T-Head publishes no throughput figure of any kind for this chip, so the precision table is deliberately empty rather than incomplete. Its product page names no datatype at all, the closest it comes being "实现全精度计算加速", which names none, and T-Head positions the part only by relative claims: level with the NVIDIA H20, and one third the performance of its own Zhenwu M890. A relative claim against a part that itself has no published rate cannot be resolved into a figure. The 2.7 TB/s on this row is memory bandwidth, from a row of the vendor's specification table headed 内存带宽; the 700 GB/s is inter-chip interconnect bandwidth across seven ICN ports and is a different quantity. This part is not sold as a component: it exists only as Alibaba Cloud capacity, which is why there is no form factor, no board power and no list price.
T-Head logo

T-Head Zhenwu 810E NEW

Type: NPUArchitecture: ZhenwuReleased: 2026Spec confidence: Official
Memory
96 GB
HBM2e
Bandwidth
2.7 TB/s
memory

Overview

The Zhenwu 810E is the first Alibaba AI accelerator the company put a name and a product page to, on 29 January 2026, although it had been shipping inside Alibaba Cloud for a year before that under the internal label PPU. It pairs 96 GB of HBM2e at 2.7 TB/s with seven independent ICN ports carrying 700 GB/s of inter-chip bandwidth, and it is the part that proved the architecture at scale: Alibaba says it runs several clusters of more than ten thousand cards, and that more than 100,000 of them serve autonomous driving customers on Alibaba Cloud alone. What the specification cannot tell you is how fast it is. T-Head publishes no TFLOPS or TOPS figure for this chip, names no datatype on its product page, and positions it only by comparison, claiming overall performance above the NVIDIA A800 and level with the NVIDIA H20. It is not sold as a component. The only way to use one is to rent Alibaba Cloud capacity.

Performance

Peak theoretical throughput by precision type

PrecisionPeak
FP64
No verified data available
FP32
No verified data available
TF32
No verified data available
BF16
No verified data available
FP16
No verified data available
FP8
No verified data available
FP6
No verified data available
FP4
No verified data available
INT8
No verified data available

Flopper holds no verified throughput figures for this part yet. The rows stay so the gaps are visible.

Specifications

Architecture

Zhenwu

Form Factor

No verified data available

Launch Year

2026

Process Node

No verified data available

Memory

96 GB HBM2e

Bandwidth

2,700 GB/s

TDP

No verified data available

Max power

No verified data available

Interconnect

700 GB/s ICN

direction and scope not stated by vendor

Spec Confidence

Official

Full Specifications

Memory
Memory 96 GB
Memory Type HBM2e
Bandwidth 2.7 TB/s
Interface Width No verified data available
Interconnect & I/O
GPU-to-GPU ICN
Interconnect Bandwidth 700 GB/s direction and scope not stated by vendor
Interconnect Links 7
Power & Thermal
TDP No verified data available
Max power No verified data available
General
Form Factor No verified data available
Architecture Zhenwu
Process Node No verified data available
Launch Year 2026

Datasheet & Resources

Flopper Datasheet

T-Head Zhenwu 810E specifications, generated from our database. Printable.

Zhenwu 810E product page (T-Head)

T-Head · Latest version

View

Data Provenance

Every figure traced to a source

Primary Source

Publisher
T-Head
Published
No verified data available

Data Quality

Spec confidence
Official
Core precisions with figures
0 of 9
Normalization
All values in TFLOPS

Compare cloud providers offering on-demand GPU instances for AI training, inference, and HPC workloads.

Browse GPU Cloud Providers

Similar GPUs

T-Head Zhenwu M890

· 2026
FP32: No verified data available
Compare vs Zhenwu M890

T-Head Zhenwu V900

· 2026
FP32: No verified data available
Compare vs Zhenwu V900

Frequently Asked Questions

How many TFLOPS does the T-Head Zhenwu 810E have?

Flopper does not currently have any verified throughput figures for the T-Head Zhenwu 810E.

What is the power consumption of the T-Head Zhenwu 810E?

Flopper does not currently have a verified TDP (Thermal Design Power) figure for the T-Head Zhenwu 810E.

How much memory does the T-Head Zhenwu 810E have?

The T-Head Zhenwu 810E is equipped with 96 GB of memory with 2,700 GB/s of memory bandwidth.

What architecture is the T-Head Zhenwu 810E based on?

The T-Head Zhenwu 810E is based on the Zhenwu architecture, launched in 2026.

Stay Updated on GPU Releases

Get notified when new GPUs are added or specifications are updated.

Loading verification...

No spam, unsubscribe anytime.

Back to GPUs

© 2026 Flopper.io - Compare the GPUs Powering AI