64GB max RAM · 24 bays · 1GbE
AINAS Score 43
No verified price

AI CapableUGREEN
Why AI Capable: AMD Ryzen Embedded R2514 (4C/8T Zen+, no NPU) with Radeon Vega 8, no GPU expansion path and 64GB max DDR4: older Ryzen embedded with no expansion is capable at best.
Updated
Can the UGREEN DXP2800 GT run local AI?
The UGREEN DXP2800 GT can run small local models, with limits. Its AMD Ryzen Embedded R2514 and 64GB memory ceiling suit 32B class models comfortably, and 70B class models at 4-bit with little headroom (Estimated), so treat it as a storage-first NAS that can host light AI tools.
There is no way to add a discrete GPU, so all inference runs on the CPU and basic integrated graphics.
Local LLM support is documented by the manufacturer.
The DXP2800 GT suits a buyer who wants a 2-bay NAS that can also run small local AI jobs in Docker. It has an AMD Ryzen Embedded R2514 with 4 cores, one 10GbE port, 2 M.2 NVMe slots and 2 drive bays. Memory is 8GB DDR4 SO-DIMM as shipped, expandable to 64GB, and ECC is supported if you fit ECC modules. It runs UGOS Pro with Docker and virtual machines. UGREEN has guides for Ollama and Open WebUI, and AI photo albums and Surveillance Center are vendor confirmed. The AINAS Score is 44. On the listed models, Llama 3.3 70B is the largest that fits at Q4 (Estimated), once memory is upgraded.
The main limit is that the DXP2800 GT has no GPU path. There is no PCIe slot, Thunderbolt or OCuLink, and no NPU, so all inference runs on the CPU. UGREEN lists local LLM support as limited for that reason. Anyone planning heavier local AI, such as larger models used interactively or image generation, will be held back and cannot add a card later. Those buyers should look for a system with a PCIe slot or OCuLink for an external GPU, or an NPU with a published TOPS figure. They should also check that the memory ceiling suits the model sizes they want to load.
Amazon UK
8GB
£349.99£459.99
Verified
In stock
Buy on AmazonProduct prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon.co.uk or Amazon.com, as applicable, at the time of purchase will apply to the purchase of this product.
Each capability shows how it was confirmed: Verified by AINAS, Vendor confirmed, Community confirmed or Unverified.
| Capability | Supported | Evidence | Note |
|---|---|---|---|
| Local LLM | Limited | Vendor confirmed | CPU-only inference via Ollama in Docker per UGREEN KB; no NPU. |
| Ollama | Yes | Vendor confirmed | UGREEN KB Docker Compose guide for Ollama/Open WebUI on UGOS Pro. |
| Open WebUI | Yes | Vendor confirmed | Official UGOS Pro Open WebUI guide. |
| RAG | Limited | Vendor confirmed | RAG via Open WebUI document chat only. |
| AI search | Limited | Vendor confirmed | Photos description and OCR search via Model Manager; document semantic search applicability per model not stated. |
| GPU expansion | No | Vendor confirmed | No PCIe slot, Thunderbolt or OCuLink. |
| Hardware acceleration | Limited | Unverified | Radeon Vega 8 iGPU; UGREEN does not document hardware transcoding or AI acceleration for GT; no NPU. |
| AI photos | Yes | Vendor confirmed | AI photo album: faces, places, objects, duplicates. |
| AI video | Limited | Vendor confirmed | Surveillance Center local AI detection (people, vehicle, package, pet); features vary by model. |
| Surveillance AI | Yes | Vendor confirmed | Surveillance Center supports DXP series. |
Calculated from the DXP2800 GT's memory and each model's published size at each quantization. Rows marked Estimated are calculations, not benchmarks. Check any model.
| Model | Best quantization | Runs? | Memory needed | Expected performance | Verification |
|---|---|---|---|---|---|
| Llama 3.2 3BLlama, 3.21B | Q8 | Yes | 13GB | Not measured | Estimated |
| Llama 3.1 8BLlama, 8.03B | Q8 | Yes | 19GB | Not measured | Estimated |
| Qwen3 8BQwen, 8.19B | Q8 | Yes | 19GB | Not measured | Estimated |
| Gemma 3 12BGemma, 11.8B | Q8 | Yes | 24GB | Not measured | Estimated |
| Qwen3 14BQwen, 14.8B | Q8 | Yes | 28GB | Not measured | Estimated |
| Phi-4 14BPhi, 14.7B | Q8 | Yes | 28GB | Not measured | Estimated |
| Mistral Small 3 24BMistral, 23.6B | Q8 | Yes | 38GB | Not measured | Estimated |
| Gemma 3 27BGemma, 27B | Q8 | Yes | 44GB | Not measured | Estimated |
| Qwen3 30B A3BQwen, 30.5B | Q8 | Yes | 47GB | Not measured | Estimated |
| Qwen3 32BQwen, 32.8B | Q8 | Yes | 50GB | Not measured | Estimated |
| DeepSeek-R1 32B (Qwen distill)DeepSeek, 32.8B | Q8 | Yes | 50GB | Not measured | Estimated |
| Llama 3.3 70BLlama, 70.6B | Q4 | Yes | 60GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | None fits | No | 89GB at Q4 | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q3 | Yes | 54GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q2 | Yes | 62GB | Not measured | Estimated |
One verdict per workload, from the DXP2800 GT's use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
The DXP2800 GT is an OK choice for local LLM work. UGREEN confirms CPU-only inference through Ollama in Docker, marks support as limited, and lists no NPU.
Private RAG is OK on the DXP2800 GT, with a narrow scope. UGREEN documents Open WebUI and Ollama guides, but RAG is limited to document chat in Open WebUI.
AI search is rated OK, with partial coverage. UGREEN confirms photo description and OCR search through Model Manager, but does not state whether document semantic search applies per model.
AI photos is OK, and it is a fully supported feature. UGREEN confirms an AI photo album covering faces, places, objects and duplicates on the DXP2800 GT.
Surveillance is rated OK on the DXP2800 GT. Surveillance Center supports the DXP series, and local AI detection for people, vehicles, packages and pets is limited, with features varying by model.
The DXP2800 GT is OK for a home lab. It runs Docker and virtual machines on UGOS Pro, has 2 M.2 NVMe slots, 10GbE, and memory expandable to 64GB.
Business use is rated OK. The DXP2800 GT supports ECC memory if you fit ECC modules, has 10GbE and a 2-year warranty, but offers only 2 drive bays.
Creative AI is rated OK, with clear limits. GPU expansion is not supported, and hardware acceleration on the Radeon Vega 8 is unconfirmed, so work stays on the CPU.
The DXP2800 GT ships with 8GB of DDR4 SO-DIMM 2666 and supports up to 64GB with ECC. Upgrading to the maximum lifts the model ceiling from small models up to around 7B parameters to 32B class models comfortably, and 70B class models at 4-bit with little headroom (both Estimated).
As a rule of thumb, a 4-bit model needs about 0.6GB per billion parameters plus room for context. After leaving around 8GB for the NAS operating system, 64GB leaves space for a model of roughly 93B parameters at most (Estimated).
ECC support protects against silent memory errors on a box that runs around the clock.
No. The DXP2800 GT has no PCIe slot or OCuLink port for a graphics card. Inference runs on the CPU, with AMD Radeon Vega 8 handling media tasks, so keep to small models.
With limits. The DXP2800 GT can host a small RAG stack such as AnythingLLM or Open WebUI with a compact embedding model, but memory and compute restrict how large the answering model can be. It suits a personal document library more than a team knowledge base.
Networking is 10GbE. That is ample for indexing large libraries and for several people querying at once.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| 8GB | 8GB | None | Diskless, 8GB DDR4, 64GB eMMC system storage. |
The closest UGREEN model by bay count and AINAS Score. Only the specifications that differ are listed.
The DXP2800 GT scores best for Business (53/100) and Home Lab (50/100). It suits small offices that need shared storage with private AI tools and room to grow.
It is a weaker fit for buyers who expect to add a graphics card later and for anyone who needs a lot of raw storage.
Other AI Capable systems closest in price.
Buying guides that list it
Yes, at Q4 only, and only with a memory upgrade (Estimated). Llama 3.3 70B needs 43GB at Q4 and 75GB at Q8, and the maximum is 64GB, so Q8 does not fit (Estimated). It ships with 8GB. The model runs on the CPU, with no GPU path or NPU.
Yes. It uses DDR4 SO-DIMM 2666 with 2 slots, ships with 8GB, and takes up to 64GB. The memory sits in slots rather than being soldered. Plan the upgrade if you want larger models, since the shipped 8GB limits what loads.
Yes, the platform supports ECC. It ships with a non-ECC 8GB module, so you must buy and fit ECC modules yourself. UGREEN's US FAQ confirms this. Check module compatibility before buying.
No. UGREEN confirms there is no PCIe slot, no Thunderbolt and no OCuLink, and the USB ports are not USB4. The only graphics hardware is the integrated Radeon Vega 8, and hardware acceleration on it is Unverified.
Yes. The DXP2800 GT has 2 M.2 NVMe slots beside its 2 drive bays, and UGOS Pro runs from 64GB of eMMC system storage, so both slots are free for flash. With only two bays, those slots matter for capacity as well as for fast storage behind the single 10GbE port. Model files and Docker containers can live on NVMe while the bays hold bulk data, and memory can grow from 8GB to 64GB alongside.
Yes. It runs UGOS Pro with Docker and virtual machines. UGREEN publishes a Docker Compose guide for Ollama and Open WebUI on UGOS Pro. Inference is CPU-only, and UGREEN rates local LLM support as limited.
Yes. Ollama support is documented by the manufacturer. See the capability table above for the source.
Yes. Open WebUI support is documented by the manufacturer. See the capability table above for the source.
Yes. AI photos support is documented by the manufacturer. See the capability table above for the source.
Yes. Surveillance AI support is documented by the manufacturer. See the capability table above for the source.
Partly. Hardware acceleration support is not yet verified. See the capability table above for the source.
It has 10GbE built in. The fastest port is 10GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.