Skip to content
All hardware
GMKtecAMD Strix HaloDesk tier · £2.8k – £8.5k

EVO-X5 Pro

The first 192 GB Strix-class mini-PC you can order in the UK — AMD's Gorgon Halo silicon, with up to 160 GB for the GPU.

Price

£5,299

gmktec.uk 192 GB / 2 TB (4 TB £5,549) · ships from China, VAT basis not stated · US list $6,799

verified 2026-11 · supply & lease options in every proposal

192 GB
Unified memory

up to 160 GB to the GPU

~273 GB/s
Memory bandwidth

LPDDR5X-8533

55 TOPS
XDNA 2 NPU

Ryzen AI Max+ PRO 495

2× 10 GbE
Networking

+ 2× USB4 v2 (80 Gb/s)

The machine

AMD's Ryzen AI Max 400 series ('Gorgon Halo') is the Strix Halo refresh: the Ryzen AI Max+ PRO 495 keeps 16 Zen 5 cores and a 40-CU GPU (Radeon 8065S) but raises the memory ceiling to 192 GB of LPDDR5X-8533 — about 273 GB/s, with up to 160 GB allocatable to the GPU. GMKtec's EVO-X5 Pro is the one you can buy in the UK today.

The 160 GB GPU window is the point: Qwen3.8-Flash-Next (~111 GB at 4-bit) fits with real context headroom, and a 120B-class model can sit resident alongside a small triage model. Bandwidth barely moves (≈273 vs 256 GB/s), so per-stream speed stays Strix-class — the gain is capacity, not tokens per second.

It is £5,299 for 192 GB / 2 TB on GMKtec's UK store (shipped from China; VAT basis not stated). Dual 10 GbE and USB4 v2 make it the best-networked box in the family; gpt-oss-120b benchmarks on this silicon weren't published at time of writing.

vs its siblings: The first 192 GB Gorgon Halo box with UK supply.

Memory, to scale

192 GB model-visible · bandwidth is the speed limit

Unified LPDDR5X

192 GB · ~273 GB/s

LPDDR5x

For scale

DGX Spark — 128 GB @ 273 GB/s

RTX PRO 6000 — 96 GB @ 1.79 TB/s

Mac Studio M5 Ultra — 512 GB @ 1.2 TB/s

DGX Station GB300 — 748 GB coherent

Capability

What it actually runs

Declared from research and benchmarks, not computed marketing — tokens-per-second figures are cited where a real measurement exists.

  • Qwen3.8-Flash-NextQ4 (~111 GB)with headroomfits the 160 GB GPU window with context headroom
  • Qwen3.5 122BInt4 (~79 GB)with headroomroom for long context or a second model
  • gpt-oss-120bMXFP4with headroomStrix-class speed — Gorgon Halo benchmarks not yet published
  • Qwen3.6 35BQ4with headroomrun it alongside a 120B model for triage
Specification

The full sheet

Compute

APU
AMD Ryzen AI Max+ PRO 495 — 16× Zen 5 (32 threads) to 5.2 GHz
GPU
Radeon 8065S — 40 CU RDNA 3.5
NPU
XDNA 2 — 55 TOPS

Memory

Unified memory
192 GB LPDDR5X-8533 (soldered)
GPU allocation
Up to 160 GB

Storage & I/O

Storage
3× M.2 (to 24 TB)
Networking
2× 10 GbE
Ports
2× USB4 v2 (80 Gb/s) + 2× USB4
Management
AMD DASH remote management · TPM 2.0

Where it shines

  • 192 GB — the most unified memory under £6k in the catalogue
  • Up to 160 GB for the GPU: 180B-class MoE models at 4-bit with headroom
  • Dual 10 GbE and USB4 v2 — the best networking in the family
  • AMD DASH remote management and TPM 2.0 for office deployment

The trade-offs

  • Bandwidth barely moves (~273 GB/s) — more capacity, same speed per stream
  • Ships from China; UK VAT and import treatment not stated
  • New silicon — few published LLM benchmarks yet
  • Consumer-grade warranty and support

Buy this box for

Running a bigger model, or two models side by side, on one deskPilots that need 180B-class capability on a budgetSmall-team private inference where capacity matters more than speed
The platform

Understanding AMD Strix Halo

Ryzen AI Max+ 395 · Max+ PRO 495

AMD's Ryzen AI Max+ 395 ('Strix Halo') puts 16 Zen 5 cores, a 40-CU Radeon 8060S GPU and up to 128 GB of unified LPDDR5X-8000 on one package. Framework, HP, GMKtec, Beelink, Corsair and AMD itself all ship it as a compact desktop — and the memory shortage has pushed UK prices for 128 GB machines to £2,830 (GMKtec) to £5,304 (HP).

It is a mixture-of-experts machine: gpt-oss-120b runs at 30–55 tok/s and even Qwen3-235B squeezes in at 3-bit (~11 tok/s), but dense 70B models crawl at ~5 tok/s because 256 GB/s of bandwidth is the ceiling. The software stack (Vulkan / ROCm / llama.cpp / LM Studio) is genuinely usable in 2026 but remains a step behind CUDA for production serving and fine-tuning.

The refresh has arrived: AMD's Ryzen AI Max 400 series ('Gorgon Halo') — the Max+ PRO 495 with a Radeon 8065S — lifts the ceiling to 192 GB of LPDDR5X-8533 (about 273 GB/s, up to 160 GB for the GPU). GMKtec's EVO-X5 Pro is £5,299 in the UK and Framework's 192 GB board is £6,799 on pre-order, shipping in December: more capacity, much the same speed.

Sources & verification

Specifications and prices verified 2026-11 against the sources below. The memory shortage is repricing this market monthly — we re-verify at quote.

Compare against the rest of the catalogue, or have us spec it against your workloads.

Back to catalogue
Make it real

Spec the EVO-X5 Pro for your workload

We'll size it against your real token volumes, quote supply and lease options, and stand up the models with evals that prove they do the job — or tell you honestly that a different box (or the API) wins.