Alibaba revealed its first chip to power the processes of artificial intelligence (AI) on Tuesday. The action can enhance the already fast-growing cloud computing business and signals China’s developing ambitions in creating a personal homegrown semiconductor industry. The chip revealed is called the Hanguang 800, which can reduce the computing tasks that normally would take an hour to five minutes.
Alibaba stated that the chip is being used within the company’s business operations presently. Most especially in the aspect of intelligent customer services, the product search and automatic translation on e-commerce sites, advertisement, and customized recommendation. Mostly, these are the aspects that need extensive computing tasks in which the chip can help speed them up.
Growing Computing Business
Using the chip in the areas of cloud computing business would enable Alibaba to sell its new cloud services to customers. There is no doubt that one of Alibaba’s fastest-growing divisions is the cloud business and it has been highlighted by the management as an important area for the future.
The giant of Chinese e-commerce is among the top five in the world, and it is also the largest and major cloud driver by market share in the mainland.
Alibaba’s action on AI chips made Huawei revealed its own AI semiconductor called the Ascend910 last month. It is targeted at being utilized mainly in data centres. Most companies making use of AI applications need a large amount of data to develop smart algorithms, which takes almost a week or even more than. Huawei stated that its chip is the “world’s most powerful AI processor.”
The release of the AI chips by the two of China’s biggest tech companies, Alibaba and Huawei emphasized the country’s aim to develop its homegrown semiconductor industry. Most of the Chinese companies depend majorly on American technology presently. But, when Washington banned Huawei and restricts its access to American technology, it has helped enhanced China’s attention on chips.
Beijing even emphasized semiconductors as the major aspect of the “Made in China 2025” plan, a government initiative that intends to enhance the production and release of high-quality products. One of China’s aspirations is to produce about 40% of the semiconductors it uses by 2020 and 70% by 2025.
Potential of the Chip
Normally, it takes Taobao about an hour to divide the one billion product images that are being uploaded to the e-commerce platform by business owners and properly assemble them for search and customized recommendations. With Hanguang 800, Taobao was able to complete the task in five minutes.
Alibaba is making use of Hanguang 800 in almost all of its business operations that require machine processing.
The president of Alibaba Cloud Intelligence stated that “The launch of Hanguang 800 is an important step in our pursuit of next-generation technologies, boosting computing capabilities that will drive both our current and emerging businesses while improving energy-efficiency.” He also stated that “Soon, we plan to empower our clients by providing access through our cloud business to the advanced computing that is made possible by the chip, anytime and anywhere.”
Plans of Alibaba with the Chip
Jeff Zhang, Alibaba chief technology officer stated that the Hanguang 800, has a computing power that is similar to that of ten graphics processing units (GPUs). Which is a more general-purpose type of computing chip used to process AI workloads. The launching of this chip is already contributing positively to Alibaba’s e-commerce business.
Alibaba did not divulge whether it would substitute the uses of GPUs with the AI chip on a largescale. Nevertheless, the company’s billionaire co-founder Jack Ma has pointed out his desire to decrease the company’s dependence on U.S. chips. Almost each of the processors used to compute complicated AI workloads and analyze videos, images, and languages in data centre servers are made by U.S. chipmakers. Such as Nvidia, Advanced Micro Devices (AMD), Intel’sAltera, as well as Xilinx.
Alibaba stated that it has no plans to sell the chips. But analysts say furnishing its cloud service with the high-quality and highly-efficient AI chips would increase the aim of alluring more customers to China’s increasingly competitive cloud market.
The customized AI chips were developed to perform specific tasks with lightning speed and lower power consumption.
The Chinese e-commerce corporation launched its first cloud AI accelerator chip on stage during its Aspara Conference held in Hangzhou, China. Which is the Hanguang 800Presently, Alibaba has no plans of selling the hardware, although the new AI chip is expected to reach the business cloud servers at some point for customers to rent.
The specs of Hanguang 800’s include, it is a 12nm fabrication technology and has about 17billion transistors, based on Synced. Alibaba gave a little detail about its performance by declaring a few benchmark figures, using the popular convolutional neural network ResNet-50.
A Hanguang 800 chip running the ResNet-50 can process about 78,563 images per second (IPS)and operates at peak efficiency is 500 IPS per watt. It was stated that Alibaba’s silicon was 15times and 46 times more powerful than Nvidia’s T4 GPU and its older P4 model respectively.
Ever since Jack Ma declared that China had to stop depending on US imports of AI Chips, Alibaba has been stepping up its hardware efforts. Currently, the two countries are still locked ina trade tariff war, setting higher prices for technical equipment. Jack Ma has created a new semiconductor division, T-Head or Pingtouge – meaning honeybadger in Chinese, to make the company more self-sufficient and self-dependent.
Hanguang 800 is the first AI chip created by T-Head engineers in partnership with researchers from the DAMO Academy, Alibaba’s research arm.
The Middle Kingdom movers and shakers have also released blueprints for its RISC-V based chip, the Xuantie 910 or XT 910 for short in July. The processor was publicized as the fastest and most powerful RISC-V chip presently, having a 12nm 64-bit (RV64GCV) RISC-V design that packs 16 cores clocked at up to 2.5GHz. The XT 910 includes 50 extra instructions to help achieve a range of tasks, including memory and storage access, on top of the open-source base RISC-V instructions.