SupermicroARS-511GD-NB-LCC-01-G2
The Super AI Station puts the NVIDIA GB300 superchip, a 72-core Grace CPU and a Blackwell Ultra B300 GPU into a liquid-cooled tower, with 748 GB of memory shared between the two.
Tower · Gold Series G2The exact configuration
- Warranty
- 3 years parts and labour
- On-site engineer
- Mandatory on this model: next business day, for 3 years
A data-centre AI node for one team, right by the desk
It is designed for teams that build agents and fine-tune large models and currently depend on booking hours on a cluster or in the cloud. The GPU has 252 GB of HBM3e (memory stacked next to the chip itself) and the CPU another 496 GB of LPDDR5X, low-power memory that here includes error correction; the two are joined by NVLink-C2C, NVIDIA's direct CPU-to-GPU link, so the model sees a single 748 GB space. NVIDIA talks of models of up to a trillion parameters quantised to 4 bits, which take about 500 GB and fit in that space, and two 400 GbE ports on a ConnectX-8 connect it to storage or to another station.
The CPU is Arm, so check that your container images and libraries exist for arm64; that is the norm across the NVIDIA stack today but not always for in-house tools. It also has one integrated GPU (with a slot for one more PCIe card) and a single 1,600 W supply with no redundancy, which makes it a working tool for a team; serving many users calls for a rack server.
What we do
We deliver it with the operating system and the NVIDIA stack installed, with containers for fine-tuning and vLLM to serve whatever the team is testing, and connect it to your network and storage through the 400 GbE ports. This model is only sold with next-business-day on-site engineer cover; we are your first point of contact when something goes wrong.
Compared with the rest of the family
| Model | What it has | When to choose it |
|---|---|---|
| SYS-822GS-NB3RT-01-G2 | HGX B300 8-GPU · 2 TB | The HGX B300 with Intel; the choice if your standard is Xeon. |
| AS-8126GS-NB3RT-01-G2 | HGX B300 8-GPU · 3 TB | Same HGX B300 with AMD, 3 TB and a separate storage network. |
| AS-5126GS-TNRT-01-G2 | 2x RTX PRO 6000 · 1.5 TB | Two GPUs today, space for eight and memory to spare. |
| SYS-212GB-FNR-01-G2 | 2x RTX PRO 6000 · 512 GB | The most compact: 2U for a two-GPU service. |
| SYS-422GA-NRT-01-G2 | 4x RTX PRO 6000 · 1 TB | Four GPUs that MIG splits into up to sixteen instances for several teams. |
| SYS-422GA-NRT-02-G2 | 8x RTX PRO 6000 · 1.5 TB | The most PCIe can do: eight GPUs and 400 GbE networking. |
| SYS-C542i-11302U | RTX 50-ready · 32 GB | Development tower without a standard GPU; you pick the card. |
What the delivery includes
From the factory
- Configuration assembled and tested by Supermicro
- Warranty: 3 years parts and labour, with an on-site engineer next business day: mandatory on this model and always included in the quote
From us
- Operating system, drivers and AI stack installed
- Testing with your workload before delivery
If you need it
- 24×7 on-call support with SLA
How we build it
What runs on top of these servers, explained in detail:
Support and training for your team
Questions about the ARS-511GD-NB-LCC-01-G2
How does it differ from an HGX B300 server?
The Super AI Station has one B300 GPU with 748 GB of memory shared with the CPU, in a liquid-cooled tower with a 1,600 W supply. An HGX B300 such as the SYS-822GS-NB3RT-01-G2 has eight, joined by NVLink, in an 8U chassis that needs a properly prepared data centre. The station is for a team developing and fine-tuning models; the HGX is for training them at scale or serving them to many users.
Does it need a data centre, or can it sit in an office?
It runs as a tower and the liquid cooling is a closed loop inside the unit itself, with no external water connections, so it can sit in a technical room or an office with a power socket suited to its 1,600 W supply. Supermicro also offers it in a 5U format for rack mounting if you would rather keep it in the data centre.
Does my software run on Arm?
The NVIDIA stack (drivers, CUDA, its containers and vLLM) is published for arm64, so the usual tools work. What needs checking is in-house images and libraries built only for x86: we review them with you before delivery and, if any are missing, help you rebuild them or find an alternative.
Is the ARS-511GD-NB-LCC-01-G2 right for what you're building?
Tell us which models you'll run and who will be working on it, and we'll come back with the price, a confirmed lead time and what it would take to get it running.
Blog!
Our mission
SIXE ensures that your infrastructures based on IBM®, Red Hat®, Canonical®, and SUSE® technologies remain stable, secure, and optimized for as long as you need them.
At SIXE, we are committed to sustainability. That’s why we help you design, implement, and maintain durable and efficient technological solutions.
