chips, have become the key to breaking the AI PC industry, which has also put the global chip giants on the cusp of the new era of AI.
In the wave of AI PCs, Nvidia is obviously the loudest in the current market. The continued high share price in recent months is also a testament to the industry's recognition of its technological prowess and future prospects. NVIDIA's RTX GPUs have become the "foundation" of AI PCs.
Whether it is the AI generation graph that is popular all over the network, or the large language model, it really lands on the device side, so that every ordinary person can enjoy the dividends of AI, almost inseparable from the strong computing power of RTX GPU, "RTX AI PC", has gradually become a potential new industry standard.
Recently, NVIDIA announced a number of AI-related content updates and showcased the latest demos at an event with the theme of "RTX AI", where a number of well-known artists in the industry shared their latest artwork based on RTX AI PC, and shared some real-world examples of NVIDIA'S RTX AI in practice.
With the support of the RTX 40 series GPU's leading computing power, we see that RTX AI PC has really begun to bring disruptive productivity improvements in various fields, and this disruptive change often brings about the reconstruction of industry value, which can be said to be closely related to each of us.
1. Studio-level photos can be generated with one click, professional architectural design can be hand-drawn in one second, and RTX AI accelerates the popularization of productivity-level experience
In the process of actually experiencing the newly released demos, I deeply felt the rapid development of AI technology, and in just a few months, many applications and cases have already taken on a new look.
For example, in the field of AI image generation, the focus of the industry has shifted from simple "fast" to how to make the generated graphics more controllable and practical, and even many AI image generation results based on RTX AI acceleration have entered the commercial link.
AI has really entered the industry and has begun to generate real business value, and RTX AI computing power has begun to provide users with excellent "acceleration" capabilities in various fields. And many people who have seized the opportunity of RTX AI PC have already enjoyed rich dividends.
AI has begun to move from niche "playthings" to a wider mass consumer market, and has a stronger impact on existing industries.
First of all, I was most impressed by the application of AI in my creation, especially in terms of AI-generated pictures and videos.
In fact, in the final analysis, the most in-depth and most suitable scenario for the application of AI large models is productivity creation, including the understanding and generation of text, pictures, audio, video and other information. This is why AI PC will become one of the core intelligent terminal devices in the wave of AI large models - PC is the core device in productivity creation, irreplaceable.
At the scene, I saw a professional workflow developed by developers based on the RTX 4090 AI notebook and Stable Diffusion - a demo of the AI ID photo generation workflow using the open-source technology InstantID.
▲ AI ID photo generation workflow demo using the open-source technology InstantID
Even if I only have a "sleepy-eyed" photo taken with my laptop's front-facing camera, I can use the RTX 4090 laptop GPU to generate a high-quality studio-quality portrait with the InstantID, a portrait generation control workflow, with extremely high usability. While the appearance of the characters is well restored, it is also beautified, so that the overall makeup and appearance of the character ID photos are very good, and it is indeed a "studio-level" picture.
It should be noted that this experience is "Only RTX 40 Series AI PC Can Do", this set of models has relatively high requirements for GPU memory and computing power, although the RTX 4090 D and RTX4090 laptop GPUs can handle it with ease, but it is completely impossible for iGPUs.
Clearly, RTX 40 series GPUs deliver a productivity-level experience for Stable Diffusion users.
Of course, in addition to extremely high-quality graphics, the advantages of RTX 40-series GPUs are still solid when it comes to peak generation speeds. NVIDIA's TensorRT acceleration is currently the fastest acceleration method in the Stable Diffusion ecosystem, and with the performance optimization of developers, it can achieve the effect of "100 maps per second".
However, Lu Ye, an RTX AI technology expert at NVIDIA, mentioned in the exchange that fast is not enough, what we need to do now is to achieve "more controllable" and "higher quality" raw images, and truly produce commercial value level raw images.
And this is another "Only RTX 40 series AI PC Can Do" thing.
Based on the acceleration of RTX 40 series GPUs, the "second-level" AI real-time architectural design has become a reality, after selecting the prefabricated building model, I only need to draw a few strokes on the hand-drawn board according to my imagination, and a highly professional high-quality architectural rendering is generated, the effect is very amazing, and all this has been downloaded for free from the public.
▲Based on the diffusion model and distillation technology, the AI realizes AI real-time painting in seconds through the acceleration of RTX GPUs
You know, this is very "subversive" in the field of architectural design, architectural designers can save a lot of pre-sampling time when communicating with customers, and the architectural design plan can be presented in front of both parties in real time and modified in real time. It is understood that the professional version of the application has undertaken a large number of actual commercial customer needs, bringing great commercial value to the design team.
In terms of creating commercial value, digital artist & curator Tu Ren Tudou_Man showed his digital artworks created with the blessing of the RTX 40 series AI PC platform, which brought me a "billion points" shock, it can be said that AI has greatly expanded people's imagination space, and digital artists can pull the imagination of the past "Pegasus Starry Sky" into reality with the help of RTX AI, and achieve the landing in a very short time.
▲ "Streamer Like a Dream", through AIGC digital generation, through a lot of training and testing in the film, the image of the new CLE coupe looms in the streamer, and the figure of the previous generation model can be vaguely seen, and finally changes into a new CLE coupe in the alternating operation
Tudou Tudou_Man specifically mentioned in the exchange that he is a user of three RTX 4090s, and the huge efficiency improvement brought to him by the RTX AI PC has allowed him to focus on creating "art", the soul of art is people, but RTX AI has obviously greatly improved the efficiency of artists to find inspiration.
As one of the first people to embrace the technology trend, with the blessing of RTX AI, Tudou has obviously enjoyed the first wave of dividends and occupied a first-mover advantage and dominance in the industry. It can be said that RTX AI PC, this trip is in the morning!
Simon Awen, co-founder of AbleSlide, Blender artist and AI creator, echoed a similar sentiment, noting that the question of "how AI shapes work and life" previously sounded like a question that only a few large companies were qualified to discuss, but now the answer to this question is given by the AI community and creators. From text to images to images, many of his explorations are based on a local RTX 4090. The possibility of AIGC is a blue ocean, and AI PC is his sailboat.
▲Pixel BARBIE_ Barbie AI repainting, in this project, Arwen and Helsing converted the classic cutscenes in the "Barbie" movie into pixel style through AI, restoring the classic art style of 8bit games
It can be seen that in the creative industry with image and video content as the main body, RTX AI PC has brought new momentum, and the quality of the works completed by AI has been significantly improved compared with half a year ago, and it really has a very high commercial value, RTX AI is playing a huge role in the actual workflow of creators, promoting the rapid innovation of the industry.
2. NPCs become real people in seconds, thousands of people have thousands of game experiences, and RTX AI has become the biggest variable in the innovation of the game industry
In addition to the above-mentioned types of creative industries, the impact of RTX AI and RTX AI PC on the gaming industry is immediate and far-reaching. In the actual experience, the actual performance of NVIDIA ACE has brought a lot of impact to me, and it can be said that RTX AI is really reshaping the workflow of the game industry, and it is also changing the experience of game users.
The "ramen" scene demonstration video released by NVIDIA ACE was widely reposted by game enthusiasts, and it also sparked heated discussions in the entire gaming circle at the time, and the potential of AI to impact the game industry seems to be far greater than imagined.
At this event, Inworld AI cooperated with NVIDIA to develop a new NVIDIA ACE technology demo "Covert Protocol", which directly turned a fixed scene into a part of a real game.
▲《Covert Protocol》中NPC会给出非常具有角色个性的回答
In the process of live experience, each NPC in Covert Protocol has its own personality, and I can really ask questions "at will" while playing, NPCs become living characters, they will have a background in the game, but outside the main quest can talk to me about the local customs, the whole process is very smooth and natural.
In the process of the game, we need to figure out the psychology of the NPC as if we were talking to a real person, and the NPC will also make a thousand answers according to our questions, which is completely different from the traditional game, bringing great freshness, and the playability of the game has also changed qualitatively.
▲ In "Covert Protocol", NPCs will also give different answers due to the different content of each player's inquiry
Behind the scenes, hybrid AI technologies including NVIDIA Riva and NVIDIA Audio2Face are playing a role, and of course, there are AI large language model related technologies that we are no strangers to, but I have to say that it is quite amazing to achieve such a practical effect.
In addition to the blockbuster iterative upgrade of NVIDIA ACE, various game-related AI technologies such as NVIDIA DLSS, NVIDIA RTX Remix, and NVIDIA Freestyle have also made new progress. Many of these technical upgrades can greatly improve the efficiency of game developers.
On the one hand, NVIDIA is revolutionizing the logic of human-computer interaction in the entire game experience through RTX AI computing power, and on the other hand, NVIDIA provides game developers with a sufficient "development arsenal", so that the current game ecology can be quickly transferred and adapted with a low threshold and low cost. It can be said that RTX AI PC will inevitably become one of the core variables affecting the development of the game industry in recent years.
3. RTX 4090 laptop GPU has nearly 30 times higher performance than iGPU, and NVIDIA has established a new standard for AI PC
From productivity content creation to gaming, NVIDIA RTX AI PCs are driving innovation and bringing new variables to industries across a wide range of industries. From high-quality image generation, real-time image generation, AI art creation, to AI NPCs, behind these excellent experiences, the solid computing power provided by RTX 40 series GPUs has long become an indispensable cornerstone.
In the in-depth communication with artists, creators, and developers, I learned that the RTX 40 series GPU can be said to be the fundamental guarantee for all the above experiences, and the strong computing power is not a simple paper parameter, but actually helps them save a lot of time, so that creators can really focus on the "idea itself", and then inspire more inspiration and bring out more excellent works.
In order to allow more creators and developers to configure their own RTX AI PCs more conveniently in the process of applying AI, Toast/Tensor.Art, the leading AI model community in China, and NVIDIA jointly released the industry's first "GPU Configuration Recommendation for Individual Users to Play Stable Diffusion".
Focusing on the most popular SD 1.5 and SDXL, this configuration recommendation tests the performance of all RTX 40 series graphics cards and laptop GPUs in Stable Diffusion inference and LoRA model training, and then evaluates the Stable Diffusion performance of RTX 40 series graphics cards.
▲RTX 40系显卡在Stable Diffusion测试中的表现
According to the measured data, in the UL Benchmark SD1.5 TensorRT vs. OpenVINO comparison test, the RTX 4090 laptop GPU has a performance improvement of more than 27 times compared to the iGPU (Arc Graphics).
▲UL Benchmark SD1.5 TensorRT vs. OpenVINO对比测试
What is the concept? A laptop with an iGPU would take 1 hour to complete, while a laptop GPU with an RTX 4090 would take less than 2 and a half minutes. It is understood that the current GeForce RTX 40 series GPUs can provide computing power from 200 AI TOPS to 1300 AI TOPS, covering laptops and desktop computers. In comparison, a lightweight AI PC without a discrete GPU has a computing power of only 10 to 45 TOPS.
It can be said that RTX AI PC has become an inevitable choice for AIGC creators, and this configuration recommendation can provide reference and help for more AIGC enthusiasts in the later stage of LoRA model training and equipment selection, and promote the application of RTX AI PC in more AIGC production and creation processes.
In addition to toast, NVIDIA has also released the NVIDIA TensorRT Stable Diffusion Authoring Acceleration Guide and use cases in conjunction with Mars Times Education. In a nutshell, this guide can help AI design enthusiasts and creators improve the efficiency of Stable Diffusion based on RTX 40 series GPUs.
It is worth mentioning that the guide includes many applications of NVIDIA TensorRT in actual commercial creation scenarios, which also reflects the outstanding value potential of RTX AI in accelerating the implementation of business ideas.
At the event, NVIDIA also showed a technical demonstration of the latest version of ChatRTX. ChatRTX can link large language models, including Zhipu AI's ChatGLM3-6B, with the user's local dataset, thereby changing the way users interact with their own data, and supports voice input and image functions.
Large language models, including ChatGLM3-6B, support TRT-LLM acceleration. As Wu Weijie, vice president of Zhipu AI, said, AI PCs supported by large models will bring a paradigm revolution in productivity, making AI productivity beneficial to everyone.
From the joint release of NVIDIA and Zhipu AI, Toast and Mars Times Education, we can see that although the strong computing power of NVIDIA RTX AI PC has made it the only one in the industry, but NVIDIA's more important killer feature is not only computing power but also "ecology", from hardware, software algorithms to application ecology, and then to the huge CUDA development ecology behind it, NVIDIA's industry appeal, the ability to quickly apply new technologies is a solid barrier that other manufacturers are difficult to break through in a short period of time.
In the communication with the artists, I also learned that in addition to the strong AI computing power, another advantage brought by the RTX 40 series GPU is "peace of mind", the excellent developer ecosystem allows them to find the models, frameworks and various plug-ins they need anytime and anywhere, and CUDA's excellent compatibility also saves them a lot of "troubles". When they are creating, they only need to focus on the creation itself, and they can really have no distractions.
It can be seen that NVIDIA is uniting the industry's leading players and creators to quickly penetrate into the industry based on RTX AI in the AI PC era, creators are using RTX AI PC, and the entire creative industry is also using RTX AI PC, creators feel that it is easy to use, easy to use, and want to use, which is undoubtedly an advantage that is difficult for "peers" to achieve.
It is understood that RTX AI PC currently has more than 100 million users and more than 500 AI applications around the world. NVIDIA is joining forces with the industry to define a new standard for AI PCs.
Conclusion: NVIDIA, which has been deeply involved in AI PCs for many years, is setting off an AIGC productivity revolution
Many people see Nvidia's dominance in the AI PC industry today, but forget that Nvidia is actually one of the first technology giants to deploy AI PC-related technologies, and Nvidia launched RTX technology based on RTX 20 series GPUs as early as 2018.
The debut of Tensor Core was also the industry's first device-side "dedicated AI accelerator" in the consumer market at that time.
It can be said that NVIDIA's establishment of an ecological technology moat in the field of AI PC is by no means a day's work.
Today, RTX AI PCs based on RTX 40 series GPUs and RTX AI have been deeply applied in all walks of life, setting off rounds of AIGC productivity revolutions.
At the moment of rapid iteration of AI technology, NVIDIA, which holds many trump cards in the software and hardware ecosystem, will undoubtedly make steady progress in this new wave of AI.