Qualcomm is aiming to take on standard nvidia and Intel with its Cloud AI 100 reasoning chip
Published time: 2019-04-12
When it comes to qualcomm, the industry has a direct impression of the mobile chip giant.Indeed, qualcomm has been heavily involved in mobile communications, from chips to the underlying platforms.But now, qualcomm unexpectedly throws out a "bomb" will also change the previous everybody to its cognition.
At the Qualcomm AI Day event in San Francisco, the giant company announced its entry into the Cloud computing market and unveiled a dedicated AI accelerator for AI reasoning: the Qualcomm Cloud AI 100.Qualcomm said the Cloud AI 100 series accelerator, based on a 7-nanometer chip process, will be available in 2020, with a prototype to be announced later this year.Without warning, qualcomm is the fourth company after Google, amazon and nvidia to successfully launch chips on cloud reasoning.
According to the definition of qualcomm, Cloud AI 100 is a special "artificial intelligence reasoning" AI Accelerator (purpose - built AI Accelerator), it integrates a variety of development tools include compilers, analyzer, monitor, and service, chip debugger and quantitative, so that customers can be adjusted according to the AI reasoning processing tasks demand and module design, appearance and power levels.
Cloud AI 100 has peak performance 3 to 50 times that of Snapdragon 855 and Snapdragon 820.Compared with the traditional FPGA, its reasoning speed is about 10 times higher.On the software side, it supports ONNX, Glow and XLA, as well as most of the deep learning frameworks including Google's TensorFlow, Facebook's PyTorch, Keras, MXNet, baidu's PaddlePaddle and Microsoft's cognitive toolkit.
As for the specific technical details, qualcomm did not disclose too much, and it is not suitable to make a simple horizontal comparison with the competing products at present.However, for the convenience of understanding, the data of Google TPU generation can be released for reference: the computing performance of Cloud AI 100 can reach 350TOPS;The test result of Google TPU generation under 250W is 92TOPS.
Undoubtedly, with the deepening of cloud computing and artificial intelligence technology for various scenarios, the demand for inference computing will continue to rise. In front of this blue ocean, qualcomm is undoubtedly to grab a part of the market, and intends to face the "determined" nvidia and Intel.
- 1.Qualcomm is aiming to take on standard nvidia and Intel with its Cloud AI 100 reasoning chip
- 2.Qualcomm AI 100 AI reasoning chip was born, claiming to disrupt the cloud
- 3.Qualcomm will launch the new Snapdragon flagship processor chip
- 4.Qualcomm High-performance chip_ Snapdragon 710 mobile platform released
- 5.Qualcomm Cellular Solution is suitable for PC5-based direct communications.
- 6.Processors bring premium-tier features to smartphones.
- 7.Reference Designs and SoC support wearable devices.
- 8.Quad- and Octa-Core Chipsets integrate 5-mode global LTE.
- 9.Processor Chipsets feature integrated 4G LTE World Mode.
- 10.Chipset supports carrier-grade multi-room HD streaming.
Field Programmable Gate Array (FPGA) >
- MSM6300 CP70-V4404-4