The recent rise of artificial intelligence (AI) and AI PCs has driven a tremendous amount of market buzz and raised a host of questions. Everyone from individual end-users to major corporate buyers are asking what an AI PC is, what hardware is required to take advantage of AI, which applications leverage it, and whether it is better to deploy such services locally, via the cloud, or in a hybrid environment that blends aspects of both.
This confusion is normal. AI represents a new frontier for computing. In the long term, it may fundamentally change both how we interact with computers and the ways we integrate them into our everyday lives. Let’s tackle some of these questions, starting with one of the basic ones:
What is an AI PC?
An AI PC is a PC designed to optimally execute local AI workloads across a range of hardware, including the CPU (central processing unit), GPU (graphics processing unit), and NPU (neural processing unit). Each of these components has a different part to play when it comes to enabling AI workloads. CPUs offer maximum flexibility, GPUs are the fastest and popular choice to run AI workloads, and NPUs are designed to execute AI workloads with maximum power efficiency. Combined, these capabilities allow AI PCs to run artificial intelligence aka machine learning tasks more effectively than previous generations of PC hardware. Check out this video to learn more about AMD Ryzen™ AI PCs.
What is the Difference Between Local and Cloud Computing?
If a workload is processed locally, that means it runs on specialized silicon inside the user’s laptop or desktop. Select models of the AMD Ryzen™ Mobile 7040 Series and AMD Ryzen™ Mobile 8040 Series processors contain NPUs specifically designed to handle emerging AI workloads, as do our recently launched Ryzen™ 8000G desktop processors. AI workloads can also be run locally via a discrete GPU (if present), an integrated graphics solution, or directly on the CPU depending on how the application is optimized. For best viewing, make sure you set the video to the highest resolution your monitor supports.
If a workload is processed in the cloud, that means information is being relayed from the end-user’s PC to a remote service provided by a third-party. Major AI services like ChatGPT, and Stable Diffusion that are commonly discussed today are all cloud-based services, for example. Cloud-based AI services typically rely on high-end server-class discrete GPUs or specialized data center accelerators like the AMD Instinct™ MI300X and AMD Instinct™ MI300A.
If you ask a cloud-based generative AI service to draw you a landscape or a bouquet of flowers, your request is relayed and processed by a remote server. Once complete, the image is returned to you. If you’ve experimented with any of the freely available generative AI services for text or speech, you are aware that it can take these cloud-based applications up to several minutes to return results depending on the complexity of your request and the number of other requests that the cloud service is processing.
What are the Strengths and Weaknesses of Cloud vs. Local Computing?
Each of these approaches has its own merits and demerits. The advantage of local AI computing is that the work is handled locally on your device. It takes less time for the CPU, GPU, or NPU built into a system to spin up and start processing a task than it does to send that same task to a server located hundreds or thousands of miles away. Keeping data local may also help to improve user privacy since these devices are designed to keep sensitive information from being inadvertently transmitted or shared. The advantage of running a workload locally on your device is improved latency.
Cloud computing is not without its own advantages, however. Sending data to a distant server may take a measurable amount of time, but remote data center services, aka cloud-based services, can run a given query on an array of hardware that is far more powerful than any single laptop, desktop, or workstation. The advantage of running a workload in the cloud is scale, which at times outweighs the need for quick response times or the intrinsic desire to keep certain data private.
The question of whether local AI or cloud-based AI is better depends on the end-users’ needs and the characteristics of the application. Both cloud and local AI services are complimentary to each other, creating an opportunity for future hybrid services. Cloud-based providers want to reduce the gap between question and response to as little as possible, while AI PC hardware available for local AI processing is rapidly improving. Imagine a conversational chatbot that relied on a cloud service to provide general background information on various topics but relied on local processing any time it needed to reference documents or other files stored on your AI PC.
AMD is investing in AI at every level and powering the new era of AI PCs
Succeeding in AI requires investing in every part of the business, from strategic partnerships with leading ISVs (Independent Software Vendors) to software investments in toolchains, platforms, and libraries. It means supporting AI capabilities across our entire product line of CPUs, GPUs, and NPUs. That’s exactly what AMD has done. Some of AMD's significant AI advances are listed below.
- AMD acquired Xilinx partly to integrate their hardware into our chips (AMD Ryzen™ 7040 Series, 8040 Series, and 8000G series).
- AMD introduced the first NPU, AMD Ryzen™ AI in an x86 processor in June 2023. This video discusses how AMD is advancing AI PCs in 2024 in greater detail.
- AMD is working with ISVs like Topaz Labs, Microsoft, Blackmagic, and Adobe to develop software products for AMD Ryzen™ AI based devices. These videos cover AMD’s partnership with Topaz Labs and AI-Powered Editing with Adobe Premiere Pro.
- AMD launched unique data center chips like AMD Instinct MI300X and MI300A with features and capabilities that are unmatched in their respective markets.
- AMD acquired Nod.AI, the open-source AI software development house, last year and released the Ryzen™ AI software stack in December for general availability.
- AMD released ROCm 6.0 in Feb 2024, adding ONNX compatibility, support for new GPUs and math types, and newly implemented support for workloads that mix FP32 and FP16 data types. Additional details are available here.
AMD is committed to Advancing AI.
The optimizations and partnerships we’ve already publicly discussed with our ISV partners are proof that AMD is working to make certain AI won’t just be a gimmick. We’re building a deep reservoir of talent at every level of the company to support the development of artificial intelligence and we’re excited about how this technology could change computing in the future.
Chetna Gupta is Director, Global Commercial Client Marketing for AMD. Her postings are her own opinions and may not represent AMD’s positions, strategies, or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied. GD-5.
Use or mention of third-party marks, logos, products, services, or solutions herein is for informational purposes only and no endorsement by AMD is intended or implied. GD-83.
