Samsung Electronics has increased the production volume of its 8GB high bandwidth memory-2 (HBM2) to meet growing market needs across a wide range of applications including AI, high-performance computing (HPC), advanced graphics, network systems and enterprise servers. Samsung's 8GB HBM2 consists of eight 8Gb HBM2 dies and a buffer die at the bottom of the stack, which are all vertically interconnected by TSVs and microbumps. With each die containing over 5,000 TSVs, a single Samsung 8GB HBM2 package has over 40,000 TSVs, according to the vendor. The utilization of so many TSVs, including spares, ensures high performance, by enabling data paths to be switched to different TSVs when a delay in data transmission occurs. The HBM2 is also designed to prevent overheating beyond certain temperature to guarantee high reliability. In meeting increasing market demand, Samsung anticipates that its volume production of the 8GB HBM2 will cover more than 50% of its HBM2 production by the first half of 2018.

Interconnect solution supplier Mellanox Technologies has announced the ConnectX-4 single/dual-port 100Gb/s virtual protocol interconnect (VPI) adapter, an end-to-end 100Gb/s InfiniBand interconnect solution. Doubling the throughput of the previous generation, the ConnectX-4 adapter delivers the consistent, high-performance and low latency required for high performance computing (HPC), cloud, Web 2.0 and enterprise applications to process and fulfill requests in real-time, according to the vendor. Mellanox's ConnectX-4 VPI adapter delivers 10, 20, 25, 40, 50, 56 and 100Gb/s throughput supporting both the InfiniBand and the Ethernet standard protocols, and the flexibility to connect any CPU architecture - x86, GPU, POWER, ARM, FPGA and more. With performance at 150 million messages per second, latency of 0.7usec, and smart acceleration engines such as RDMA, GPUDirect and SR-IOV, the ConnectX-4 will enable efficient compute and storage platforms.

Nvidia has unveiled the Tesla K40 GPU accelerator. Providing double the memory and up to 40% higher performance than its predecessor, the Tesla K20X GPU accelerator, and 10 times higher performance than today's fastest CPU, the Tesla K40 GPU is optimized for big data analytics and large-scale scientific workloads, the company claimed. GPU accelerators have gone mainstream in the HPC and supercomputing industries, enabling engineers and researchers to consistently drive innovation and scientific discovery, said Sumit Gupta, general manager of Tesla Accelerated Computing products at Nvidia. With the breakthrough performance and higher memory capacity of the Tesla K40 GPU, enterprise customers can quickly crunch through massive volumes of data generated by their big data analytics applications. Key features of the Tesla K40 GPU accelerator include: 12GB of ultra-fast GDDR5 memory; 2,880 CUDA parallel processing cores; dynamic parallelism; and the PCIe Gen-3 interconnect support. In a related announcement, the Texas Advanced Computing Center (TACC) at the University of Texas at Austin plans to deploy "Maverick," a new interactive, remote visualization and data analysis system powered by Nvidia's Tesla K40 GPU accelerators. Maverick is expected to be fully operational in January 2014.