Google Cloud is super keen to keep certain customers on pricey Intel VMs

Unless your app really needs AVX 512, AMD and Ampere's cores may be a cheaper, better bet

Google Cloud really, really wants enterprises to keep using its Intel-powered virtual machines. This week it unveiled a “white-glove” service to convince customers the silicon is worth paying extra for.

It's no secret Google Cloud's Intel-based instances come at a premium over those using AMD Epyc or Ampere’s Altra CPU cores. A fairly modest, eight-thread, 32GB N2-Standard instance based on Intel’s Xeon Scalable platform, for example, will run you about 25 percent more than an Ampere Altra-based instance with twice the physical cores. Meanwhile, AMD's Epyc instances slot in somewhere in the middle.

Despite the higher cost of Intel-based instances, Google says that, depending on your workload, opting for a cheaper option may mean leaving performance on the table, and you may therefore get more oomph for your dollar from the Intel-powered instances. Google has also spent a bunch of money on these Xeons and by golly, it wants you to use all of them.

Intel has long touted the AI and cryptographic accelerators built into its Xeon Scalable processors. This really boils down to its AVX 512 vector extensions, Deep Learning Boost, and cryptographic engines. These can be used to accelerate a variety of workloads ranging from transcoding, database, and inferencing, without resorting to GPU instances. But that’s only if the customer’s software is optimized to take advantage of these features.

And that’s a big enough if that Google and Intel are willing to work one-on-one with select “high-growth enterprise” customers to refactor deployments and even tweak their application code base in pursuit of higher performance.

At launch, Google is explicitly targeting customers running database, analytics, AI inference, Nginx, language runtime, and/or media transcoding workloads.

Under the three-phase program, Google and Intel engineers will review a customer’s performance objectives and identify opportunities for improvement. Based on this information, the web giant and the x86 titan's teams will take resource utilization into account and provide a performance report detailing the recommended changes to the customer’s environment, we're told.

Equifax, one of the first corporations to participate in the program, claimed the optimizations doubled throughput and halved the latency of its deployments.

While the program is being made available at no cost, customers must apply for the service. This suggests Intel and Google may be gating access to the program as to avoid inadvertently sending customers running for the competition.

It’s worth noting that customers who rely on AVX 512 instructions to accelerate their workloads won’t be stuck on Intel much longer. As we saw with AMD’s Zen 4 launch, the Intel rival will finally roll out support for large vector extensions to its Epyc processors this fall. ®