Cloud server provider Linode has launched new graphics processing unit-optimized cloud computing instances for users requiring massive parallel computational power.
The cloud computing instances are built on Nvidia Quadro RTX 6000 graphics processing units (GPU) with Compute Unified Device Architecture, Tensor and Real-Time cores available, supporting parallel-processing, machine and deep learning, and ray tracing, respectively.
Linode GPU instances are tailored for enterprise-grade users working with compute, graphics and data-intensive use cases, such as AI, graphics visualization, video processing and complex modeling in industries including science, art, engineering, media and healthcare.
Linode's GPU cloud instances feature the following capabilities:
- On-demand GPU enables users to turn capital expenditures into operational expenditures.
- Removing barriers to complex use cases enables users to focus on core competencies.
- High-performance Nvidia GPUs support intensive use cases.
- Flexible pricing and plan types accommodate any scope of workload.
The vendor claims this release is a cost-effective alternative to other leading hyperscale providers. Linode GPU plans start at $1,000 per month for one 32 GB instance and up to $4,000 per month for four instances. While Linode's plans are less expensive than many of the high-performance plans with similar memory offered by Google, AWS and Microsoft Azure, the larger vendors offer more variety in plan options -- with as little as 1 GB to as much as 256 GB of memory -- to suit a wider range of projects and workloads.
Determining the value of a plan is difficult to measure and requires consideration of a variety of factors, according to Alan Priestley, vice president analyst at Gartner.
"Value is largely dependent on customer uses and how they utilize GPUs," Priestley said. "Large scale vendors offer plans on a smaller time interval, like per hour, which can be more cost-effective if you're only using the GPU service for a few hours per week. It's hard to tell whether [small vendor offerings] are actually cheaper or not."
While large providers have dominated GPU-enabled virtual machines, smaller vendors are creeping into the market. In addition to Linode, Scaleway and Clever Cloud have both recently introduced GPU-based instances that range from approximately $336 per month to $1,400 per month.
"The need for GPU in the cloud is growing and becoming very competitive as workload increases, especially with the growth of AI," Priestley said. "There are some shifts happening which say there will be more workload accelerators in the cloud, GPU or otherwise, especially for rendering applications."
Currently, Linode's GPU-based instances are only offered as a pilot in limited quantities, but the vendor offers to notify interested customers when additional GPU instances and plans are available.