NVIDIA H100 Tensor Core GPUs are now available and L40S GPUs are coming to Oracle Cloud Infrastructure.
Boosting Performance with NVIDIA GPUs
Because ground-breaking discoveries are being driven by generative artificial intelligence and large language models (LLMs), the computing requirements for training and inference are rapidly increasing.
These cutting-edge applications of generative artificial intelligence need for full-stack accelerated computation, beginning with infrastructure that is up to date and capable of managing huge workloads in a timely and accurate manner. Oracle Cloud Infrastructure made the announcement today that NVIDIA H100 Tensor Core GPUs are now generally available on OCI Compute. In addition, NVIDIA L40S GPUs are scheduled to become available in the near future.
GPU Instance of the NVIDIA H100 Tensor Core Available on OCI
OCI Compute bare-metal instances with NVIDIA H100 GPUs, which are powered by the NVIDIA Hopper architecture, provide an order-of-magnitude jump for large-scale AI and high-performance computing. These instances provide unrivaled performance, scalability, and agility for any kind of task.
When compared to utilizing the NVIDIA A100 Tensor Core GPU, organizations who use NVIDIA H100 GPUs get up to a 30x improvement in the performance of AI inference and a 4x jump in the performance of AI training. The H100 graphics processing unit was developed specifically for resource-intensive computational operations such as the training of LLMs and inference while they were being performed.
The BM.GPU.H100.8 OCI Compute form has been outfitted with a total of eight NVIDIA H100 GPUs, each of which has 80GB of HBM2 GPU memory. Through the use of NVIDIA NVSwitch and NVLink 4.0 technology, the 3.2 terabits per second (TB/s) of bisectional bandwidth that is shared across the eight GPUs allows each GPU to connect directly with all seven of the other GPUs. The form factor consists of 16 local NVMe SSDs, each of which has a capacity of 3.84 terabytes (TB), as well as 112 cores of system memory and 4th Generation Intel Xeon CPU processors with a total of 112 cores apiece.
In a word, this form is tailored specifically for the most difficult tasks faced by enterprises.
OCI Supercluster gives businesses the ability to expand their NVIDIA H100 GPU utilization via a high-performance, ultra-low-latency network from a single node all the way up to tens of thousands of H100 GPUs, depending on the timeframes and sizes of the workloads they need to complete.
NVIDIA L40S Graphics Processing Unit Instance on OCI
The NVIDIA L40S GPU is a universal GPU for the data center. It is built on the NVIDIA Ada Lovelace architecture and delivers groundbreaking multi-workload acceleration for LLM inference and training, visual computing, and video applications. Later on in 2018, early access to the OCI Compute bare-metal instances equipped with NVIDIA L40S GPUs will be made accessible. The public availability of these instances is not expected until the beginning of 2024.
For AI workloads ranging from small to medium in scale, as well as for graphics and video computing work, these instances will provide an alternative to the NVIDIA H100 and A100 GPU instances. When compared to the NVIDIA A100, the NVIDIA L40S GPU offers up to a 20% performance gain for generative artificial intelligence workloads and as much as a 70% improvement in fine-tuning artificial intelligence models.
The BM.GPU.L40S.4 OCI Compute form is equipped with four NVIDIA L40S GPUs, the most recent version of Intel Xeon CPU with up to 112 cores, 1 terabyte (TB) of system memory, 15.36 terabytes (TB) of low-latency NVMe local storage for caching data, and 400 gigabytes per second (GB/s) of cluster network bandwidth. This instance was built to tackle a broad variety of use cases, ranging from LLM training, fine-tuning, and inference to NVIDIA Omniverse workloads and industrial digitalization, 3D graphics and rendering, video transcoding, and FP32 HPC. Other use cases that this instance can handle include industrial digitalization and video transcoding.
NVIDIA and OCI: Enterprise Artificial Intelligence
By providing them with state-of-the-art NVIDIA H100 and L40S GPU-accelerated infrastructure, this cooperation between OCI and NVIDIA will make it possible for companies of all sizes to participate in the generative AI revolution.
Having access to NVIDIA GPU-accelerated instances, on the other hand, may not be enough. Having the most effective software layer is necessary in order to release the full power of NVIDIA GPUs running on OCI Compute. With the assistance of support services, NVIDIA AI Enterprise makes the creation and deployment of enterprise-grade accelerated artificial intelligence applications more simpler. This is accomplished by providing open-source containers and frameworks that are optimized for the underlying NVIDIA GPU infrastructure.
[…] Trust Authority SaaS verifies evidence with NRAS and collects the NVIDIA-signed token using the […]
[…] ML for applications like product recommendations, scientific computing, and gaming often turn to NVIDIA GPUs on Google Cloud. They must keep an eye on the GPU performance indicators in order to comprehend the […]
[…] NVIDIA was one of the first companies to lay bets on artificial intelligence (AI), and the company is now reaping the benefits of those investments, as seen by the phenomenal growth it has had over the past year. Deep Learning Super Sampling, also known as DLSS, is a technology that was developed with the purpose of boosting the performance of video games by utilizing the capabilities of artificial intelligence (more specifically, a trained neural network). […]
[…] Samsung’s performance-enhanced SF3P technology, which was created with data center CPUs and GPUs in mind, will be made available in […]