IT/Datacenter & Super Computing News - Page 1
The National Oceanic and Atmospheric Administration has just replaced two outdated supercomputers with new supercomputers that will bring massive upgrades to weather forecasting.
The National Oceanic and Atmospheric Administration (NOAA) announced back in 2020 that it would be replacing some of its older, now-outdated supercomputers that are used to run weather forecasts and predict patterns. NOAA has now replaced the older models, which were Cry and IBM supercomputers located in Reston, Virginia, Florida, and Orlando - with two Hewlett Packard Enterprise (HPE) Cray supercomputers.
These two supercomputers come with 2,560 AMD Epyc Rome 64-core 7742 server CPUs that collectively provide 327,680 cores that are capable of operating up to 12.1 petaflops. This new setup is approximately three times faster than what NOAA was previously using, and with the new upgraded power, the agency believes it will be able to provide more accurate and detailed weather forecasts to the public. Notably, the two new NOAA supercomputer systems are called Dogwood and Cactus, and rank as the 49th and 50th fastest supercomputers in the world.
The latest generation of Sunway supercomputer rivals the performance of Frontier, which was recently labelled the world's most powerful supercomputer.
According to the researchers behind the latest Sunway supercomputer, the Sunway has over 37 million CPU cores, quadrupling the number found in Frontier, nine petabytes of memory, and 96,000 semi-independent computer systems referred to as 'nodes,' that can exchange data at rates greater than 23 petabytes per second. The Sunway is capable of exascale computing, allegedly up to 5.3 exaFLOPS (5.3 quintillion floating-point operations per second).
The research team trained an artificial intelligence (AI) model, named bagualu (meaning alchemist's pot), with 174 trillion parameters using the Sunway. According to the South China Morning Post, this number rivals that of the number of synapses in the human brain, though some estimates of the true number of synapses go as high as 1,000 trillion.
Atos is building the new MareNostrum 5 supercomputer, with a new $160 million contract signed in part of the EuroHPC JU initiative, and once it's built it will be the fastest supercomputer in the European Union.
NVIDIA's new Grace Superchips will be powering the new Atos MareNostrum 5 supercomputer, which will be built in Spain and delivered to the Barcelona Supercomputing Centre (BSC) in 2023. Inside, the new MareNostrum 5 supercomputer will pack 314 petaflops of FP64 computing performance.
The compute and storage partitions will be operational within the year, reports HPC Wire, adding that the remainder of the MareNostrum 5 supercomputer to be "operational within the year". NVIDIA has also added that the system is "expected to enter deployment in 2023". BSC says that the new MareNostrum 5 supercomputer will be "fully powered with green energy, and will utilize heat reuse technology".
The new NREL Kestrel supercomputer details have been unveiled, with the National Renewable Energy Laboratory (NREL) division of the US Department of Energy (DOE) packing some serious horsepower into its new supercomputer.
We're looking at a Minajatwa of silicon between Intel Sapphire Rapids Xeon CPUs, AMD EPYC "Genoa" CPUs, and NVIDIA H100 GPU accelerators. We have 44 petaflops of peak compute performance, up from just 8 petaflops on the previous-gen Eagle supercomputer.
CPU upgrades are big: Intel Sapphire Rapids Xeon CPUs with 52 cores, 112 threads each -- up from the 18 cores and 36 threads from the Intel Xeon-Gold Skylake CPUs in the Eagle supercomputer.
There are 2304 standard nodes with each node rocking 2 processors in a dual-socket configuration, meaning we have a total of 4608 Intel Sapphire Rapids-SP CPUs and a total of an insane 239,616 cores and even-more-insane 479,232 threads... yeah, you're reading those numbers right.
It's a huge day for supercomputers, with AMD powering the world's fastest supercomputer -- the Oak Ridge National Laboratory (ORNL) "Frontier" supercomputer -- and now NVIDIA has announced it is powering the "Venado" supercomputer.
NVIDIA announced last week that Taiwan tech giants were preparing NVIDIA Grace CPU-powered servers, and now we have our first: the new Venado supercomputer that's being constructed by the Los Alamos National Laboratory. The new VENDAO supercomputer is capable of a huge 10 exaflops of peak AI performance.
It feels like a little bit of "me too" with NVIDIA's announcement of the Venado supercomputer, but more concrete details will be provided at the International Supercomputing Conference in Hamburg, Germany later today. The new AMD CPU + GPU-powered Frontier supercomputer was teased yesterday, and now we have the NVIDIA CPU + GPU-powered Venado supercomputer here today.
The massively powerful new Oak Ridge National Laboratory (ORNL) supercomputer dubbed "Frontier" has broken the 1.1 exaflops barrier, becoming the first machine in the world to breach the historic exascale barrier.
The Department of Energy (DOE) will operate the new Frontier supercomputer in Tennessee, USA, with the system costing up to $1.8 billion to build and is now the world's fastest supercomputer, overtaking the Fugaku supercomputer in Japan. ORNL's new supercomputer is powered by AMD 3rd Gen EPYC CPUs and AMD's newest Radeon Instinct MI250X GPUs.
Inside, we have 74 purpose-built HPE Cray EX supercomputer cabinets with 9408 AMD EPYC CPUs for a total of (a bonkers) 8,730,112 processing cores, and 37,632 AMD Instinct MI250X GPUs with a power efficiency rating of 52.23 gigaflops/watt. There's 700PB of data storage with peak write speeds of an insane 5TB/sec (5000GB/sec).
We all know that turning ray tracing on radically reduces performance, so any help in the performance department is not only welcomed, it's begged for.
A team of researchers from the UK, US, and Portugal have suggested that there's some huge untapped performance using a hybrid of classical ray tracing algorithms with quantum computing. Ray tracing workloads that were boosted by up to 190% by quantum computing.
How? By limiting the amount of computations needed by each individual ray. The researchers demonstrated this by rendering a small, 128x128 ray traced image in three approaches: classical rendering, non-optimized quantum rendering, and optimized quantum rendering.
NVIDIA announced at Computex 2022 today in Taiwan that Taiwan's leading computer makers are preparing their first wave of systems powered by NVIDIA's new Grace CPU Superchip and Grace Hopper Superchip.
The likes of ASUS, Foxconn Industrial Internet, GIGABYTE, QCT, Supermicro, and Wiwynn will have NVIDIA Grace-powered systems in the first half of 2023. The upcoming servers are destined for the world of AI, high-performance computing (HPC), cloud gaming, digital twins, and so much more.
Ian Buck, vice president of Hyperscale and HPC at NVIDIA said: "A new type of data center is emerging-AI factories that process and refine mountains of data to produce intelligence-and NVIDIA is working closely with our Taiwan partners to build the systems that enable this transformation. These new systems from our partners, powered by our Grace Superchips, will bring the power of accelerated computing to new markets and industries globally".
GIGABYTE has just announced its new "supercharged, scalable server" in the G492-PD0, which supports NVIDIA's new Ampere Altra Max or Altra processor.
The new GIGABYTE G492-PD0 not only packs NVIDIA's new Ampere Altra Max or Altra processor, but also NVIDIA HGX A100 Tensor Core GPUs for the highest performance in cloud infrastructure, HPC, AI, and more. Ampere's Altra Max CPU packs 128 Armv8.2 cores per socket with Arm's M1 core, with high performance efficiency and minimized total cost of ownership.
GIGABYTE is using a novel cooling solution that dedicates a cooling chamber for the NVIDIA accelerators and GPUs used in the networking expansion slots... something that allows for the highest possible airflow to cool the high-performance components.
NVIDIA's new Hopper GPU architecture has been unveiled and with it comes some new DGX H100 systems powered by the monster new NVIDIA H100 GPU.
The new NVIDIA DGX H100 system has 8 x H100 GPUs per system, all connected as one gigantic insane GPU through 4th-Generation NVIDIA NVLink connectivity. This enables up to 32 petaflops at new FP8 precision, a gigantic 6x performance improvement over the previous-gen Ampere-based GPUs.
NVIDIA will be using the new Hopper-based DGX H100 systems as the "building blocks" of the next-gen NVIDIA DGX POD and NVIDIA DGX SuperPOD AI infrastructure platforms. NVIDIA's new DGX SuperPOD architecture has the new NVIDIA NVLink Switch System -- capable of up to 32 nodes and a total of 256 x H100 GPUs. At this level, we're talking about 1 exaflops of FP8 AI performance, also 6x more than its predecessor.