Huawei Ascend 910A vs Huawei Atlas 300V

The Ascend 910A delivers 5.1x the FP16 throughput of the Atlas 300V (256 vs 50 TFLOPS dense). The Ascend 910A also carries 8 GB more memory (32 GB vs 24 GB).

Ascend 910A: Da Vinci, 2019 Atlas 300V: Da Vinci, 2022

FP16 dense lead

5.1x

Ascend 910A: 256 vs 50 TFLOPS

Memory

32 vs 24 GB

8 GB more for the Ascend 910A

Bandwidth

1.2 TB/s vs 205 GB/s

Ascend 910A moves data faster

TDP

310 W vs 72 W

Atlas 300V draws 238 W less

Download comparison sheet Printable version →

Huawei Ascend 910A

Leads on FP16 Full specs →

What is the Huawei Ascend 910A?

The Ascend 910A is an NPU from Huawei on the Da Vinci architecture, an accelerator built for AI inference. It is an OAM module, the open accelerator format used in multi-GPU servers. Ascend 910A full specifications →

Can I rent the Ascend 910A in the cloud?

Flopper is not tracking any live rental listings for the Ascend 910A yet. The provider directory lists the clouds that rent comparable accelerators on demand. GPU cloud providers →

How does the Ascend 910A compare with other accelerators?

This page sets the Ascend 910A against the Atlas 300V. The Ascend 910A is the older generation of the two. Both are Huawei parts; the Ascend 910A is an OAM module while the Atlas 300V is a PCIe card. The compare tool builds the same comparison for any two parts in the catalogue. All GPU comparisons →

Where do the Ascend 910A figures come from?

Every Ascend 910A figure on this page traces to Huawei's own datasheet or specification table. How Flopper verifies specifications →

Huawei Atlas 300V

Full specs →

What is the Huawei Atlas 300V?

The Atlas 300V is an NPU from Huawei on the Da Vinci architecture, an accelerator built for AI inference. It is a PCIe add-in card, so it fits standard servers and workstations. Atlas 300V full specifications →

Can I rent the Atlas 300V in the cloud?

Flopper is not tracking any live rental listings for the Atlas 300V yet. The provider directory lists the clouds that rent comparable accelerators on demand. GPU cloud providers →

How does the Atlas 300V compare with other accelerators?

This page sets the Atlas 300V against the Ascend 910A. The Atlas 300V is the newer generation of the two. Both are Huawei parts; the Atlas 300V is a PCIe card while the Ascend 910A is an OAM module. The compare tool builds the same comparison for any two parts in the catalogue. All GPU comparisons →

Where do the Atlas 300V figures come from?

Every Atlas 300V figure on this page traces to Huawei's own datasheet or specification table. How Flopper verifies specifications →

Specifications

Ascend 910A

Atlas 300V

Architecture

Da VinciDa Vinci

Launch Year

20192022

Form Factor

OAMPCIe

Memory

32 GB24 GB

Memory Bandwidth

1.2 TB/s205 GB/s

TDP

310 W72 W

Process Node

TSMC 7nm—

Spec

Ascend 910A

Atlas 300V

Architecture Da VinciDa Vinci

Launch Year 20192022

Form Factor OAMPCIe

Memory 32 GB24 GB

Memory Bandwidth 1.2 TB/s205 GB/s

TDP 310 W72 W

Process Node TSMC 7nm—

Performance (TFLOPS)

Ascend 910A

Atlas 300V

16 TFLOPS

256 TFLOPS 50 TFLOPS

FP6

512 TOPS 100 TOPS

Precision

Ascend 910A

Atlas 300V

16 TFLOPS

256 TFLOPS 50 TFLOPS

FP6

512 TOPS 100 TOPS

Figures are dense. Where a vendor publishes a figure it sits under the dense one and never decides a winner. Hover any precision for its definition, or read the GPU glossary.

FLOPS by Precision

What actually differs

The Atlas 300V is the newer part: Da Vinci (2022) against the Ascend 910A's Da Vinci (2019).

The Ascend 910A carries 8 GB more memory (32 GB against 24 GB), so larger models fit on one GPU before you shard. Bandwidth is 1.2 TB/s against 205 GB/s, which sets the ceiling on decode speed.

Power is 310 W against 72 W, and at FP32 the Ascend 910A does more work per watt. Weight the FP8 and FP16 rows for LLM work and the FP64 row for HPC.

New to the terms? The glossary explains FP32, FP16, INT8, Sparsity and TFLOPS.

Frequently asked questions

Is the Ascend 910A faster than the Atlas 300V?

At FP16 precision the Ascend 910A reaches 256 TFLOPS dense against 50 TFLOPS for the Atlas 300V. The performance table on this page lists every published precision for both GPUs.

Which should you choose, the Ascend 910A or the Atlas 300V?

Choose the Ascend 910A for training and high-throughput inference: it has 5.1x the FP16 throughput of the Atlas 300V. It also carries 8 GB more memory, so larger models fit on one GPU before you shard. The Ascend 910A draws 238 W more power, so budget for cooling and power delivery before you commit. Packaging differs too: the Ascend 910A is an OAM module while the Atlas 300V is a PCIe card, so check what your servers accept.

Which has more memory, the Ascend 910A or the Atlas 300V?

The Ascend 910A carries 32 GB of memory versus 24 GB for the Atlas 300V. Memory bandwidth is 1.2 TB/s for the Ascend 910A and 205 GB/s for the Atlas 300V.

How much power do the Ascend 910A and the Atlas 300V draw?

The Ascend 910A is rated at 310 W TDP and the Atlas 300V at 72 W. On FP32 throughput per watt, the Ascend 910A is the more efficient part.

Get Comparison Updates

New GPUs added weekly. Be the first to see how they compare.

Open in Full Compare Tool → Ascend 910A full specs Atlas 300V full specs Estimate cluster cost in the calculator →

Huawei Ascend 910A vs Huawei Atlas 300V

Huawei Ascend 910A

What is the Huawei Ascend 910A?

Can I rent the Ascend 910A in the cloud?

How does the Ascend 910A compare with other accelerators?

Where do the Ascend 910A figures come from?

Huawei Atlas 300V

What is the Huawei Atlas 300V?

Can I rent the Atlas 300V in the cloud?

How does the Atlas 300V compare with other accelerators?

Where do the Atlas 300V figures come from?

Specifications

Performance (TFLOPS)

FLOPS by Precision

What actually differs

Frequently asked questions

Is the Ascend 910A faster than the Atlas 300V?

Which should you choose, the Ascend 910A or the Atlas 300V?

Which has more memory, the Ascend 910A or the Atlas 300V?

How much power do the Ascend 910A and the Atlas 300V draw?

Get Comparison Updates

Stay Updated on GPU Specs