AMD Ryzen AI 5 330

AMD Ryzen AI 5 330

AMD Ryzen AI 5 330 — an entry-level mobile APU with a 50-TOPS NPU

A compact, entry-level mobile processor for the Copilot+ PC platform. It combines 4 cores/8 threads on Zen 5/Zen 5c, a simple integrated Radeon 820M GPU, and a built-in XDNA 2 NPU rated up to 50 TOPS. It targets thin-and-light laptops and mini-PCs with a focus on efficiency and on-device AI scenarios.

Key specifications

  • Architecture / codename, process node: hybrid Zen 5 + Zen 5c; Ryzen AI 300 family (mobile APU generation); 4 nm node.

  • Cores / threads: 4 / 8 (1 performance Zen 5 and 3 efficiency-oriented Zen 5c; SMT enabled).

  • Frequencies (base; boost): ~2.0 GHz; up to ~4.5 GHz (actual values depend on power/thermal profiles).

  • L3 cache: 8 MB (L2 — 4 MB).

  • Power envelope: default TDP 28 W; cTDP range 15–28 W.

  • Integrated graphics: Radeon 820M (RDNA 3.5), 2 CU / 128 ALUs; aimed at basic 2D/video workloads and light 1080p gaming.

  • Memory: dual-channel DDR5-5600 or LPDDR5X-8000; a full dual-channel configuration is critical.

  • Interfaces: 14 native PCIe 4.0 lanes (allocation varies by OEM), USB4 (up to 40 Gbps, typically 2 ports), USB 3.2 Gen 2 (10 Gbps), output to up to four displays via USB-C (DP Alt Mode)/HDMI depending on implementation.

  • NPU / Ryzen AI: XDNA 2, up to 50 TOPS INT8; enables on-device features (including Copilot+ PC functions in Windows).

  • Optional: performance positioning: below Ryzen AI 5 340/AI 7 350 in multi-thread and graphics; single-thread performance can be close to Zen 5 levels of higher SKUs at similar clocks.

What this chip is and where it fits

Ryzen AI 5 330 is the most affordable member of the Ryzen AI 300 lineup for mobile. The model is intended for mainstream thin laptops, budget Copilot+ systems, and compact mini-PCs where long battery life, quiet operation, and baseline on-device AI capabilities matter. Within the 300-series hierarchy it sits below Ryzen AI 5 340, Ryzen AI 7 350, and AI 9 SKUs, emphasizing cost and efficiency.

Architecture and process

The die uses a hybrid layout: performance Zen 5 cores are paired with efficiency-focused Zen 5c cores. This allows task-type scheduling: peak single-thread bursts on the larger cores, sustained background or service workloads on the compact ones. The cache subsystem includes 4 MB of L2 (1 MB per core) and 8 MB of shared L3 — sufficient for desktop apps, heavy browser use, IDEs, and office suites at the entry tier.

The multimedia blocks support modern video codecs up to AV1 (hardware decode), along with HEVC and H.264; hardware encode capabilities depend on the iGPU media engines, but everyday streaming and video calls are covered. The memory controller supports both socketed DDR5 and high-speed LPDDR5X — the trade-off between latency and bandwidth depends on platform goals and the notebook’s design.

CPU performance

In day-to-day workloads — office apps, browsers with dozens of tabs, video conferencing, light development and builds of smaller projects, simple photo tasks — Ryzen AI 5 330 delivers snappy behavior via Zen 5 IPC and boost clocks up to ~4.5 GHz. In heavily parallel scenarios (large-scale rendering, very big codebases, multi-threaded archiving), four cores/eight threads become the limiting factor; acceleration is bounded by the core count and TDP.

A practical note: results scale with the thermal package and cooling system. In chassis holding TDP at 15–18 W, sustained clocks under long load will be lower than in designs configured at 25–28 W with more robust heatpipes. This primarily affects lengthy batch jobs: the better the cooling and the higher the power limit, the closer sustained performance gets to short-term peaks.

Graphics and multimedia (iGPU)

The integrated Radeon 820M (RDNA 3.5, 2 CU/128 ALUs) targets UI acceleration, media playback, and basic graphics. For games, think 1080p on low/occasionally medium presets in lighter and competitive titles, with wide variation by game and settings. For AAA titles and high presets, a discrete GPU or higher-tier AI 300 APU with more CUs is preferable.

Memory bandwidth strongly affects the iGPU: dual-channel LPDDR5X-8000 delivers better frame pacing than DDR5-5600, though final results depend on timings, controller behavior, and power limits. For media, AV1/HEVC/H.264 hardware decode ensures smooth 4K streaming with low power draw.

AI / NPU

The built-in XDNA 2 NPU rated up to 50 TOPS accelerates on-device AI: speech transformations (captions, noise suppression), video effects (background blur, auto-framing), on-device object recognition, metadata generation, and local inference for moderate-size models. The NPU offloads CPU/GPU, reduces power in those tasks, and satisfies Copilot+ PC requirements. NPU capability is consistent across the Ryzen AI 300 family, simplifying developer targeting.

Platform and I/O

A typical mobile platform around Ryzen AI 5 330 provides:

  • PCIe: 14 native lanes of PCIe 4.0. OEMs split them between NVMe storage and, where present, discrete graphics or high-speed controllers; exact routing varies by device.

  • USB/Thunderbolt: native USB4 (up to 40 Gbps), commonly two fully featured USB-C ports with DP Alt Mode and Power Delivery; plus USB 3.2 Gen 2 (10 Gbps). Thunderbolt interoperability depends on system-level certification.

  • Displays: up to four independent outputs (combinations of USB-C/DP Alt Mode and HDMI), subject to the notebook or mini-PC implementation.

  • Networking & storage: typically Wi-Fi 6E/7 via the OEM’s module, one or two M.2 2280 (PCIe 4.0 ×4) for NVMe, sometimes a card reader slot.

Power consumption and cooling

With a 15–28 W cTDP range and a compact die, the 330 is tailored to slim chassis. In “Balanced”/“Silent” profiles it maintains low temperatures and noise at the expense of sustained performance; in “Performance” profiles sustained clocks are higher but cooling becomes the priority. With sensible fan curves and a competent heatpipe, the CPU avoids pronounced frequency “sawtoothing” under long loads.

Where it appears

Expect Ryzen AI 5 330 in mainstream 13–15-inch entry-level laptops, education and office-oriented models, and compact mini-PCs focused on office work, multimedia, light development, and everyday tasks. Specific series and timelines depend on OEMs.

Comparison and positioning

  • Versus Ryzen AI 5 340: the 340 brings 6 cores/12 threads and Radeon 840M (4 CU), clearly boosting multi-threaded and graphics performance. NPU is the same (50 TOPS); TDP is similar. The 330 trades raw throughput for lower cost and simpler graphics/core counts.

  • Versus Ryzen AI 7 350: 8 cores/16 threads and a stronger iGPU (860M, 8 CU) move the 350 into a different class for rendering, encoding, and gaming. The power envelope is comparable, with greater cooling demands.

  • Versus Ryzen AI 9 (365/370/HX): double-digit core counts, higher boosts, and large iGPUs (880M/890M) define the premium tier. The 330 is the baseline entry to Copilot+ PCs, optimized for efficiency and value.

Who it suits

  • Office and study: documents, spreadsheets, presentations, video calls, note-taking with local transcription — with long battery life.

  • Browsing and everyday tasks: heavy tab use, mail, messaging, online services.

  • Light development and automation: IDE work, builds for small/medium projects, local scripts; for CI with very high parallelism, higher SKUs are advisable.

  • Multimedia: 4K playback, basic clip editing, format conversion without strict deadlines.

  • Entry-level gaming: lighter titles at 1080p on low/medium settings; for AAA or high FPS, aim for a dGPU or higher APU.

Pros and cons

Pros

  • XDNA 2 NPU at 50 TOPS — same AI acceleration level as higher AI 300 models.

  • Modern I/O (USB4, 14 lanes of PCIe 4.0), support for up to four displays.

  • Strong efficiency with a wide 15–28 W cTDP span.

  • Support for fast LPDDR5X memory and hardware AV1 decode.

Cons

  • Only 4 cores/8 threads limit heavily parallel workloads.

  • Radeon 820M with 2 CUs is the minimal 3D tier in the family.

  • Sustained clocks depend noticeably on cooling and power limits.

  • I/O routing and port speeds vary with each notebook design.

Configuration recommendations

  • Memory: strictly dual-channel; for iGPU, prefer LPDDR5X-7500/8000. Capacity — at least 16 GB; 32 GB gives headroom for IDEs, browsers, and media apps.

  • Storage: NVMe PCIe 4.0 ×4; for working datasets, consider two M.2 drives (system + data/cache).

  • Cooling: prioritize designs with a mid-sized fan and heatpipe; “Balanced/Performance” profiles with sensible curves help sustain clocks in long jobs.

  • Power modes: on AC, use “Performance” and allow higher power limits within 25–28 W; on battery, “Balanced” extends runtime with modest performance trade-offs.

  • Graphics/displays: for multi-monitor use, ensure two USB4 ports with DP Alt Mode; for occasional editing and encode offload, lean on AV1/HEVC paths where hardware engines are most efficient.

Summary

Ryzen AI 5 330 is a cost-effective entry to the Copilot+ PC ecosystem: modern Zen 5/Zen 5c architecture, a platform-standard 50-TOPS NPU, and a basic iGPU adequate for UI, video, and light gaming. For heavy multi-thread and 3D workloads, AI 5 340/AI 7 350 with more cores and CUs are better fits, but in compact and affordable laptops the 330 strikes the needed balance of efficiency, contemporary I/O, and on-device AI capability. When priorities are battery life, office/browsing/media, and local AI features, this processor fits the brief; for intensive rendering, wide parallel builds, or high-end gaming, choose a higher-tier SKU or add a discrete GPU.

기초적인

라벨 이름
AMD
플랫폼
Laptop
출시일
July 2025
모델명
?
인텔 프로세서 번호는 프로세서 브랜드, 시스템 구성, 시스템 수준 벤치마크와 함께 컴퓨팅 요구 사항에 적합한 프로세서를 선택할 때 고려해야 할 여러 요소 중 하나일 뿐입니다.
Ryzen AI 5 330
코어 아키텍처
Krackan Point 2
주조소
TSMC
세대
Ryzen AI (Zen 5)

CPU 사양

전체 코어 개수
?
코어는 단일 컴퓨팅 구성 요소(다이 또는 칩)에 있는 독립적인 중앙 처리 장치의 수를 설명하는 하드웨어 용어입니다.
4
전체 스레드 개수
?
해당하는 경우 인텔® 하이퍼 스레딩 기술은 성능 코어에서만 사용할 수 있습니다.
8
성능 코어 기본 주파수
2 GHz
효율 코어 기본 주파수
2000 MHz
성능 코어 터보 주파수
?
Intel® 터보 부스트 기술에서 파생된 최대 P-코어 터보 주파수.
4.5 GHz
L1 캐시
80 KB per core
L2 캐시
1 MB per core (4MB total)
L3 캐시
8 MB
버스 주파수
100 MHz
곱셈기
20.0
잠금 해제된 곱셈기
No
소켓
?
소켓은 프로세서와 마더보드 사이의 기계적, 전기적 연결을 제공하는 구성 요소입니다.
AMD Socket FP8
제조 공정
?
리소그래피는 집적 회로를 제조하는 데 사용되는 반도체 기술을 말하며, 반도체 위에 구축된 형상의 크기를 나타내는 나노미터(nm) 단위로 보고됩니다.
4 nm
전력 소비
28 W
최고 온도
?
접합 온도는 프로세서 다이에서 허용되는 최대 온도입니다.
100°C
PCIe 버전
?
PCI Express는 고속 직렬 컴퓨터 확장 버스 표준으로, 고속 컴포넌트를 연결하는 데 사용되며, AGP, PCI 및 PCI-X와 같은 이전 표준을 대체합니다. 처음 출시 이후 여러 번의 개정과 개선이 이루어졌습니다. PCIe 1.0은 2002년에 처음 도입되었으며, 높은 대역폭에 대한 요구를 충족시키기 위해 시간이 지남에 따라 후속 버전이 출시되었습니다.
4

메모리 사양

메모리 유형
?
인텔® 프로세서는 단일 채널, 듀얼 채널, 삼중 채널 및 플렉스 모드의 네 가지 유형으로 제공됩니다. 다중 메모리 채널을 지원하는 제품에서 채널당 다중 DIMM을 장착하는 경우 지원되는 최대 메모리 속도가 낮아질 수 있습니다.
LPDDR5x-8000, DDR5-5600
최대 메모리 채널
?
메모리 채널 수는 실제 적용을 위한 대역폭 작동을 나타냅니다.
2
ECC 메모리 지원
No

GPU 사양

통합 그래픽스
?
통합 GPU는 CPU 프로세서에 통합된 그래픽스 코어를 지칭합니다. 프로세서의 강력한 연산 능력과 지능적인 전력 효율 관리를 활용하여, 더 낮은 전력 소비로 뛰어난 그래픽 성능과 부드러운 응용 프로그램 경험을 제공합니다.
Radeon 820M
그래픽 주파수
?
그래픽 최대 동적 주파수는 동적 주파수 기능이 있는 인텔® HD 그래픽을 사용하여 지원할 수 있는 최대 기회적 그래픽 렌더링 클럭 주파수(MHz)를 나타냅니다.
2800 MHz
Graphics Core Count
2
DirectX Support
?
DirectX* Support indicates support for a specific version of Microsoft’s collection of APIs (Application Programming Interfaces) for handling multimedia compute tasks.
12
최대 해상도
7680x4320 @ 60Hz , 3840x2160 @ 240Hz , 3440x1440 @ 360Hz , 2560x1440 @ 480Hz , 1920x1080 @ 600Hz
Number of Displays Supported
4

여러 가지 잡다한

PCIe 레인 수
14

벤치마크

Geekbench 6
싱글 코어 점수
2131
Geekbench 6
멀티 코어 점수
7079

다른 CPU와 비교

Geekbench 6 싱글 코어
2314 +8.6%
2207 +3.6%
2034 -4.6%
1934 -9.2%
Geekbench 6 멀티 코어
8035 +13.5%
7510 +6.1%
6673 -5.7%
6311 -10.8%