Huawei Atlas 900 A3 SuperPoD
Huawei's 384-NPU supernode, built from twelve 47U liquid-cooled compute cabinets plus four air-cooled bus-equipment cabinets, each 2,250 by 600 by 1,150 mm. It carries up to 384 Ascend 910 NPUs and 192 Kunpeng 920 CPUs, with 384 x 128 GB of on-chip memory at up to 3.2 TB/s each and 784 GB/s of bidirectional device-to-device bandwidth. Huawei rates it at up to 307.2 PFLOPS of FP16, and also lists 288.7 and 240.3 PFLOPS variants; the top figure works out at exactly 800 TFLOPS per NPU, which is the Ascend 910C's own published rate, so the BF16 and INT8 figures here are that chip's published 800 and 1,600 TFLOPS multiplied by 384. Cooling is liquid for the compute cabinets and air for the bus cabinets, with an operating range of 5 to 40 C. Note that Huawei publishes no power figure for this machine: its supply row states only the input voltage, three-phase 380V AC on two feeds. A figure of 559 kW previously held here appeared in no Huawei source and has been removed rather than shipped. Huawei labels nothing dense or sparse. This machine is also sold as CloudMatrix 384, which is a deployment name for the same 384-NPU configuration rather than a separate product.
We will point you at suppliers who have it. Free, and no signup.
System Details
GPU Configuration
System Specifications
Precision Performance Breakdown
| Precision | System Performance | Per GPU | Efficiency |
|---|---|---|---|
| 307 PFLOPS | 800 TFLOPS | — | |
| 307 PFLOPS | 800 TFLOPS | — | |
| 614 POPS | 1,600 TOPS | — |
Where a vendor states a basis, the figure shown is the dense one, and any figure it publishes is listed separately. Vendors commonly headline the sparse number instead: for NVIDIA's 2:4 structured sparsity that is exactly twice the dense figure, though other vendors' sparse modes do not all follow that ratio.
Powered by Huawei Ascend 910C
This system utilizes 384 × Huawei Ascend 910C GPUs, each delivering exceptional performance for AI and HPC workloads.
Per GPU TDP
600W
Per GPU Memory
128 GB
Process Node
SMIC 7nm
Architecture
Da Vinci v2 (dual-die)
Frequently asked questions
How much does the Atlas 900 A3 SuperPoD cost?
Huawei sells the Atlas 900 A3 SuperPoD on quote through OEMs and resellers, and Flopper has no live rental listing for its Ascend 910C accelerator to derive an hourly figure from. System rental pricing →
How much power does the Atlas 900 A3 SuperPoD require?
Huawei has not published a power figure for the Atlas 900 A3 SuperPoD, and Flopper does not derive one from the accelerator count because the rest of the system draws power too.
How many GPUs are in the Atlas 900 A3 SuperPoD?
384. The Atlas 900 A3 SuperPoD carries 384 Huawei Ascend 910C accelerators in a pod form factor, and every throughput figure on this page is the total across them. Ascend 910C specifications →
Documentation & Resources
Need this system?
Tell us the quantity, timeframe and region. We reply by email with who can supply it.
Flopper Spec Sheet
Huawei Atlas 900 A3 SuperPoD specifications, generated from our database. Printable.
Vendor product page
Huawei Atlas 900 A3 SuperPoD documentation
Huawei HiSilicon Da Vinci architecture, Hot Chips 31 (2019)
Huawei • 2019-08-19
Typical Use Cases
The Huawei Atlas 900 A3 SuperPoD runs 384× Ascend 910C GPUs over UnifiedBus (UB) all-to-all optical; 784 GB/s bidirectional D2D per NPU, delivering 307 PFLOPS BF16 dense.