Alibaba’s PPU Chip Outperforms NVIDIA A800 – How China’s AI Hardware Is Catching Up

A recent report comparing Alibaba’s home‑grown PPU AI accelerator with NVIDIA’s A800 shows the PPU’s 96 GB HBM2e memory, 700 GB/s bandwidth, PCIe 5.0×15 interface and 400 W power draw surpass the A800 and rival the H20, while large‑scale deployments in China’s telecom sector demonstrate rapidly growing compute capacity.

21CTO
21CTO
21CTO
Alibaba’s PPU Chip Outperforms NVIDIA A800 – How China’s AI Hardware Is Catching Up

On September 17, a comprehensive report citing CCTV and other news sources compared the key parameters of domestic AI accelerator cards with NVIDIA’s offerings. Alibaba’s subsidiary Pingtouge (平头哥) unveiled its latest artificial‑intelligence PPU chip, which surpasses NVIDIA’s A800 on all major metrics and is comparable to the H20.

Compared with other domestic AI chips, the Pingtouge PPU leads in most performance indicators.

Specific parameters include: the PPU is equipped with 96 GB HBM2e memory, exceeding the A800’s 80 GB HBM2e and matching the H20’s capacity (though the H20 uses the newer HBM3). Its inter‑chip bandwidth reaches 700 GB/s, higher than the A800’s 400 GB/s but slightly below the H20. The PPU supports PCIe 5.0×15, outperforming the A800’s PCIe 4.0×16 and equal to the H20. Power consumption stays at 400 W, matching the A800 and lower than the H20’s 550 W.

The report also noted that China Unicom’s Sanjiangyuan Green Power Intelligent Computing Center project has signed contracts with multiple domestic AI‑chip vendors. The signed equipment totals 1,747 units, comprising 22,832 compute cards and delivering a combined compute power of 3,479 P. Breakdown: Alibaba Cloud’s “Wan Card” – 1,024 units with 16,384 Pingtouge compute cards delivering 1,945 P; Chinese Academy of Sciences – 512 units with 4,096 MuMing cards delivering 984 P; Beijing Jingtai – 83 units with 1,328 BR100 cards delivering 450 P; Zhonghao Semiconductor – 128 units delivering 200 P.

In addition, the planned contracts for the project anticipate a total compute power of 2,002 P, involving cards from companies such as Tai​chu Yuanjie, SuiZhi Technology, and Moore Thread.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Alibabaperformance comparisonAI ChipPPUcompute powerHBM2ePCIe 5.0
21CTO
Written by

21CTO

21CTO (21CTO.com) offers developers community, training, and services, making it your go‑to learning and service platform.

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.