Alternative cloud provider Linode has published a report titled “GPU Price-Performance Benchmarking,” highlighting commissioned research by Cloud Spectator evaluating the price-performance of Linode GPU-based virtual machines against those of Amazon Web Services, Microsoft Azure and Google Compute Engine.
Cloud Spectator performed more than 100 benchmark iterations over multiple days to capture accurate performance data. The 16-page report illustrates Cloud Spectator’s findings, concluding that “based on this study, Linode delivers the most cost-effective GPU-based VMs on the market, especially for rendering workloads.”
By using benchmarks on multiple GPU VMs, Cloud Spectator found that Linode offers excellent value when it comes to its introductory GPU-based VMs. The study found that NVIDIA Quadro RTX6000-based VMs from Linode outperformed others in the study when it came to rendering workloads, as well as in performance per dollar spent. The report cites the primary drivers for Linode’s strong price-performance as:
A competitive pricing model compared to the other providers. Linode provides the same hourly and monthly pricing for all regions. Bandwidth is bundled with each virtual machine (VM): the 32GB GPU VM comes with 10TB included in the $1000/mo price. With other providers, pricing is often more complicated, sometimes varying by region and often requiring users to pay per GB of outgoing bandwidth, which is not always easy to plan for from month to month.
Solid performance scores, especially for video and image rendering workloads. Performance per dollar on 32GB Linodes with the RTX6000 consistently outperformed competitors in the study for Gaussian blur, depth of field and V-ray (a 3-D rendering benchmark).
“The report shows how various workloads perform on Linode compared to GPU-based virtual machines from the three largest hyperscale cloud service providers,” said Alex Peña, Lead Research and Development Engineer, Linode, “Using a standardized and repeatable testing methodology, Cloud Spectator tested four different GPU-based VMs to evaluate performance for a wide range of workloads, and the results show how users can achieve improved performance and cost per instance with Linode.”