OVHcloud, the European Cloud leader and a part of the NVIDIA Partner Network continues to develop its portfolio of AI solutions, today adding new NVIDIA GPU product offerings that form an integral part of its strategic vision for artificial intelligence. Acknowledging the tremendous impact AI will have in the years to come, OVHcloud is on a mission to help customers grow their businesses, uniting an ecosystem through innovative, easy and affordable AI solutions, featuring transparent, ethical and open models that preserve data privacy.
With over 20 years of expertise in infrastructure, through a unique vertically integrated industrial model, OVHcloud is designing AI-enabled infrastructures, which include new best-of-breed NVIDIA H100 and A100 Tensor Core GPUs. Customers will be able to choose from many options to power their most ambitious machine learning workloads, including large language models.
Adding to critically acclaimed options already offered at a competitive price with older generation NVIDIA V100 and NVIDIA V100S GPUs, the Group today announced new offerings based on the following GPUs: NVIDIA H100, NVIDIA A100, NVIDIA L40S and NVIDIA L4, with effective deployment ramping up in the coming weeks.
New GPU instances with NVIDIA A100 for deep learning training and inference
New NVIDIA A100 80GB powered GPU instances are immediately available and let AI specialists run complex projects on highly specialized NVIDIA Tensor Cores. With exceptional abilities in deep learning training, A100 is also ideally suited to run inference thanks to various optimizations in tackling those workloads, including LLM-related projects. High-performance computing is another playing field where the A100 GPU instances help unlock the next-generation of discoveries, through advanced simulations, thanks to double-precision compute and high-bandwidth memory.
A100-based public cloud instances can be configured as A100-180 with 1x A100, 15 vCore and 180GB of RAM, A100-360 with 2x A100, 30 vCore and 360GB of RAM and A100-720 with 4x A100, 60 vCore and 720GB of RAM.
New GPU instances with NVIDIA H100 for deep learning training
OVHcloud is also announcing upcoming H100-based GPU instances built around NVIDIA’s latest accelerator with a compute power starting at 26 petaFLOPS (FP64) per PCIe GPU. Purpose-built for the most demanding AI models, the NVIDIA H100 is the de facto choice for innovation in AI, whether accelerating LLMs with its Transformer Engine, or creating generative AI applications.
For the most demanding use cases, such as extreme fine-tuning and training, the Group will offer NVIDIA H100 SXM-based solutions. With 67 TFlops of FP64 compute power and a higher GPU bandwidth, this select offering will showcase the full power of the NVIDIA Hopper GPU architecture.
New GPU instances and bare-metal servers with NVIDIA L4 and L40S
The Group also today unveiled GPU instances featuring NVIDIA L4 GPUs with 24GB of memory. The L4, based on the NVIDIA Ada Lovelace GPU architecture, is a universal GPU for every workload with enhanced AI and video capabilities. It provides efficient compute resources for graphics, simulation, data science and data analytics.
The NVIDIA L40S GPU with 48GB of memory is also joining the Group’s GPU instances. NVIDIA L40S benefits from fourth-generation Tensor Cores and FP8 Transformer Engine providing robust performance for AI workloads both in training and inferencing.
These GPUs will be available through public cloud instances as well as in dedicated bare-metal servers with L4 in SCALE-GPU and L40S in HGR-AI.
Establishing a foundation to supercharge customer AI journeys
Thanks to an unprecedented choice of NVIDIA GPU architectures, OVHcloud now delivers an AI-designed infrastructure for AI engineers, researchers, data scientists and data practitioners that can leverage the elasticity of the cloud to support their needs from training to running inference.
Furthermore, OVHcloud will gradually add NVIDIA H100 and A100 options to its set of comprehensive AI PaaS solutions designed to accompany the data life cycle: AI Notebooks, AI Training and AI Deploy. OVHcloud AI solutions act as a complete set of tools that are easy to use, and designed to explore data, train models and serve them into production.
Executing on a strong dedicated AI roadmap, OVHcloud is set to announce in the weeks to come a wave of AI innovations designed to further help its customers in navigating this new paradigm.
Pricing & availability
GPU instances based on NVIDIA A100 are available now on public cloud starting at 2,75 EUR per hour for A100-180, 5,50 EUR per hour for A100-360 and 11 EUR per hour for A100-720. Managed AI solutions based on A100 will be available later at 3 EUR per hour. NVIDIA H100 PCIe-based instances in the public cloud universe will launch later this fall with pre-reservation starting now. Customers interested in upcoming NVIDIA H100 SXM offers can apply for a personalized quote.
NVIDIA L4 and L40S GPU instances will be available in October. New bare-metal servers featuring NVIDIA L4 and L40S GPUs are expected to launch this fall.
OVHcloud Launches Second Generation Bare Metal Scale Servers
Posted in Commentary with tags OVH on November 15, 2023 by itnerdOVHcloud, the European cloud leader, today announced the launch of its second generation of dedicated Bare Metal Scale servers. Hot on the heels of the critically-acclaimed first Scale line-up launched in 2021, this new generation of high-performance dedicated servers brings even more compute power.
Addressing verticals such as healthcare, finance, industry, public sectors and media, the new generation of Scale Bare Metal servers uses powerful processors from AMD and NVIDIA to answer customers’ most intensive use cases such as confidential computing, big data, analytics as well as virtualization and container orchestration or grid computing.
Modern CPU for Integrity, Compute Power for Big Data and Analytics
In a world where data is at the center of every modern business, customers need robust platforms where they can process their sensitive data. The new OVHcloud Bare Metal second-generation Scale servers benefit from AMD Secure Encrypted Virtualization (SEV) found in the fourth generation of AMD EPYC processors to allow for strengthening VM isolation.
Offering high-performance and scalable Bare Metal servers, OVHcloud can take advantage of the high core count of AMD EPYC processors, in single- or dual-socket configurations. This allows the hardware to deliver tremendous compute power so customers can process and analyze large sets of data, in real time. Customers can gain insights from their data, facilitating decision-making. Not only can customers process, treat and harness data in a snap, they can also store large amounts of data with up to 46 TB of NVMe storage.
Best-of-Breed CPU, DDR5 and Large IO
Leveraging AMD’s most advanced Zen 4 CPU architecture, the second generation of Scale servers boasts fourth-generation EPYC processors, code named Genoa. The range starts with AMD EPYC 16 cores / 32 threads and scales to 96 cores / 192 threads in a single-socket configuration (SCALE-a1 to SCALE-a6).
The second-generation Scale line-up is offered in a dual socket configuration with core counts all the way up to 256 cores and 512 threads, leveraging fourth-generation AMD EPYC processors, code named Genoa and Bergamo, with the latter featuring Zen 4C cores. SCALE-a7 offers a total of 192 cores / 384 threads (2P AMD EPYC 9654 code named Genoa) and the SCALE-a8 variant comes with 256 cores / 512 threads (2P AMD EPYC 9754 code named Bergamo).
Available for the first time at OVHcloud with DDR5 memory configuration ranging from 128 GB to 1 TB (4800 MHz), new Scale references can benefit from generous storage options – from the included 2 x 960 GB NVMe up to 6 x 7,68 TB SSD NVMe.
NVIDIA L4 Tensor Core GPU for Machine Learning
Alongside new second-generation Scale CPU-based references, OVHcloud is also introducing SCALE-GPU dedicated servers leveraging the NVIDIA L4 GPU for data centers.
Based on the NVIDIA Ada Lovelace GPU architecture, the NVIDIA L4 is a universal GPU for efficient video, AI and graphics-intensive workloads. The hardware is optimized for use cases like machine learning with image classification, natural language processing, object detection and VDI.
Unmetered Guaranteed Bandwidth with up to 25 Gbps Private Network
OVHcloud’s new second-generation Bare Metal Scale benefits from all the advantages of dedicated servers when it comes to flexibility, control, performance, or availability as well as 1 Gbps to 10 Gbps guaranteed and unmetered public bandwidth and up to 25 Gbps of unmetered private bandwidth. Through this offering, OVHcloud also provides companies with the best possible support for their workloads, environments, and ad-hoc technologies, while guaranteeing predictable pricing.
Data Protection and Sustainability
Second generation Scale Bare Metal servers also benefit from OVHcloud’s well-known expertise in infrastructure, offering a trusted cloud in environmentally friendly data centers. With the highest security and data protection standards in the form of ISO 27001 certification and HDS, data is immune to extra-territorial laws. OVHcloud data centers take advantage of the company’s unique industrial model with a water-cooling system that contributes in a sustainable cloud allowing to reach best-in-class PUE/WUE indexes (see more data here).
Availability
Bare Metal SCALE-a1 (16c/32t), SCALE-a2 (24c/48t), SCALE-a3 (32c/64t), SCALE-a4 (48c/96t), SCALE-a5 (64c/128t) and SCALE-a6 (96c/192t) are available for order now, in our datacenters located in Canada, France, Germany, Poland, UK and USA. Roll-out in the APAC region is planned from December to March. Bare Metal SCALE-a7 (192c/384t) and SCALE-a8 (256c/512t) are available now as build-to-order references. SCALE-GPU with NVIDIA L4 GPUs are planned for early December availability.
Leave a comment »