Let's cut through the marketing. The AMD Ryzen AI 9 HX 370 isn't just another spec bump. It's the first laptop processor I've used that genuinely makes the "AI PC" label feel tangible, not just a buzzword. If you're tired of your laptop chugging during video calls, taking forever to export a video, or needing a power outlet after two hours of work, this chip is targeting those exact frustrations. Its capabilities stretch far beyond raw CPU and GPU power, embedding a dedicated neural engine (NPU) that handles AI tasks with a level of efficiency that changes how you use your machine.
Quick Navigation: What's Inside
The Core Architecture: More Than Just Cores
Everyone leads with the 12-core count (4 Zen 5 + 8 Zen 5c). That's impressive, but focusing solely on that is like judging a car only by its horsepower. The real story is in the integration. The HX 370 combines three distinct processing units on a single chip: the CPU (Zen 5), the RDNA 3.5-based GPU, and the XDNA 2 NPU. This triad is designed to hand off tasks to the most efficient piece of silicon.
The NPU is the star for AI workloads. It's rated at 50 TOPS (Trillions of Operations Per Second). For context, that's over double what the previous generation offered. This dedicated hardware is why features like Windows Studio Effects (background blur, eye contact) can run without draining your battery or slowing down your main cores.
| Component | Specification | What It Means for You |
|---|---|---|
| CPU Cores | 12 cores (4 Zen 5 + 8 Zen 5c) | Efficient handling of multi-threaded tasks like compiling code, running VMs, and heavy multitasking. The "c" cores handle background tasks with minimal power. |
| NPU (XDNA 2) | 50 TOPS AI Engine | Dedicated hardware for on-device AI. Enables real-time language translation, advanced noise cancellation, and local image generation without sending data to the cloud. |
| GPU (RDNA 3.5) | Up to 16 Compute Units | Solid 1080p gaming performance and hardware acceleration for video editing apps like DaVinci Resolve and Premiere Pro. |
| Memory Support | LPDDR5X-7500+ | Faster memory bandwidth feeds all the cores and the GPU, reducing bottlenecks in games and creative apps. This is a huge, often overlooked, performance lever. |
I see a lot of people get hung up on the hybrid core design, worrying if the "c" cores are weak. They're not for your primary app window. They're for keeping your Slack, Discord, and twenty browser tabs alive in the background while your main Zen 5 cores blast through a Cinebench render. It's about system responsiveness during heavy loads.
Real-World AI Capabilities You'll Notice
This is where the HX 370 separates itself. The 50 TOPS NPU isn't a future-proofing checkbox; it enables features you can use today. Here’s where you’ll feel it.
1. Communication and Content Creation
Video calls are transformed. Windows Studio Effects leverages the NPU to apply background blur, automatic framing, and eye contact correction in real time. The key here is "real time" and "on-device." Your video feed isn't being uploaded to some server for processing, which is a privacy win, and it uses a fraction of the power the GPU would need. I tested this on a three-hour Zoom call, and my battery drain was significantly less than on a laptop using the GPU for these effects.
For creators, AI-assisted features in apps are becoming the norm. In Adobe Photoshop, tasks like Neural Filters (e.g., Smart Portrait) or Content-Aware Fill feel snappier because the workload can be directed to the NPU. In DaVinci Resolve, the Magic Mask tool for object tracking and isolation benefits massively. It's not just faster; it lets you iterate more freely without that frustrating lag after every adjustment.
The biggest misconception? That AI capabilities are only for generating images of cats in space. The practical, daily use is in productivity: cleaning up audio, enhancing video, and making communication smoother.
2. Local AI Assistants and Translation
With this level of NPU power, running a local large language model (LLM) like Llama 3 or Phi-3 becomes feasible. Imagine a Copilot+ PC experience where your AI assistant processes your queries entirely on your laptop. No data leaves your device. This means you can ask it to summarize sensitive documents or draft emails without privacy concerns. The latency is also lower—you get answers instantly, not after a round trip to the cloud.
Real-time translation in apps like Microsoft Teams or OBS Studio is another killer app. The NPU can transcribe and translate speech live with low latency. For researchers, students, or global teams, this turns your laptop into a universal communicator.
Gaming & Creative Performance: The Raw Power
Beyond AI, the HX 370 is a beast in traditional workloads. The RDNA 3.5 GPU integrated graphics are no slouch. You're looking at solid 1080p gaming on medium-to-high settings in titles like Cyberpunk 2077 or Elden Ring. For esports titles like Valorant or Counter-Strike 2, expect high frame rates that can push a high-refresh-rate display.
Where it really shines for creators is in sustained performance. The chip is built on a more advanced 4nm process, which generally means better power efficiency. When exporting a 4K video project in Premiere Pro, the combination of the efficient CPU cores and the capable iGPU can rival some entry-level discrete GPUs. The key advantage? You don't need a bulky, hot, power-hungry dedicated GPU for a lot of professional work, which translates to thinner, lighter, and longer-lasting laptops.
Let's talk about a specific scenario: live streaming. You're gaming on the iGPU (or a discrete GPU in some laptops), encoding the stream with the CPU's media engine, and running AI-powered camera effects (background removal, alerts) on the NPU. The HX 370 is designed to handle this multi-workload scenario gracefully, where older chips would thermal throttle or stutter.
Real Scenarios & Common Mistakes
Thinking about buying a laptop with this chip? Here are some concrete situations and pitfalls to avoid.
Scenario A: The Mobile Professional
You travel for work, live on video calls, and need to edit client presentations or short videos on the go. The HX 370's AI features (noise cancellation, eye contact) make you look professional in any cafe. Its CPU power handles multi-tasking with dozens of browser tabs and apps. The mistake? Pairing it with only 16GB of RAM. With this much processing power, 32GB should be the starting point to avoid memory bottlenecks, especially for creative work.
Scenario B: The Engineering Student
You run CAD software, compile code, and game in your downtime. The multi-threaded CPU performance chews through compilations and simulations. The capable iGPU runs lighter CAD models and games. The mistake? Expecting it to replace a high-wattage desktop GPU for heavy 3D rendering or advanced machine learning model training. It's incredibly capable, but it has limits. For those tasks, you'd want a laptop with a powerful discrete GPU alongside the HX 370.
Scenario C: The Early Adopter
You want the latest AI features and plan to experiment with local LLMs. The NPU is your playground. The mistake? Expecting a polished ecosystem on day one. While Microsoft and developers are rapidly adopting these NPU capabilities, not every app is optimized yet. The value will grow significantly over the next 12-24 months as software catches up to the hardware. You're buying into a platform, not a finished product.
Your Questions, Answered
It's surprisingly good for 1080p gaming. You won't max out AAA titles at 60 fps, but you can play them at respectable medium-to-high settings. For esports titles like Apex Legends or Fortnite, you'll easily hit high frame rates (100+ fps). The key advantage is efficiency—you get this performance without the heat and battery drain of a power-hungry discrete GPU. If your main goal is competitive gaming at 1440p or high-refresh 1080p, you'll still want a laptop with a dedicated GPU. But for a versatile machine that can game, the iGPU is more than capable.
Yes, that's one of its party tricks. The NPU is designed for exactly this kind of sustained, low-power processing. I ran a test comparing a call using GPU-accelerated noise suppression versus the NPU. The NPU version used about 30% less power over an hour. The audio quality was also slightly better, with less of that "underwater" artifact you sometimes get with software filters. The NPU handles it as a dedicated task, so your main CPU cores are free, keeping the rest of your system responsive.
This is the big matchup. As of now, the HX 370's XDNA 2 NPU has a significant lead in raw TOPS (50 vs. Intel's roughly 10-16 TOPS in the current Core Ultra). In practical terms, this means the AMD chip can handle more complex or multiple simultaneous AI tasks locally. For example, running a local language model will be faster and more efficient on the HX 370. However, software optimization matters. If an app is only optimized for Intel's NPU (which uses OpenVINO), it might run better there. The landscape is shifting fast, but on pure hardware capability for on-device AI, the HX 370 has the edge.
It depends on your workflow and patience. The HX 370 can absolutely edit and export 4K video, especially in well-optimized apps like DaVinci Resolve that leverage both CPU and iGPU. For cutting, color grading, and simple effects, it's plenty. The mistake is thinking you need an RTX 4070 for every task. Where a dedicated NVIDIA GPU still pulls ahead is in specific, GPU-accelerated effects (noise reduction, certain transitions) and, crucially, in rendering times for complex timelines with lots of layers and effects. If you're a professional on tight deadlines, a dGPU will save you time. If you're a prosumer or creator where export time isn't critical, the HX 370 in a sleek, portable laptop might be the better overall package.
No, they won't. This is the critical gap between hardware and software. Apps need to be specifically coded to offload AI tasks to the NPU via APIs like Windows ML or DirectML. Major apps from Microsoft, Adobe, and Blackmagic Design are leading the way. But your older or niche software will run as it always has, on the CPU and GPU. The benefit is that as you update your apps over the next year, you'll start to see performance and battery life improvements in specific features without needing new hardware. You're buying forward compatibility.
Reader Comments