InfiniBand
PulseAugur coverage of InfiniBand — every cluster mentioning InfiniBand across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
SF Compute enables programmable data centers for GPU trading
SF Compute, a company focused on the GPU market, is enabling programmatic control over data centers, akin to software-defined networking's impact on cloud infrastructure. CEO Evan Conrad likens their expansion to operat…
-
Nvidia GPU Terms Explained for AI Engineers
This article defines essential GPU terms for AI engineers, covering concepts from CUDA cores to NVLink. It explains how these NVIDIA GPU concepts influence model fitting, execution speed, and scalability. The piece aims…
-
Nvidia's NVSwitch faces competition in AI networking from emerging standards
Nvidia's NVSwitch technology is currently the dominant solution for scaling AI networks, similar to how InfiniBand has historically led in low-latency networking. However, emerging standards such as UALink and the Ultra…
-
New tools and research tackle GPU optimization for AI workloads
Several research papers and a new open-source tool address challenges in optimizing AI workloads on GPUs. COMPASS-ABS aims to reduce fragmentation in shared GPU clusters for deep learning training, improving resource ut…
-
GPU rental costs rise as AI infrastructure complexity outpaces standardization
The cost of renting GPUs, even older models like the NVIDIA H100, has not decreased as expected due to the complex systems required for AI tasks. Unlike traditional computing, AI workloads depend heavily on the integrat…
-
NVIDIA AI Infrastructure Certification Program Launched with Training Resources
A training program and associated resources are available for the NVIDIA-Certified Associate: AI Infrastructure and Operations (NCA-AIIO) certification. The program includes recorded sessions, an exam guide, and practic…
-
China unveils 100,000-card AI supercluster, Sugon 8000
China has unveiled the Sugon 8000 (Dengfeng), its first domestically produced AI supercluster system capable of housing 100,000 cards. This system, which debuted at the World Artificial Intelligence Conference (WAIC), h…
-
Nvidia launches Vera Rubin supercomputing platform for AI and HPC
Nvidia has launched its Vera Rubin supercomputing platform, designed for demanding HPC and AI workloads like climate modeling and energy exploration. The platform integrates Rubin GPUs and Vera CPUs, featuring NVLink, I…
-
NVIDIA Vera Rubin NVL4 system to launch Q4 2026 with 7 exaflops AI power · 2 sources tracked
NVIDIA has announced its Vera Rubin NVL4 system, designed for demanding HPC and AI workloads like climate modeling and fluid dynamics. This platform integrates Rubin GPUs with Vera CPUs, featuring NVLink, InfiniBand, an…
-
Intel-born networking tech eyed as InfiniBand alternative for US DoE
Intel-born networking technology, originally developed for high-performance computing, is being re-evaluated as an alternative to InfiniBand for the United States Department of Energy's supercomputing initiatives. This …
-
InfiniBand networking achieved over Thunderbolt fabric
A developer has created a project called thunderbolt-ibverbs that enables InfiniBand networking over Thunderbolt fabric. This setup connects two AMD Strix Halo mini-PCs using USB4 cables and a custom Linux kernel module…
-
Multi-node training enables scaling foundation models across GPU clusters
Training large foundation models necessitates distributing the workload across numerous GPUs housed in multiple interconnected machines, a process known as multi-node training. This approach is essential for handling mo…