Skip to main content
PromptQuorumBuilt for humans. Structured for AI.
Home/Local LLMs
Overview & Reference

GMKtec EVO-X5 Pro vs DGX Spark: Specs, Price and Verdict

·8 min readBy · Founder of PromptQuorum · Discovery engine for open-weight & open-source AI

The GMKtec EVO-X5 Pro is a Ryzen AI Max+ PRO 495 mini PC with 192 GB of unified memory, which is 64 GB more than the 128 GB DGX Spark, at a list price roughly $2,000 above the Spark's $4,699 Founders Edition. It lists the same 273 GB/s memory bandwidth as the Spark but no CUDA and no 200 Gbps fabric, and we found no tokens-per-second figure on GMKtec's launch post or product page, so its speed advantage or disadvantage is unproven.

Key Takeaways

  • Memory: 192 GB of soldered LPDDR5X-8533 on the EVO-X5 Pro against 128 GB on the DGX Spark; GMKtec says up to 160 GB can be allocated to graphics.
  • Speed: both machines list 273 GB/s of memory bandwidth, and we found no tokens-per-second figure on GMKtec's launch post or product page, so we cannot say which is faster.
  • Price: GMKtec's page shows $6,799 for 192 GB / 2 TB; the Spark Founders Edition is $4,699 per NVIDIA, so the EVO-X5 Pro has a list price about $2,100 higher at the cited dates.
  • Software: the EVO-X5 Pro ships with Windows 11 Pro and supports Linux; the Spark runs only NVIDIA's DGX OS and offers CUDA.
  • Networking: two 10GbE ports on the EVO-X5 Pro versus a 200 Gbps ConnectX-7 adapter on the Spark for linking boxes.

The GMKtec EVO-X5 Pro is a mini PC built around the AMD Ryzen AI Max+ PRO 495 with 192 GB of soldered LPDDR5X-8533, launched by GMKtec on 2026-09-28. NVIDIA's DGX Spark has 128 GB and a lower list price at the cited dates, but the two machines list the same 273 GB/s memory bandwidth, so the real questions are memory size, price, software and networking. This page compares published specifications and prices; PromptQuorum has not tested either machine, and no independent LLM speed test of the EVO-X5 Pro was available to us.

GMKtec EVO-X5 Pro vs DGX Spark: Specs, Price and Verdict

What the GMKtec EVO-X5 Pro Is

The EVO-X5 Pro is a mini PC from GMKtec, built on AMD's Ryzen AI Max+ PRO 495 processor and sold only with 192 GB of memory. GMKtec's launch post, dated 2026-09-28, describes a 16-core Zen 5 chip, a Radeon 8065S graphics engine with 40 compute units, and an XDNA 2 NPU rated up to 55 TOPS. Notebookcheck reported the launch the same day.

GMKtec states that the machine runs language models of up to 320 billion parameters fully offline, and adds that actual compatibility and performance depend on the model, software, quantization, workload and system configuration. The company's earlier design target was 300 billion. We treat both numbers as a vendor claim, because we found no speed figures from GMKtec with them.

Specifications

Rows below are taken from GMKtec's launch post or official product page unless the Detail cell says otherwise.

Spec
Detail
CPUAMD Ryzen AI Max+ PRO 495, 16 Zen 5 cores / 32 threads, up to 5.2 GHz
GPURadeon 8065S, 40 compute units, RDNA 3.5
NPUXDNA 2, up to 55 TOPS
Memory192 GB LPDDR5X-8533, onboard, not upgradeable
GPU-addressable memoryUp to 160 GB, per GMKtec
Memory bandwidthUp to 273 GB/s, per the launch post
Storage2 TB or 4 TB PCIe 4.0 SSD; three more M.2 slots; up to 24 TB total; SATA not supported
NetworkingTwo 10GbE RJ45 ports, Wi-Fi 7, Bluetooth 5.4
PortsFront: 2x USB4 (40 Gbps), 2x USB 3.2 Gen2, audio. Rear: HDMI 2.1, 2x USB4 v2 (80 Gbps), 2x USB 2.0, audio out
Cooling and modesVapor chamber, three fans; Silent 54 W, Balanced 85 W, Performance 120 W sustained (160 W peak)
Power19.5 V / 16.9 A (330 W) adapter; USB-C power input not supported
Size and weight221 x 80 x 195 mm, about 2.5 kg
Operating systemWindows 11 Pro pre-installed; Linux supported
SecuritydTPM 2.0, fingerprint reader, AMD DASH management
Warranty1 year, plus 2 years of replacement cover for RAM, SSD and adapter

Sources: GMKtec launch post and product page, read 2026-11-13. Parts of both long pages were read in summary form, so check the live page before relying on a single row. The 273 GB/s figure appears in the launch post only; it matches the arithmetic for a 256-bit bus at 8533 MT/s.

EVO-X5 Pro vs DGX Spark

The two machines list identical memory bandwidth, so the gap is capacity, price, software and networking rather than a stated speed difference. NVIDIA's figures come from its DGX Spark product page as recorded on our DGX Spark page.

Spec
EVO-X5 Pro
DGX Spark
ProcessorRyzen AI Max+ PRO 495, 16 coresGB10, 20 Arm cores
Memory192 GB LPDDR5X-8533128 GB LPDDR5x
Bandwidth273 GB/s (GMKtec)273 GB/s (NVIDIA)
GPU softwareAMD ROCm and Vulkan stackCUDA
Operating systemWindows 11 Pro or LinuxDGX OS only
Fast networking2x 10GbE200 Gbps ConnectX-7 plus 10GbE
Power supply330 W240 W
Size and weight221 x 80 x 195 mm, 2.5 kg150 x 150 x 50.5 mm, 1.2 kg
US price$6,799 (2 TB)$4,699 (Founders)

The "GPU software" row for the EVO-X5 Pro describes AMD's general platform, not a statement on GMKtec's product pages, which list no software stack. NVIDIA's 128 GB unit has also been reported at $6,950 by ServeTheHome (2026-10-03), so the price gap depends on the Spark listing you compare.

Price and Availability

GMKtec's launch post and product page show different US prices, so none of the figures below should be treated as the current price. Memory costs have moved prices repeatedly in 2026; open the live listing before ordering.

Configuration
US price
Source and date
Status
192 GB / 2 TB$6,799GMKtec product page, read 2026-11-13Regular and sale price; not marked sold out
192 GB / 4 TB$6,799GMKtec product page, read 2026-11-13Variant marked sold out or unavailable
192 GB / 2 TB$6,399 super early bird, $6,599 early bird, $6,799 regularGMKtec launch post, 2026-09-28Launch promotions; not shown on the product page
192 GB / 4 TB$6,699 super early bird, $6,899 early bird, $7,099 regularGMKtec launch post, 2026-09-28Launch promotions; not shown on the product page

The product page shows the same $6,799 for both storage sizes while the launch post lists $7,099 for 4 TB; we do not know which is current. Reports on the launch quote $6,599 and $6,899 early-bird prices, for example Notebookcheck (2026-09-28). We have not repeated GMKtec's discount code because it may have expired.

What Independent Reports Show

We found no independent LLM-speed test of the EVO-X5 Pro that we could read, and none comparing it with the DGX Spark. A TechRadar article about the machine was listed, but only its headline loaded for us, so we do not rely on it. A Tom's Hardware piece on the launch pricing was listed, and again only the headline loaded, so we do not rely on it.

Notebookcheck's launch report is a news item rather than a review: it lists specifications and prices and contains no benchmarks, noise, power or temperature measurements. We found no tokens-per-second figure on GMKtec's own pages. A GMKtec blog headline mentions 75 tokens per second, but we could not read the post, so we do not use the figure.

For an independent comparison of a Strix Halo machine against the DGX Spark, see our category comparison, which covers AMD's Ryzen AI Halo rather than this unit. Its results do not transfer to the EVO-X5 Pro, which uses a different chip.

Who It Suits and Who Should Skip It

  • Suits: buyers who need 192 GB of unified memory in one box and prefer Windows or Linux to NVIDIA's DGX OS.
  • Suits: users who want two 10GbE ports and USB4 v2 for a fast local network or external storage.
  • Skip: developers who need CUDA, fine-tuning on NVIDIA tooling or a 200 Gbps link between two boxes; compare the DGX Spark.
  • Skip: anyone who only needs 128 GB; compare the spec sheets of other Strix Halo mini PCs and the Spark.
  • Skip: buyers who need an independently measured tokens-per-second figure for this exact unit before paying, because we found none.

Alternatives in the Same Class

  • DGX Spark: NVIDIA's CUDA system with 128 GB, 200 Gbps networking and, at the cited dates, a lower list price than the EVO-X5 Pro; see the category comparison.
  • GMKtec EVO-X2: GMKtec's earlier Ryzen AI Max+ 395 box with up to 128 GB and a single 2.5G Ethernet port.
  • Ryzen AI Halo: AMD's own 128 GB Ryzen AI Max+ 395 system with an independent Spark comparison.
  • Framework Desktop: a modular case with a 192 GB option on pre-order.
  • Minisforum MS-S1 MAX: 128 GB with dual 10GbE and a PCIe x16 slot (x4 electrically) per Minisforum and reviews.

Frequently Asked Questions

Can I upgrade the memory later?

No. GMKtec lists the 192 GB of LPDDR5X as onboard and non-upgradeable. Storage can be expanded through the additional M.2 slots, up to 24 TB in total.

How much of the 192 GB can the GPU use?

GMKtec says up to 160 GB can be allocated as graphics memory. It is shared system memory, not a separate pool, so what remains is left for the operating system and applications.

Does the EVO-X5 Pro run 320B-parameter models?

GMKtec states it supports fully offline inference of up to 320 billion parameters, depending on model, software, quantization and configuration. On GMKtec's launch post and product page we found no speed figures (a blog headline mentions 75 tokens per second, which we could not verify), and we did not test it.

Is the EVO-X5 Pro faster than the DGX Spark?

We cannot say. Both list 273 GB/s of memory bandwidth, and we found no independent speed test comparing them. NVIDIA's CUDA software and AMD's stack also differ, which can change results on the same model.

Which operating systems does it support?

GMKtec lists Windows 11 Pro pre-installed and Linux as supported. The DGX Spark runs only NVIDIA's DGX OS.

Sources

A Note on Third-Party Facts

This article references third-party AI models, benchmarks, prices, and licenses. The AI landscape changes rapidly. Benchmark scores, license terms, model names, and API prices can shift between the time of writing and the time you read this. Before making deployment or compliance decisions based on this article, verify current figures on each provider’s official source: Hugging Face model cards for licenses and benchmarks, provider websites for API pricing, and EUR-Lex for current GDPR and EU AI Act text.

Run PromptQuorum with a local LLM, your own API keys, or both — you pick the backend.

Download the PromptQuorum Beta →

← Back to Local LLMs

MCP