We take cryptocurrency — so the coins you hold can buy real hardware.Pay in crypto, spend it on real hardware. Free insured shipping worldwide over $399. Plain, unbranded boxes, sent from Germany.Free insured shipping over $399. Your rate is locked for 30 minutes once checkout opens.Rate locked 30 minutes. Every item sealed, serial-checked and covered by a 24-month warranty.Sealed, 24-month warranty.

Saved items Sign in
Tower workstation with closed-loop liquid cooling, dark metal finish

Supermicro

Supermicro ARS-511GD-NB-LCC-01-G2 workstation 72-core 496GB LPDDR5X 252GB GPU

SKU CH-PMHJ-7DWJE MPN ARS-511GD-NB-LCC-01-G2

Deskside AI supercomputer with 72-core Grace CPU, 496GB LPDDR5X and 252GB GPU memory

  • TypeTower
  • Rack Height5U/Tower
  • Server Scalability1-Way
  • LEDsHDD activity LED Power status LED
  • Redundant Hot-swap PowerNo
  • AI FeaturesAI Accelerated
  • 72-core Grace CPU delivers data-centre AI compute
  • 496GB LPDDR5X unified memory handles trillion-parameter models
  • 252GB HBM3e GPU memory accelerates large-model training
  • Liquid-cooled 5U tower runs quietly at the desk
  • Two 1.92TB plus two 960GB NVMe drives store datasets fast

Deskside AI supercomputer

The Supermicro Super AI Station ARS-511GD-NB-LCC-01-G2 is a tower workstation built around the NVIDIA GB300 Grace Blackwell Ultra Superchip. It integrates a 72-core Grace CPU with a Blackwell Ultra GPU that provides 252GB of HBM3e video memory. The system delivers up to 20 PFLOPS of AI performance in a liquid-cooled chassis designed for quiet office use.

Coherent memory and local NVMe limits

The platform unifies 496GB of LPDDR5X system memory with GPU memory for up to 748GB of coherent capacity. Local storage consists of two 1.92TB and two 960GB M.2 NVMe PCIe 4.0 drives. This fixed configuration suits large-model workloads but cannot be expanded with additional drive bays or memory modules.

Replaces rack AI servers for desk development

Teams moving model training and agentic AI development out of the data centre gain a plug-and-play system that runs from a standard NEMA 5-20 outlet. The jump is worth it when data privacy, low latency iteration, and native NVIDIA software stack support outweigh the need for multi-node scaling or hot-swap redundancy.

Highlights

  • 72-core Grace CPU delivers data-centre AI compute
  • 496GB LPDDR5X unified memory handles trillion-parameter models
  • 252GB HBM3e GPU memory accelerates large-model training
  • Liquid-cooled 5U tower runs quietly at the desk
  • Two 1.92TB plus two 960GB NVMe drives store datasets fast

Specifications

BrandSUPERMICRO
SeriesSuper AI Station
ModelARS-511GD-NB-LCC-OTO-1
TypeTower
Rack Height5U/Tower
Server Scalability1-Way
LEDsHDD activity LED
Power status LED
Redundant Hot-swap PowerNo
AI FeaturesAI Accelerated
CPU TypeNVIDIA Grace
CPU SpeedNVIDIA Grace CPU on NVIDIA GB300 Grace Blackwell Ultra Superchip
CPU SocketSuperchip module
Number of CPU Cores72-Core Processor
# of Processors Installed1
MAX Processors1
Installed Memory Size496GB
Memory TypeLPDDR5X
Servers_Graphic Installed1
GPU TypeNVIDIA Blackwell Ultra (B300) GPU
Video Memory252GB HBM3e
LAN PortsDual-port QSFP 400 GbE LAN via NVIDIA ConnectX-8 SuperNIC
Dedicated 1GbE BMC LAN
10GbE LAN
Expansion Slots1 PCIe 5.0 x16 FHFL double-width slot
2 PCIe 5.0 x8 (in x16) FHHL slots
Expansion Bays4 M.2 PCIe 5.0 NVMe
Operating SystemUbuntu
Power Characteristics1 x 1600W Titanium Level (94%) power supply
TemperatureOperating Temperature: 10°C to 25°C (50°F to 77°F)
Non-operating Temperature: -40°C to 60°C (-40°F to 140°F)
HumidityOperating Relative Humidity: 8% to 90% (non-condensing)
Non-operating Relative Humidity: 5% to 95% (non-condensing)
Dimensions17.9" x 8.6" x 27.6"
Weight88 lbs.
Shipping weight50.98 kg
Package size955 × 655 × 389 mm

Questions about this item

Which memory configuration should I select for the 496GB LPDDR5X system?

The workstation ships with four 128GB LPDDR5X SOCAMM modules totalling 496GB; this is the only memory configuration offered for the Grace CPU superchip.

What does the 20 PFLOPS AI performance figure mean when the system is running locally?

It represents up to 20 petaFLOPS of AI compute from the integrated GB300 Grace Blackwell Ultra superchip, enabling trillion-parameter model work on the desk.

What interface do the four M.2 drives use and what speed does it cap at?

Two 1.92TB and two 960GB M.2 drives run on PCIe 4.0 NVMe, limiting each drive to the 4.0 lane bandwidth.

Also in this aisle

People compared these

Same shelf, same checkout — eight coins and a 30-minute rate lock.