Industry News

Google Launches New AI Chip TPU v5e: Training Performance Boosted by 2x, Reasoning Performance Boosted by 2.5x

Update time : 2023-09-01 10:36:47

Google has announced a series of new artificial intelligence technologies and partnerships aimed at introducing more constantly evolving technologies to large enterprises. Google has also released a new version of customized artificial intelligence chips.

Google stated that Cloud TPU v5e has now launched a preview version and is the latest version of its internal Tensor processing unit. Google stated that compared to the TPU v4 released in 2021, for large language models and generative AI models, the chip's training performance per dollar has been improved by up to 2 times, and the inference performance per dollar has been improved by up to 2.5 times. The cost of TPU v5e is less than half that of TPU v4, enabling more organizations to train and deploy larger and more complex AI models.
Google stated that "the TPU v5e Pod balances performance, flexibility, and efficiency, allowing up to 256 chips to interconnect, with a total bandwidth exceeding 400Tb/s and INT8 performance reaching 100 petaOps
Google Blog stated that the new TPU will provide eight different virtual machine configurations, ranging from one TPU chip to over 250 TPU chips in a single chip. For customers who need more computing, the company is launching "Multislice", a service that hands over models to tens of thousands of TPU chips for computing.
So far, training assignments using TPU are limited to a single TPU chip slice, and the maximum job slice size for TPU v4 is 3072 chips
With Multislice, developers can scale workloads across multiple Pods to tens of thousands of chips within a single Pod through inter chip interconnection (ICI) or through data center networks (DCN)
In addition to the new TPU, Google also stated that the A3 Virtual Machine (VM) will be fully launched in September, equipped with 8 Nvidia H100 GPUs, dual fourth generation Intel Xeon scalable processors, and 2TB of memory. These instances were initially announced in May this year and can be increased to 26000 Nvidia H100 Hopper GPUs. However, given the continuous shortage of GPUs, it is currently unclear how many H100 units Google will have.
Google stated that Anthropic, a generative artificial intelligence startup, was an early user of the new TPU v5e and A3 VM. Although Google has invested $300 million in this startup, it is also a loyal user of Amazon AWS

Previous : US restricts NVIDIA and AMD's AI chip exports to Middle Eastern countries

Next : Russian Chipmaker Baikal Electronics to be Auctioned Off, Worth $5 Million

How many chips does a car need?

Sep .19.2024

Automotive chips can be divided into four types according to their functions: control (MCU and AI chips), power, sensors, and others (such as memory). The market is monopolized by international giants. The automotive chips people often talk about refer to

Position and Function of Main Automotive Sensors

Sep .18.2024

The function of the air flow sensor is to convert the amount of air inhaled into the engine into an electrical signal and provide it to the electronic control unit (ECU). It is the main basis for determining the basic fuel injection volume. Vane type: The

Chip: The increasingly intelligent electronic brain

Sep .14.2024

In this era of rapid technological development, we often marvel at how mobile phones can run various application software smoothly, how online classes can be free of lag and achieve zero latency, and how the functions of electronic devices are becoming mo

LDA100 Optocoupler: Outstanding Performance, Wide Applications

Sep .13.2024

In terms of characteristics, LDA100 is outstanding. It offers AC and DC input versions for optional selection, enabling it to work stably in different power supply environments. The small 6-pin DIP package not only saves space but also facilitates install

About Us

About Us What we do Quality People

Products

CHIP/IC Active devices Passive device Components OTHER

China electronic components supplier----Wah Luen Shing Electronic.----www.wlschip.com

Leave a message