Intel has installed a powerful supercomputer: Equipped with more than 63 thousand GPUs and 21 thousand CPUs, performing 2 billion billion calculations / second

Intel and Argonne National Laboratory (USA) recently successfully installed the Aurora supercomputer, thereby achieving an important step to bring this powerful system into official operation. This is considered good news for Intel, as this supercomputer has continuously faced delays in going into operation for a long time.

Equipped with Xeon Max CPUs and GPU Max family for systems that need extreme computing power, the Aurora supercomputer’s performance is up to 2 ExaFLOPs (ExaFLOPs are a measure of the performance of a supercomputer, can be perform 1 billion billion operations/second). This means that, if Aurora goes into operation, it will be the most powerful supercomputer in the world today, topping the top 500 most powerful supercomputers. For comparison, HP’s Frontier supercomputer is currently the world’s No. 1 supercomputer with typical performance of 1,102 ExaFLOPs, and about 1.68 ExaFLOPs at full capacity.

To do this, Aurora integrates 21.2 thousand Xeon CPUs belonging to Intel’s latest Sapphire-Rapid SP product line today. This is accompanied by about 63.7 thousand graphics processing units (GPUs) of the Ponte Vecchio architecture, allowing it to provide a maximum gain of 2.12 PB/s and a maximum halved bandwidth of 0.69 PB/s.

To build Aurora, engineers installed 10,624 compute slabs, with each blade integrating 2 Xeon Max CPUs, 6 GPU Max and weighing nearly 32kg. All compute slabs are housed in 166 racks, each containing 64 slabs. They are divided into eight rows, occupying the area of ​​two professional basketball courts in the ALCF data center.

As for memory capacity, the Aurora supercomputer has 10.9 PB of DDR5 RAM, 1.36 PB of HBM memory through the CPU, and 8.16 PB of HBM memory through the GPU. Finally, Aurora’s total storage capacity is 220TB. In terms of performance, Intel’s GPU Max is said to be 30% more powerful than Nvidia’s H100 series of GPUs on various tasks, and about 50% for AI tasks (according to software vendor data). independent Ansys).

It is known that one of the tasks Aurora will perform includes the application of generative AI technology in scientific research. This includes curing deadly diseases, or understanding the effects of climate change, which require advanced computing technology on a large scale.

Along with Intel, Meta, Microsoft and Nvidia are some of the other big companies that are currently developing their own supercomputer projects. IBM recently outlined its plans to create a quantum supercomputer. Supercomputers have more memory and are capable of performing more tasks at higher speeds than regular computers. When it comes to high-performance computing tasks that require a system to operate at high power on AI, simulation, and data analysis, supercomputers are key. The researchers believe that supercomputers have the ability to promote research and innovation, which in turn can help them solve social problems.

Refer to Wccftech