Supermicro
Supermicro GPU-XGMIMI210-4P Radeon MI210 64GB HBM2 AI accelerator card
64 GB HBM2 AI accelerator with 6,656 stream processors and 181 TFLOPs peak FP16 performance
- 6656 stream processors deliver massive parallel compute
- 104 compute units scale HPC and AI workloads
- 64 GB HBM2 memory feeds data-intensive models
- 181 TFLOPs FP16 peak accelerates training and inference
- 300 W dual-slot design fits standard server chassis
Supermicro MI210 64GB HBM2 AI Accelerator
The Supermicro GPU-XGMIMI210-4P is a new AMD Radeon Instinct MI210 accelerator built for enterprise HPC and AI workloads. It delivers 64 GB of HBM2 memory and 6,656 stream processors organised in 104 compute units. Peak half-precision throughput reaches 181 TFLOPs at a 1,700 MHz boost clock.
Suits single-server AI and HPC.
This card targets researchers and enterprises running large-model training or inference on a single server. The 300 W dual-slot design fits standard PCIe slots. Buyers who need multi-accelerator scaling or workloads that exceed 64 GB memory should evaluate alternative configurations.
64 GB HBM2 memory decides fit
The 64 GB HBM2 frame buffer is the single specification that determines whether this accelerator matches a workload. Models and datasets that fit within this capacity benefit from high-bandwidth memory; larger problems will require cards with more memory or distributed solutions.
Highlights
- 6656 stream processors deliver massive parallel compute
- 104 compute units scale HPC and AI workloads
- 64 GB HBM2 memory feeds data-intensive models
- 181 TFLOPs FP16 peak accelerates training and inference
- 300 W dual-slot design fits standard server chassis
Specifications
No published specifications for this item yet.
Questions about this item
How much power and heat does the 300 W MI210 produce when it runs AI workloads all day?
The card draws up to 300 W under full load, so sustained HPC or training jobs will generate significant heat that requires adequate server airflow and cooling capacity.
What server infrastructure must already be in place for this dual-slot, full-height accelerator to operate?
A PCIe x16 slot with sufficient physical clearance, a power supply able to deliver 300 W to the card, and chassis airflow rated for high-density compute modules are required.
Will the MI210 work alongside older CPUs and memory already installed in my rack?
The accelerator communicates over standard PCIe, so it functions with existing CPUs and system memory as long as the motherboard provides a compatible x16 link and the platform BIOS supports add-in compute devices.
New to this? Buying server hardware with crypto — what to check, what it costs to pay, and what is never asked for.
Also in this aisle
People compared these
Same shelf, same checkout — eight coins and a 30-minute rate lock.
NVIDIA
NVIDIA 900-2G183-0000-001 Tesla T4 16 GB GDDR6 ai accelerator card
$1,305.99
0.01566198BTC
In stock