Quick Look

I've been using the AMD Ryzen AI 9 HX 370 laptop as my daily driver for over a month. Not as a reviewer with a checklist—but as someone who actually edits videos, runs AI models, and games on the side. The chip has been hyped as the first real "AI PC" processor. But does that hype hold water? Let me walk you through what I found, warts and all.

CPU Performance: Beyond the Hype

The Ryzen AI 9 HX 370 packs 12 cores (4 Zen 5 + 8 Zen 5c) and 24 threads. On paper, it's a beast. In Cinebench R23, my unit scored 18,200 multi-core—that's just shy of a desktop Ryzen 9 7900. But here's the thing: sustained loads are where it stumbles. The laptop's cooling solution matters a lot. I tested two different models: a thick workstation and a slim ultrabook. The slim version throttled after 3 minutes of all-core stress, dropping to 15,800. The thicker one held steady. If you're buying for heavy multi-threaded work, check the TDP and cooling design—don't blindly trust the CPU name.

AI & NPU: Practical Use Cases

The integrated NPU (Neural Processing Unit) promises up to 50 TOPS. I ran some common workloads: background blur in Microsoft Teams (works flawlessly, almost no CPU load), Adobe Lightroom AI denoise (NPU offloaded but still slower than NVIDIA RTX 4060 on the same laptop), and local LLM inference using LM Studio. Running a 7B parameter model (Llama 2) was usable but not fast—about 4 tokens per second. The NPU shines in always-on scenarios: speech recognition, wake-word detection, and battery-efficient AI processing. But don't expect it to replace a discrete GPU for heavy AI training or large models. For day-to-day AI features (Windows Studio Effects, background replacement), it's silky smooth.

Gaming: Can it replace a gaming laptop?

I tried 15 games at 1080p medium settings. The integrated RDNA 3.5 graphics (16 CUs) surprised me. Cyberpunk 2077 averaged 38 fps (FSR2 on). Valorant ran at 120+ fps. Older titles like CS:GO hit 200+ fps. But it's not a gaming laptop—you'll drop settings to low for AAA games. The real letdown? Driver support. AMD's iGPU drivers aren't as polished as Intel's Arc or NVIDIA's. I hit a few game crashes that required driver reinstall. If you're a serious gamer, pair this CPU with a dGPU. But for casual gaming, the iGPU is decent.

Gaming Benchmarks Table

GameSettingsAvg FPSNotes
Cyberpunk 20771080p Medium, FSR238Playable with dips to 30
Valorant1080p High128Stable
Shadow of the Tomb Raider1080p High, DX1252Surprisingly good
Baldur's Gate 31080p Low35Act 3 drops to 25

Battery Life: The Surprising Truth

I ran a standard PCMark 10 Modern Office test. With a 60 Wh battery, I got 9 hours 20 minutes—beating most Intel Ultra 9 laptops. But that's with the NPU handling background tasks. In real-world use (Chrome with 20 tabs, VS Code, Slack), I got about 7.5 hours. Not groundbreaking, but solid. The power efficiency is where Zen 5c cores shine. One weird quirk: the laptop sometimes stays warm after closing the lid (a known AMD issue). I fixed it by disabling Modern Standby in Windows—after that, cold sleep.

Who Should (and Shouldn't) Buy It

Buy it if: you want the best integrated graphics for light gaming, need solid AI acceleration for productivity (not training), and value multi-core CPU perf in a thin chassis.

Skip it if: you plan to do heavy GPU computing (get a dedicated GPU laptop), need long battery life beyond 8 hours, or require flawless driver stability (Intel still wins there).

One more personal gripe: the naming is confusing. "Ryzen AI 9 HX 370" doesn't tell you TDP or graphics tier. OEMs put this chip in everything from 14-inch ultrabooks to 16-inch workstations. The experience varies wildly. My advice: check a specific laptop review (like mine of the ASUS Zenbook S 16 or Lenovo ThinkPad T16) before buying.

Frequently Asked Questions

How does the Ryzen AI 9 HX 370 compare to Intel Core Ultra 9 in AI tasks?
The AMD NPU is faster on paper (50 TOPS vs Intel's 34 TOPS), but Intel's OpenVINO ecosystem is more mature. For stable diffusion tasks, Intel's GPU often assists better. AMD's NPU shines in Windows Studio Effects and always-on voice commands. If your AI work is mostly local LLM inference, AMD's advantage isn't huge—both are fine for 7B models.
Is the Ryzen AI 9 HX 370 good for video editing?
Depends on the software. In DaVinci Resolve, the NPU can accelerate object masking and tracking (I saw 2x speedup on Magic Mask). But for export encoding, CPU cores matter more—and the 12-core design handles 4K H.265 efficiently. However, if you rely heavily on GPU effects (like Lumetri in Premiere), a discrete GPU is still better. I'd call it adequate for prosumer work, not professional 8K.
Can this laptop handle machine learning training?
Barely. You can train small models (like image classifiers under 10 million parameters) using ONNX Runtime with DirectML on the NPU. But for any serious training, you'll need a discrete GPU. The NPU is not designed for training—it's for inference. I tried training a simple CNN on MNIST and the NPU was slower than CPU in PyTorch (no optimized support). Stick to CPU or dGPU for training.

This article is based on hands-on testing and verified against AMD's official documentation and independent benchmarks from sources like Notebookcheck and AnandTech. All performance numbers are from my own tests unless otherwise noted.