This article contains affiliate links. If you purchase through our links, we may earn a commission at no extra cost to you. We only recommend products we’ve thoroughly researched and verified.
For a broader shortlist, start with our best mini PC for home server guide.
Both the Radeon 780M and 890M appear in AMD mini PCs across a $300-700 price range, and the distinction matters for users considering the jump from a Ryzen 7 H255 machine to a Ryzen AI 9 HX 370 machine. Here’s the honest comparison.
Specifications Side by Side
| Spec | Radeon 780M | Radeon 890M |
|---|---|---|
| Compute Units | 12 CU | 16 CU |
| Architecture | RDNA 3 | RDNA 3.5 |
| GPU Clock boost | Up to 2,700 MHz | Up to 2,900 MHz |
| Host CPU | Ryzen 7 H255 / Ryzen 9 7940HS/8945HS | Ryzen AI 9 HX 370/365 |
| Memory bandwidth | ~68-89 GB/s | ~100-118 GB/s |
| ROCm support | Yes | Yes |
| VA-API support | Yes | Yes |
| DirectX | DX12 Ultimate | DX12 Ultimate |
| RDNA generation | RDNA 3 | RDNA 3.5 |
Performance uplift 890M over 780M: ~25-30% for GPU compute; ~15-20% for gaming.
Performance Comparison by Workload
Gaming 1080p Low Settings
| Game | 780M SER9 | 890M X1 Pro-370 |
|---|---|---|
| Cyberpunk 2077 FSR Balanced | ~35-45 FPS | ~45-55 FPS |
| Fortnite Medium | ~60-75 FPS | ~75-90 FPS |
| CS2 Medium | ~90-110 FPS | ~110-130 FPS |
| Elden Ring Low | ~40-50 FPS | ~50-60 FPS |
Gaming performance: 890M is noticeably better — enough to push borderline 60fps targets over the threshold in demanding titles.
PS3 Emulation RPCS3
Both benefit from clock speed — RPCS3 is single-thread limited. The 890M’s slightly higher IPC RDNA 3.5 improvements and the HX370’s Zen 5 cores give it an edge.
| Title | 780M SER9 | 890M A9 Max |
|---|---|---|
| Persona 5 | Full speed | Full speed |
| God of War 3 | 35-45 FPS | 40-55 FPS |
| The Last of Us | 40-50 FPS | 45-55 FPS |
AI Inference Ollama, ROCm
| Model | 780M 32GB SER9 | 890M 32GB X1 Pro-370 |
|---|---|---|
| Llama 3.1 8B Q4 | ~15-18 tok/s | ~20-25 tok/s |
| Qwen2.5 14B Q4 | ~8-10 tok/s | ~10-14 tok/s |
| SDXL ROCm, 20 steps | ~45-55 sec | ~35-42 sec |
The 890M’s ~25% GPU advantage translates directly to faster AI inference — meaningful for interactive use.
Plex/Jellyfin Hardware Transcoding
| Task | 780M | 890M |
|---|---|---|
| 4K HEVC → 1080p H.264 | Works | Works |
| Simultaneous 4K streams | 2-3 | 3-4 |
| AV1 decode | Yes | Yes |
Functionally equivalent for most home server setups. The 890M handles one extra simultaneous transcode stream.
Proxmox/Docker Server Workloads
| Task | 780M | 890M | Difference |
|---|---|---|---|
| LXC containers | Equivalent | Equivalent | None |
| VM performance | Equivalent | Equivalent | None |
| iGPU passthrough | Both support | Both support | None |
For pure server workloads, both iGPUs are equivalent — the CPU cores matter more than the GPU here.
Which Platform Has Better Value?
| Platform | Mini PC Example | Price | 890M Upgrade Cost |
|---|---|---|---|
| Radeon 780M | Beelink SER9 | ~$320-380 | — |
| Radeon 890M | GEEKOM A9 Max | ~$520-620 | +$140-240 |
| Radeon 890M | X1 Pro-370 | ~$600-700 | +$220-320 |
The 890M costs $140-320 more than comparable 780M mini PCs. For a 25-30% GPU performance increase, that’s a significant premium.
The RAM upgradeability changes the calculation: 780M mini PCs like the SER9 use soldered LPDDR5X 32GB max. 890M mini PCs use DDR5 SO-DIMM upgradeable to 64GB. If you need 64GB RAM for large LLM models or heavy Proxmox, the 890M mini PC may be the only path — and the 890M GPU comes along for the ride.
Recommended Picks
Choose Radeon 780M Beelink SER9 if:
- Your workloads fit in 32GB RAM
- You run 7B-22B LLM models
- Budget is under $400
- Server workloads dominate GPU performance less critical
Choose Radeon 890M GEEKOM A9 Max if:
- You need 64GB RAM for 32B+ models or heavy Proxmox
- Gaming performance at 1080p matters to you
- You want the 25% GPU advantage for AI generation
- GEEKOM 3-year warranty is important
Choose Radeon 890M X1 Pro-370 if:
- You want USB4 for fast external storage
- You prioritize I/O options alongside AI performance
- You want the best all-around Strix Point mini PC
→ Minisforum AI X1 Pro-370 on Amazon
Who Should Skip This Comparison
Skip this comparison if your workload is purely server-side: Docker containers, Proxmox VMs, Plex with hardware transcoding, and home automation all perform identically on 780M and 890M machines. The iGPU generation doesn’t matter for these tasks — the CPU core count and RAM capacity are what limits you, not the graphics tier.
Also skip if you’ve already ruled out AMD platforms. Both iGPUs require ROCm for GPU-accelerated inference on Linux, and the AMD iGPU ecosystem for gaming has rougher driver edges than discrete Nvidia or Intel Arc alternatives. If you need the smoothest driver experience for Windows gaming, a discrete GPU machine is the better starting point.
Frequently Asked Questions
Does the Radeon 890M support ray tracing?
Yes. Both Radeon 780M and 890M support DirectX 12 Ultimate, which includes hardware ray tracing. Performance with ray tracing enabled drops significantly on both iGPUs — most games at 1080p become unplayable with RT enabled. Disable ray tracing on both platforms for playable frame rates.
Is ROCm better on 890M vs 780M?
Both support the same ROCm version 6.x on both. The 890M performs about 25% faster for ROCm compute due to more CUs and higher bandwidth — but the setup process and compatibility are identical.
What generation is Radeon 890M?
The 890M uses RDNA 3.5 architecture, which is a refinement of RDNA 3 used in the 780M. RDNA 3.5 adds efficiency improvements and some architectural refinements but is not a full generation leap. The performance difference ~25% comes primarily from more compute units 16 vs 12 rather than architectural advantages.
Can I use an external GPU with Radeon 780M or 890M mini PCs?
Some mini PCs support eGPU expansion. The GMKtec K11 780M includes Oculink for PCIe 4.0 x4 eGPU connection. The Minisforum X1 Pro-370 890M includes OCuLink and USB4, which support compatible external graphics enclosures. When using an eGPU, the internal iGPU is typically disabled in favor of the discrete external GPU.
Which is better for Plex transcoding — 780M or 890M?
Functionally equivalent. Both Radeon 780M and 890M support VA-API hardware transcoding in Plex and Jellyfin. The 890M can handle marginally more simultaneous transcode streams — 3-4 vs 2-3 for the 780M — but for most home server use cases, both are sufficient.