Introduction
COMPUTEX 2024 put my AI series on break but I’m still working on a State of AI article which will look into the functional aspect of AI currently. That said, I’ll be borrowing a segment from that draft where in this monthly update we’ll check what’s happening on the side of AI that most of us will actually find useful. From gamers, creators to remote work, as the saying goes: “there’s an app for that” but did you know those apps are AI-powered?
Many modern applications now feature AI in one way or another and in this article we’ll put a spotlight on some of them as well as how your hardware of choice will affect how it performs which ultimately dictates how you work. But before we dive in, here’s a refresher for those not familiar with an intro to AI.
What is AI?
Despite all of us being drowned in AI marketing in the past few months, we’re still years away from a potential Skynet AI being developed somewhere. As of today, we use AI as a catch-all term for some specific methods of computing, the most prevalent of which are large-language models with chatbots and generative AI ala Midjourney, Stable Diffusion, etc.
Many of those who may be versed in computer science or at least in IT would assume current AI is just a long “if,else” statement turned into an algorithm and that’s not technically wrong, but if we look at the core logic and just how big the lines of code needed for this “just a bunch of if/else statements” would be, its certainly a massive oversimplification.
AI currently can be defined as a smarter way of computing, with AI computing being able to recognize patterns, what we define as “training”. The trained model can then be applied to applications where they can now recognize a person sitting on a chair, have the application isolate the person based on the trained data and then blur the background. That’s simple background removal for you.
With data recognition comes data generation or as we generally call it, generative AI. There is an abundance of generative AI applications out right now, many of which have become a giant, transformative tool for many people primarily ChatGPT and Stable Diffusion.
With that, you’ll be saying the word “AI PC” being thrown around a lot lately. It’s a marketing term but with COMPUTEX and Microsoft putting it up on a pedestal, you’ll definitely be seeing it more. Still, if you’re PC has a CPU, a GPU and optionally an NPU, then you already have a functional AI PC. And with that AI PC you have access to a bunch of tools that will aid you in daily life.
But don’t you need an NPU for AI?
Not really. An NPU is a different subject altogether but to quickly summarize, it’s a dedicated, fixed-function hardware meant for accelerating processing of AI tasks with very efficient power draw. And that’s great for laptops or lower power devices but for tasks that take time to finish and require a lot of power, the GPU is just so much more powerful.
GeForce RTX for AI, Not Just Gaming
Did you know that DLSS is one of the most widely used forms of AI in modern computing today. And it didn’t start out in 2023, not in 2022, not even in 2020. DLSS came out in 2019 alongside the GeForce RTX 20-series of graphics cards. The debut of real-time ray tracing and deep-learning for gaming.
NVIDIA GPUs have been the core foundation in driving AI forward and GeForce RTX graphics cards receive all the benefits for its users. Since 2019, deep-learning and machine-learning has pushed AI exponentially and now there’s a lot of applications that take advantage of AI to deliver faster, better and more engaging experiences across multiple categories of computing.
And now we proceed to the applications that benefit from GeForce RTX:
For Gaming
AI has revolutionized gaming, making experiences more immersive and realistic than ever before. Here are some of the AI-powered tools and features that enhance gaming:
DLSS (Deep Learning Super Sampling): This technology leverages AI to boost frame rates and generate high-quality images in real-time. Gamers can enjoy smoother gameplay and stunning visuals without compromising on performance. DLSS analyzes lower-resolution images and reconstructs them at a higher resolution, providing a significant performance boost while maintaining image clarity.
With the latest update, DLSS 3.5 introduces Ray Reconstruction which corrects ray-traced effects further which improves realism and improves the visual experience overall. Working in tandem with DLSS 3.0 Frame Generation, DLSS has reached the point where full path tracing can now be experienced in modern games, another step towards the holy grail of computer graphics.
Supporting DLSS is NVIDIA Reflex. This feature reduces system latency, ensuring faster response times in competitive gaming scenarios. By optimizing the rendering pipeline, imagine a smart frame capping method, games are more responsive, so games are snappy even in the most intense of situations, particularly competitive games like esports.
NVIDIA Broadcast: Tailored for streamers, this app uses AI to enhance audio and video quality. It can remove background noise, apply virtual backgrounds, and auto-frame the webcam, making live streams look and sound professional without the need for complex setups.
NVIDIA ACE (Avatar Cloud Engine): NVIDIA ACE is a suite of AI technologies designed to bring digital humans, AI NPCs, and interactive avatars to life. It uses generative AI for natural language interactions, realistic speech, and expressive animations.
ACE is more of game developer thing where they create more immersive and interactive experiences with NPCs that can understand and respond to players in a human-like manner. With NVIDIA ACE, the final game can include virtual characters who can engage in dynamic conversations, enhancing the depth and realism of gaming environments.
For Creators
Here’s a very broad category. Content creators benefit immensely from AI, as it streamlines workflows and enhances creativity. AI-powered tools for creators include a lot of sub-categories ranging from image generation to automatic rotoscoping all the way to audio clean-up.
And that’s just on the multimedia end of the spectrum. Architectural visualization and engineering, whether you’re in product development or material simulation.
NVIDIA Canvas: This application uses AI to transform simple brushstrokes into realistic landscapes. It allows artists to quickly create backgrounds and scenes, which can be further refined in other design software.
Adobe Creative Cloud: Many Adobe applications, such as Photoshop and Premiere Pro, integrate AI to simplify complex tasks. AI-powered features like auto-tagging in Lightroom, content-aware fill in Photoshop, and auto-reframe in Premiere Pro save creators time and effort, allowing them to focus on their creative vision. And yes, these are local AI functions which uses your own hardware.
Stable Diffusion: This tool enables high-quality image generation and enhancement. By using AI to understand and recreate intricate details, creators can produce visually stunning content with ease, whether it’s for digital art, photo editing, or video production.
Software Development
Developers can leverage AI to enhance their productivity and create more intelligent applications. AI development tools include:
ChatRTX for Development: You’ll see ChatRTX pop-up here more than once as its such a useful tool. For developers, whether new or professional, NVIDIA’s ChatRTX app allows you to basically speak to your code. Easily analyze the current status of the code or the entire project folder and ask for help if you’re stuck. NVIDIA has mentioned that a VSCode plugin for ChatRTX is also in the works to allow parallel assistance live on VScode as you write your programs.
TensorFlow and PyTorch: These popular AI frameworks benefit from GPU acceleration, allowing developers to train and deploy machine learning models more efficiently. High-performance GPUs offer exceptional support for these frameworks, speeding up model training and inference times.
Daily Use
AI has become an integral part of everyday computing, enhancing productivity and improving user experiences. AI-powered applications for daily use include:
Video Conferencing: Tools like NVIDIA Maxine use AI to improve video call quality. Features such as background noise removal, video upscaling, and face alignment ensure clear and professional video calls, making remote work more efficient.
ChatRTX: Told you this will show up again, this time let’s give a bit more background: ChatRTX features a fully offline chatbot that’s capable of various functions. A smarter way to interact with a directory, ChatRTX can scan your Documents folder so you can look-up important files easily like receipts from certain dates and much more.
Video Streaming: AI enhances video streaming quality by dynamically adjusting resolution but RTX users can use RTX Video to further improve video quality on video streams and in real-time.
For more professional users, NVIDIA has a growing reference list (https://www.nvidia.com/en-us/studio/software/) of RTX-accelerated and RTX AI-powered software including tools from Autodesk, Cinema4D, D5 Render, Keyshot, Octane, Redshift, TopazLabs, Unreal Engine and so much more.
Supporting Hardware
As mentioned, all the discussion we’ve had until now is about RTX-powered software. While many of these applications have native CPU or GPU support regardless of platform, RTX features fixed function hardware on all RTX graphics cards, even the 20- and 30-series, that allows all these cards to perform much faster than bruteforcing them on native computing.
That said, AI itself has predefined metric for gauging potential performance. We refer to this in TeraOps per Second or TOPS. Conveniently, it can also be Trillions of Operations per Second. Still TOPS, right? On a more serious note, this number is somewhat arbitrary as even the king of the consumer AI hill, the GeForce RTX 4090, would take a while to process things if we’re deal with large files or even larger data sets.
But it does help us create a hierarchy of GPUs to by. And its important to note that by this point in time, there is a reason there’s not a list for cards outside of NVIDIA’s products because in general, its hard to run AI natively on other cards.
The list below shows us a list of TOPS ratings for the current RTX 40-series cards. As an example, Microsoft’s Copilot+ requires 40TOPS to operate locally. Copilot+ is Microsoft’s internal chatbot which at its current state is nothing but a glorified search box so I wouldn’t care much for it.
What matters are functional apps that we’ve highlighted that show real-world benefits to user.
It will be easier to align your performance expectations closer to gaming performance to have an easier time deciphering these numbers. And just like gaming, all these cards are ready to game. For example, the GeForce RTX 4060 has a TOPS rating of 242, easily 5x the platform TOPS rating of current AI SOCs yet lack acceleration for other functions which is quite limiting if you’re looking for more task-based AI applications.
PALIT GeForce RTX 4060 Infinity 2
Expanding on our previous example, even budget builders will be ready to enjoy and use Powered by RTX AI applications on any GeForce card, even the starter GeForce RTX 4060.
Builders who want the best value may appreciate Palit’s updated RTX 4060 with a new cooler design. The PALIT GeForce RTX 4060 Infinity 2 is a dual-slot graphics card powered off an 8-pin PCIe power connector. It retain many of the features set that the RTX 40-series is known for including RTX AI acceleration for games and creator applications.
Featuring a pair of 95mm fans, the fans compliment the subtle styling of the card and is tuned to operate and low noise levels. If you’re upgrading an older computer or building a power efficient gaming and study machine, the Palit GeForce RTX 4060 Infinity 2 is a nice option to get you started.
Closing Thoughts
With COMPUTEX 2024 behind us, we saw the deluge of AI products vying for your attention but as I’ve preached over the past few months the best advice is to focus on the things you currently do, not what you think you can do. For desktop and even laptop owners, if you have a system that’s at most 5 years old and you happen to be rocking a GeForce RTX card, you don’t need to ride the AI update hype. But if part of what you do would benefit with a GPU upgrade, particularly a mix of gaming and creator apps, then picking up a new GeForce RTX 40-series card could be for you.
All of that said, for some a GPU purchase may be a big decision and getting a cheaper alternative may be much more costly in the long run. All things considered, we have the hardware to run AI with current-gen hardware we only need the software to fully utilize and deliver the requirements.
Follow us to stay updated on AI for gaming and creators as we provide similar updates monthly right here at Back2Gaming.
