Qualcomm moves to the edge with Cloud AI 100 chip

Mobile chip-maker Qualcomm reckons all the stuff it has learned about processing AI in smartphones will come in handy in datacentres too.

The Qualcomm Cloud AI 100 Accelerator is a special chip designed to process artificial intelligence in the cloud. Specifically Qualcomm seems to think it has an advantage when it comes to ‘AI inference’ processing – i.e. using algorithms that have been trained with loads of data. This stands to reason as it has its chips in millions of smart devices, all of which will have been asked to do some inference processing of their own from time to time.

“Today, Qualcomm Snapdragon mobile platforms bring leading AI acceleration to over a billion client devices,” said Qualcomm Product Management SVP Keith Kressin. “Our all new Qualcomm Cloud AI 100 accelerator will significantly raise the bar for the AI inference processing relative to any combination of CPUs, GPUs, and/or FPGAs used in today’s datacentres. Furthermore, Qualcomm Technologies is now well positioned to support complete cloud-to-edge AI solutions all connected with high-speed and low-latency 5G connectivity.”

The datacentre chips in question will largely be provided by Intel, although Nvidia has done a great job of converting its struggling mobile chip efforts into a successful AI processing operation. Qualcomm claims a 10x performance per watt advantage over incumbent AI inference chips and, while it didn’t call out any competitors in its press release, the predominance of their names in the headlines of other stories covering this launch makes it likely that has been the angle behind the scenes.

Three UK shows off its new Nokia cloud core

Mobile operator Three UK has upgraded its network with a fully cloud-based 5G-ready core and has started internal trials of the service. It plans to launch 5G later this year.

Three announced that it is testing the world’s first fully cloud-based core network, delivered by Nokia. The software-based core network is 5G ready and is already carrying the ongoing trial for Three’s own staff. The trial is on the 3.4-3.8GHz spectrum Three bought with over £164 million in the auction concluded in April 2018.

The readiness is also achieved on the edge. Three announced that by December 2018, all its mast sites were already connected to the new cloud-based core networks, meaning when 5G is switched on all Three customers would be able to access 5G services, provided they have the 5G-enabled user devices (fixed wireless access modems, or smartphones and tablets).

Another infrastructure update Three announced is the expansion of its datacentre network. The operator used to have three datacentres in London and the Midlands. After the latest upgrade, it now has “21 data centres spread from as far North as Edinburgh to Portsmouth in the South” which are all live and “have been connected up with fibre”, said the statement. In practical terms, the more distributed datacentre network would reduce latency experienced by the users faraway from southern England, giving customers more or less equal user experience.

Indeed, “enhancing its market-leading customer experience and becoming the best loved brand in the UK by its people and customers” is the explicit target of Three’s latest network upgrading. The company reiterated its target to launch commercial 5G service later this year, after committing to invest over £2bn into 5G. “We have been planning our approach to 5G for many years and we are well positioned to lead on this next generation of technology.  These investments are the latest in a series of important building blocks to deliver the best end to end data experience for our customers,” Dave Dyson, Three UK’s CEO, said late last year.

According to the latest telecoms complaints numbers released by Ofcom in January, Three received 4 complaints per 100,000 customers, narrowly behind its mobile competitors EE and O2 (3 complaints each) but way ahead of Vodafone (8).

Huawei joins forces with Nvidia for server fun

Huawei has started the Friday party early by unveiling the FusionServer G series heterogeneous computing platform. Break out the party hats.

Launched at Huawei Connect 2017, the FusionServer G5500 and G2500 products build out Atlas, Huawei’s intelligent cloud hardware platform, as well as its Boundless Computing strategy. We can basically hear you drooling through the computer screen.

“Today’s enterprise service applications are rapidly evolving, and the types of workloads are also diversifying. These pose tremendous challenges on the efficiency and flexibility of computing platforms. With the G Series heterogeneous computing platform, we can help our customers better meet these challenges,” said Qiu Long, President, IT Server Product Line, Huawei.

“Huawei is happy to join forces with Nvidia for comprehensive, deep collaboration in the AI computing front. We believe that both parties’ innovation power will translate into powerful GPU Accelerated Datacentre platforms to help our customers travel more smoothly and swiftly in the digital transformation journey.”

Huawei claim the G series is engineered with heterogeneous resource pooling capabilities, allowing resources to be intelligently orchestrated based on application workloads. Huawei believe this will enable users to derive higher computing efficiency.

Looking more specifically at the individual products, FusionServer G5500 is a heterogeneous server with a focus on data center deployment, while FusionServer G2500 is a smart video analytics server positioned for application scenarios such as safe city and smart transportation.

We couldn’t imagine a better start to the weekend.