Google says Intel-powered VMs are the way forward, despite cost
>
Google urges some of its enterprise cloud customers to shift their “most demanding workloads” to expensive Intel virtual machines.
In a after on the Google Cloud blog, the company announced the launch of the Intel Software Center of Excellence after a successful pilot project in 2021.
Companies with use cases such as video encoding, web transactions, databasesanalytics and AI inference are invited to apply for the free scheme, with Google and Intel working with accepted applicants to access and optimize their workflow performance.
Personalized cloud collaboration
The ability for Intel to personally assess a company’s internal resources to provide a personalized report on how it can implement specific optimizations in a workflow running on Google Cloud can seem appealing.
While the scheme is free, Intel VMs will cost businesses more due to operating costs Xeon scalable (opens in new tab) processors, which provide built-in AI acceleration via the AVX-512 set of vector instructions – first introduced in 2013 – and the more recent Deep learning boost (opens in new tab) function that uses the technology.
While it’s true that VMs equipped with Xeon Scalable processors can provide an impressive speed boost to some workflows, the software that powers them needs to be optimized by developers for use with the technology.
High costs and limited use cases may be the reasons why Google and Intel only offer a practical “white glove” service and one that is only available to certain “high-growth enterprises” depending on the application.
Businesses that don’t need the latest Intel processor technology, can’t take advantage of it, or simply aren’t aware of it, may be put off by an automatic switch to service.
While Intel has long been the only option for businesses looking to take advantage of AI acceleration, AMD’s latest lineup of Epyc server processors will ship with AVX-512 support this fall.