Huawei Ascend 950PR vs Huawei Ascend 960

The Ascend 960 delivers 2.0x the FP8 throughput of the Ascend 950PR (2,000 vs 1,000 TFLOPS dense). The Ascend 960 also carries 160 GB more memory (288 GB vs 128 GB).

Ascend 950PR: Da Vinci v3, 2026 Ascend 960: Da Vinci v4 (chiplet), 2027

FP8 dense lead

2.0x

Ascend 960: 2,000 vs 1,000 TFLOPS

Memory

128 vs 288 GB

160 GB more for the Ascend 960

Bandwidth

1.6 TB/s vs 9.6 TB/s

Ascend 960 moves data faster

Download comparison sheet Printable version →

Huawei Ascend 950PR

Full specs →

What is the Huawei Ascend 950PR?

The Ascend 950PR is an NPU from Huawei on the Da Vinci v3 architecture, an accelerator built for AI inference. It is an OAM module, the open accelerator format used in multi-GPU servers. Ascend 950PR full specifications →

Can I rent the Ascend 950PR in the cloud?

Flopper is not tracking any live rental listings for the Ascend 950PR yet. The provider directory lists the clouds that rent comparable accelerators on demand. GPU cloud providers →

How does the Ascend 950PR compare with other accelerators?

This page sets the Ascend 950PR against the Ascend 960. The Ascend 950PR is the older generation of the two. Both are Huawei parts, and both ship as an OAM module, so they fit the same class of system. The compare tool builds the same comparison for any two parts in the catalogue. All GPU comparisons →

Where do the Ascend 950PR figures come from?

Huawei published the Ascend 950PR figures in marketing material rather than a datasheet, and Flopper has no independent confirmation of them. How Flopper verifies specifications →

Huawei Ascend 960

Leads on FP8 Full specs →

What is the Huawei Ascend 960?

The Ascend 960 is an NPU from Huawei on the Da Vinci v4 (chiplet) architecture, an accelerator built for AI inference. It is an OAM module, the open accelerator format used in multi-GPU servers. It has been announced but is not yet shipping. Ascend 960 full specifications →

Can I rent the Ascend 960 in the cloud?

Flopper is not tracking any live rental listings for the Ascend 960 yet. The provider directory lists the clouds that rent comparable accelerators on demand. GPU cloud providers →

How does the Ascend 960 compare with other accelerators?

This page sets the Ascend 960 against the Ascend 950PR. The Ascend 960 is the newer generation of the two. Both are Huawei parts, and both ship as an OAM module, so they fit the same class of system. The compare tool builds the same comparison for any two parts in the catalogue. All GPU comparisons →

Where do the Ascend 960 figures come from?

Huawei published the Ascend 960 figures in marketing material rather than a datasheet, and Flopper has no independent confirmation of them. How Flopper verifies specifications →

Specifications

Ascend 950PR

Ascend 960

Architecture

Da Vinci v3Da Vinci v4 (chiplet)

Launch Year

20262027

Form Factor

OAMOAM

Memory

128 GB288 GB

Memory Bandwidth

1.6 TB/s9.6 TB/s

TDP

900 W—

Process Node

——

Spec

Ascend 950PR

Ascend 960

Architecture Da Vinci v3Da Vinci v4 (chiplet)

Launch Year 20262027

Form Factor OAMOAM

Memory 128 GB288 GB

Memory Bandwidth 1.6 TB/s9.6 TB/s

TDP 900 W—

Process Node ——

Performance (TFLOPS)

Ascend 950PR

Ascend 960

500 TFLOPS 1,000 TFLOPS

1,000 TFLOPS 2,000 TFLOPS

FP6

2,000 TFLOPS 4,000 TFLOPS

Precision

Ascend 950PR

Ascend 960

500 TFLOPS 1,000 TFLOPS

1,000 TFLOPS 2,000 TFLOPS

FP6

2,000 TFLOPS 4,000 TFLOPS

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 Ascend 960 is the newer part: Da Vinci v4 (chiplet) (2027) against the Ascend 950PR's Da Vinci v3 (2026).

The Ascend 960 carries 160 GB more memory (288 GB against 128 GB), so larger models fit on one GPU before you shard. Bandwidth is 1.6 TB/s against 9.6 TB/s, which sets the ceiling on decode speed.

Weight the FP8 and FP16 rows for LLM work and the FP64 row for HPC.

New to the terms? The glossary explains BF16, FP16, FP8, FP4, Sparsity and TFLOPS.

Frequently asked questions

What is better, the Ascend 950PR or the Ascend 960?

The Ascend 960, on throughput. At FP8 precision the Ascend 960 reaches 2,000 TFLOPS dense against 1,000 TFLOPS for the Ascend 950PR. The performance table on this page lists every published precision for both GPUs.

Which should you choose, the Ascend 950PR or the Ascend 960?

Choose the Ascend 960 for training and high-throughput inference: it has 2.0x the FP8 throughput of the Ascend 950PR. It also carries 160 GB more memory, so larger models fit on one GPU before you shard.

Which has more memory, the Ascend 950PR or the Ascend 960?

The Ascend 960 carries 288 GB of memory versus 128 GB for the Ascend 950PR. Memory bandwidth is 1.6 TB/s for the Ascend 950PR and 9.6 TB/s for the Ascend 960.

Get Comparison Updates

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

Open in Full Compare Tool → Ascend 950PR full specs Ascend 960 full specs Estimate cluster cost in the calculator →

Huawei Ascend 950PR vs Huawei Ascend 960

Huawei Ascend 950PR

What is the Huawei Ascend 950PR?

Can I rent the Ascend 950PR in the cloud?

How does the Ascend 950PR compare with other accelerators?

Where do the Ascend 950PR figures come from?

Huawei Ascend 960

What is the Huawei Ascend 960?

Can I rent the Ascend 960 in the cloud?

How does the Ascend 960 compare with other accelerators?

Where do the Ascend 960 figures come from?

Specifications

Performance (TFLOPS)

FLOPS by Precision

What actually differs

Frequently asked questions

What is better, the Ascend 950PR or the Ascend 960?

Which should you choose, the Ascend 950PR or the Ascend 960?

Which has more memory, the Ascend 950PR or the Ascend 960?

Get Comparison Updates

Stay Updated on GPU Specs