What's new!

We're excited to announce updates to the Denvr Cloud! Here's what's new:

October 2024

1. Applications: early-access release for Application Catalog. These are focused on data scientists who prefer a fast-start environment like PyTorch, RAPIDS and JAX along with a JupyterLab interface.

2. Support tickets: we have integrated web support tickets into the console. This new system is easier to use than our previous ticketing platform.

3. Support Slack: Denvr Support is available on Slack for tenants with reserved clusters. We also support Teams and Discord if required.

August 2024

1. Intel Gaudi 2: general availability of Intel Gaudi 2 accelerators in bare metal or VM configurations. Gaudi2 provides superior TCO for both training and inferencing workloads.

2. NVIDIA H100: new availability of H100 nodes for on-demand and reserved use including 3,200G InfiniBand

June 2024

1. Virtual machines: simplified provisioning to optionally install GPU drivers, mount network-attached storage, and specify root disk size

April 2024

1. Virtual machines: test nodes available with NVIDIA GH200 Grace Hopper Superchip!

2. Docs: launched new documentation portal Denvr Docs including updated Quick Start guides, feature overviews, API Reference, and FAQs.

February 2024

1. Virtual machines: new configurations available including NVIDIA A40, AMD Mi250, AMD cpu-only as well as Windows operating system images.

2. Slack: launch of Slack Connect for direct access to Denvr ML support analysts for Bare Metal clusters.

November 2023

1. New region: launch of Houston region (cluster name HOU1) with NVIDIA A100 GPUs in Bare Metal configurations including 1.6 Tbps InfiniBand per node.

July 2023

1. Virtual machines: UI and API support for virtual machines. VMs should be used by teams requiring full operating systems and services. We currently provide Ubuntu 20 and 22 with pre-installed drivers for NVIDIA CUDA, NvLink, and InfiniBand.

2. Custom applications: Support for user containers in addition to our published application library. Containers are pulled from remote DockerHub and can be used with single and multi-node configuraitons.

3. InfiniBand: the MSC-1 cluster has been upgraded to 800G non-blocking networking for use by HGX A100 multi-node configurations.

April 2023

1. Scratch storage: HGX A100 nodes have upgraded scratch providing 20 GB/s read/write performance (400% improvement). Scratch including a memory cache that achieves total throughput of 20-30GB/s under common workloads, helping users to achieve higher GPU utilization.

2. Applications: PyTorch 2.0 was released by NVIDIA in their NGC container catalog and is now a supported in Denvr AI Cloud.

3. User interface: Ongoing enhancements to simplify provisioning. A new multi-node Ubuntu 22 package is available for multi-node users that deploys full InfiniBand clusters (controller and worker nodes using OpenMPI) in less than 45 seconds.

January 2023

1. New region: launch of Calgary Modular SuperCluster (cluster name MSC1) with NVIDIA A100 GPUs in both on-demand and reserved node configurations.

2. User management: tenant admins can now create users, lock and unlock accounts, and reset passwords without requiring support assistance.

3. Cost and usage: the cloud console now displays full visibility of tenant costs by application, user, and resource type.

4. Matlab: new application available, licenses are self-supplied by logging into the application with your Matlab userid.

5. Tenant limits: we have implemented a cap of 10 on-demand GPUs per Tenant to ensure availability for all Denvr Cloud users, . You can request changes to the limit by contacting support or your account manager.