48GB max RAM · 12 bays · 2 x 10GbE
AINAS Score 40
$3,449.99
Amazon US · Verified
Ships from the US; delivery and import charges shown at checkout

AI CapableASUSTOR
Why AI Capable: 2017 Intel Atom C3538 (4C/4T, no iGPU, no NPU) is not a current CPU, and its PCIe 3.0 x8 (x4 link) slot in a 1U chassis is documented only for network cards; 128GB memory ceiling but CPU-only inference via ASUSTOR's Ollama app.
Updated
Can the ASUSTOR Lockerstor 4RD (AS6504RD) run local AI?
Only lightly. Expect 70B class models at higher-quality quantization, or large mixture-of-experts models at 4-bit (Estimated) on the ASUSTOR Lockerstor 4RD (AS6504RD): Intel Atom C3538 does the work with up to 128GB of memory, which is fine for experiments rather than a daily assistant.
Without any GPU option, the processor carries the whole inference load.
Local LLM support is documented by the manufacturer.
The Lockerstor 4RD suits a buyer who wants a 1U rack NAS with four bays and a way to try private AI services on the NAS itself. It has an Intel Atom C3538 with 4 cores, 8GB of DDR4-2133 SO-DIMM memory in the box and four memory slots that take up to 128GB. Docker and virtual machines are both supported. ASUSTOR's Ollama and Open WebUI package is vendor confirmed and installs through Docker. On memory alone, Mistral Large 2 123B is the largest model listed that loads at Q4 (Estimated). Networking is 2 x 2.5GbE and 2 x 1GbE, and a PCIe slot takes a network card. The warranty is 3 years. It scores 40 on the AINAS Score, which reflects a capable but CPU-bound system.
The main limit is the lack of a GPU path. The Lockerstor 4RD has no integrated graphics and no NPU, and its PCIe slot is documented for network cards only. ASUSTOR states that its NAS models do not feature a discrete GPU and recommends an external GPU host for inference. Anyone who wants to run larger models for chat, coding or creative work will find that memory capacity does not make up for CPU-only processing. Look instead for a system with a PCIe slot documented for graphics cards, an OCuLink or Thunderbolt connection for an external GPU, or a current CPU with an NPU or integrated graphics. Its Atom C3538 dates from 2017, so the CPU is not a current part either.
No verified price
We have not verified a listing for the ASUSTOR Lockerstor 4RD (AS6504RD) at a retailer we track. The specifications below stand on their own; a price appears here as soon as a listing is confirmed.
Product 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 | ASUSTOR App Central lists Ollama + Open WebUI (v0.11.4.r01, ADM 5.0+) for the AS65 series. CPU-only inference; 4-core 2017 Atom without GPU; ASUSTOR recommends an external GPU host for inference. |
| Ollama | Yes | Vendor confirmed | Ollama + Open WebUI app lists AS65 as a supported series; installs via Docker. |
| Open WebUI | Yes | Vendor confirmed | Bundled with the Ollama + Open WebUI App Central package. |
| RAG | Limited | Vendor confirmed | ASUSTOR markets NAS as RAG knowledge base storage with vector DBs (Chroma, pgvector, Weaviate) via Docker; Open WebUI includes RAG. ASUSTOR recommends external compute for inference. |
| AI search | No | Unverified | No AI or semantic file search feature found in ADM 5.1 material. |
| GPU expansion | No | Vendor confirmed | PCIe 3.0 x8 slot (x4 link) in a 1U chassis is documented for Ethernet cards only; no iGPU. ASUSTOR states its NAS models do not feature a discrete GPU. |
| Hardware acceleration | No | Unverified | Intel ARK lists no processor graphics for Atom C3538; ASUSTOR spec lists multimedia output N/A. No NPU. |
| AI photos | Yes | Vendor confirmed | Photo Gallery 4.2 (ADM 5.1.1+) lists AS65 series; ADM 5.1 adds facial recognition grouping powered by locally run AI. |
| AI video | No | Unverified | No AI video feature found in ADM documentation. |
| Surveillance AI | Yes | Vendor confirmed | Surveillance Center 2.8.4 (app 1663) and Surveillance Center 3 (app 1661) list AS65; up to 64 IP cameras per ASUSTOR spec page licence table. |
Calculated from the Lockerstor 4RD (AS6504RD)'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 | Q8 | Yes | 98GB | Not measured | Estimated |
| Llama 4 Scout 17Bx16ELlama, 109B | Q4 | Yes | 89GB | Not measured | Estimated |
| Qwen2.5 72BQwen, 72.7B | Q8 | Yes | 101GB | Not measured | Estimated |
| Mistral Large 2 123BMistral, 123B | Q5 | Yes | 112GB | Not measured | Estimated |
One verdict per workload, from the Lockerstor 4RD (AS6504RD)'s use-case scores and capability rows. Unknown means we cannot confirm it, not that it probably works.
Local LLM is rated OK. Ollama is vendor confirmed and 128GB of memory is possible, but support is marked limited because inference is CPU-only on a 4-core Atom and ASUSTOR recommends an external GPU host.
Private RAG is rated OK, with the NAS best used as the home for the documents. RAG support is marked limited: Docker can host Chroma, pgvector and Weaviate and Open WebUI includes RAG, but ASUSTOR recommends external compute for inference.
AI search is rated poor for this system. No AI or semantic file search feature was found in ADM 5.1 material, and the row is marked Unverified, so do not buy the Lockerstor 4RD expecting it.
AI photos is rated OK, and the support is vendor confirmed. Photo Gallery 4.2 needs ADM 5.1.1 or later and lists the AS65 series, while ADM 5.1 adds facial recognition grouping powered by locally run AI.
Surveillance is rated OK for camera recording and management. Surveillance Center 2.8.4 and Surveillance Center 3 both list the AS65 series, with up to 64 IP cameras on the ASUSTOR licence table, vendor confirmed, and four bays hold the footage.
Home lab is rated OK for container and virtual machine work in a rack. Docker and VirtualBox are both supported, memory expands to 128GB across four slots, and the PCIe slot takes a network card, though the CPU has 4 cores.
Business is rated OK for a small office rack file server. Running ADM 5, it is 1U with four bays, a 250W power supply, a 3 year warranty and four network ports, but the memory is not ECC.
Creative AI is rated OK, with a clear ceiling. There is no GPU expansion, hardware acceleration is marked Unverified, local LLM support is limited to CPU-only inference, and the PCIe slot is documented for network cards only.
Plan on upgrading. 8GB is fitted as standard, while the Lockerstor 4RD (AS6504RD) accepts 128GB of DDR4-2133 SO-DIMM, which is the difference between small models up to around 7B parameters and 70B class models at higher-quality quantization, or large mixture-of-experts models at 4-bit (Estimated).
Our estimate for the biggest model that fits in 128GB is about 200B parameters at 4-bit, after setting aside 8GB for the system. Long context windows reduce that figure (Estimated).
A graphics card is not an option on the Lockerstor 4RD (AS6504RD), which has neither a PCIe slot nor OCuLink. Inference runs on the CPU, so keep to small models.
As the storage half of a RAG setup it works well; as the model host it is limited. Small embedding models run fine on the Lockerstor 4RD (AS6504RD), while the answering model needs more compute than it has.
For network access it has 2 x 2.5GbE + 2 x 1GbE. Enough for light use; large imports will take a while.
| Configuration | RAM | SSD | Notes |
|---|---|---|---|
| 8GB | 8GB | None | Diskless; 1x 8GB DDR4 pre-installed, 4 SO-DIMM slots. |
The closest ASUSTOR model by bay count and AINAS Score. Only the specifications that differ are listed.
Best fit: anyone building private search and question answering over their own documents. That is where the Lockerstor 4RD (AS6504RD) scores highest (RAG 46/100), followed by Local LLM (45/100).
Less suited to buyers who expect to add a graphics card later and teams moving large files over the network every day.
Other AI Capable systems closest in overall score.
Yes on memory alone, at Q4 (Estimated). The model needs 73GB at Q4 and the system takes up to 128GB, though it ships with 8GB. At Q8 it does not fit (Estimated), as it needs 130GB. This is a memory calculation, not a benchmark. Inference would be CPU-only on a 4-core Atom, and ASUSTOR recommends an external GPU host.
Yes. It has four DDR4-2133 SO-DIMM slots and a maximum of 128GB, and 8GB is installed as shipped. ASUSTOR's spec does not state ECC support. ASUSTOR's compatibility list covers modules up to 32GB for the AS65R series, which matches four 32GB modules for 128GB. Choose modules from that list.
No. The single PCIe 3.0 x8 slot (x4 link) is documented for network cards only, there is no integrated graphics, and ASUSTOR states its NAS models do not feature a discrete GPU. There is also no OCuLink, Thunderbolt or USB4 port for an external GPU. Buyers who need GPU inference should look for a graphics-capable slot or an external GPU connection.
Yes, once the memory is raised (Estimated). Llama 3.3 70B is 43GB at Q4 and 75GB at Q8, and both fit within the 128GB ceiling (Estimated). The system ships with 8GB, so reaching that point means filling the four slots with 32GB modules from ASUSTOR's compatibility list. This is a memory calculation, not a benchmark. Inference runs on the 4-core Atom C3538 alone, and ASUSTOR recommends an external GPU host for inference.
Yes. Docker is supported, and ASUSTOR's Docker Engine package lists the AS65 series. The Ollama and Open WebUI package installs through Docker and is vendor confirmed. Virtual machines are supported too, through VirtualBox on ADM 5.1.3 or later. The operating system is ADM 5. ASUSTOR also describes running vector stores for RAG through Docker.
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.
No. Hardware acceleration support is not yet verified. See the capability table above for the source.
It has 2 x 2.5GbE + 2 x 1GbE built in, and the 1x PCIe 3.0 x8 slot (x4 link), for network cards can take a faster network card. The fastest port is 2 x 2.5GbE.
Every field records its source. Specifications were last checked on . Spot an error? Send a correction with a manufacturer link.