Published in Cloud

Alibaba gets Xilinx FPGA for its F3 cloud

by on11 May 2018


Faster deep learning and IO

Back in October 2017 Alibaba, a $502,179 billion company announced that it had chosen Xilinx for next generation FPGA acceleration in their public cloud.   

Many people don’t realize how important and big Alibaba really is in the cloud market and therefore we have to remind you again that this is a $500 billion company. It announced to the press in China that it launched a new cloud computing instance called F3, powered with a self-developed HP Acceleration Card and a call for public testing. 

Alibaba F3 series is using “Complete elf-developed High-Performance Acceleration Card” based on the integrated Xilinx 16nm VirtexUltraScale+ Device VU9P, which can provide up to 16 VU9Ps instance specifications

Let me remind you that Amazon is using Xilinx F1 instances for its Amazon Web Services too, but Alibaba is now testing the next generation FPGA in 16nm supporting HBM 2.0 memory.

Fast AI, Deep learning and IO for cloud 

The Alibaba instance provides more than 10 million logic cells and up to 47 TeraMACs DSP computing ability. The 16nm VirtexUltraScale+ FPGA with HBM 2 started shipping for customer testing and Alibaba said last year that Xilinx FPGAs, "F2" instances give Alibaba Cloud customers access to acceleration for data analytics, genomics, video processing, and machine learning workloads. Even with the previous generation hardware, the gain can be up to 30 times higher than CPUs alone and with the new F3 hardware you can expect a significant uplift in performance.

Currently the test is targeting China based customers, but it will likely spread to other territories. 

This is an important win for Xilinx and Alibaba will probably be more vocal about this in the near future, as it is clear that Alibaba wants to challenge the big players in the cloud, especially Amazon web services.

Last modified on 11 May 2018
Rate this item
(0 votes)

Read more about: