TL;DR
- MSU-270’s impressive 400 PetaFLOPS for AI processing.
- Speculations on GPU sources for the supercomputer.
- The effects on various scientific disciplines, education, and AI research.
- The dedication of MSU to producing specialists and experts in AI.
- Integration of the supercomputer into the Russian scientific network.
Lomonosov Moscow State University (MSU) has unveiled its cutting-edge supercomputer, christened as the MSU-270, boasting a staggering peak computational prowess of 400 PetaFLOPS specialized in ‘AI’ processing. This formidable computing behemoth is slated for multifarious applications, primarily centered around artificial intelligence (AI) and high-performance computing (HPC), particularly for the intricate task of training extensive AI models.
The intricate pedigree of the MSU-270 lies in its utilization of the most contemporary graphics accelerators, although MSU has chosen to maintain a veil of secrecy concerning their origin.
“We have instigated the inauguration of this monumental undertaking,” proclaimed Viktor Sadovnichy, the luminary figure at the helm of MSU, as he unveiled the project’s grandeur at the Faculty of Computational Mathematics and Cybernetics. We truly feel the MSU-270 supercomputer is an operational masterpiece at this time.
MSU’s MSU-270 supercomputer is founded upon an ensemble of approximately 100 state-of-the-art graphics accelerators, complemented by novel systems for energy provisioning, thermal management, and communication infrastructure. This sophisticated apparatus aligns harmoniously with MSU’s ambitious long-term agenda, a vision poised for multifaceted advancements on the horizon by 2030.
Given Russia’s less prominent role in the GPU sphere, it is highly probable that the machine harnesses compute GPUs sourced from illustrious companies such as AMD, Biren Technology, Intel, or Nvidia. Notably, the most advanced GPUs from AMD and Nvidia are precluded from export to Russia and China under the latest U.S. regulatory constraints. Historically, MSU’s supercomputing endeavors have exclusively relied on Nvidia’s hardware.
It is worth noting that the MSU-270’s formidable 400 PetaFLOPS are attributed to what is colloquially referred to as ‘AI’ PetaFLOPS, likely implying the utilization of the FP16 data format. In contrast, Russia’s preeminent supercomputer boasts an Rmax performance benchmarking at approximately 21.5 FP64 PetaFLOPS and an Rpeak performance reaching an approximate threshold of 29.5 FP64 PetaFLOPs.
Alas, MSU has opted to withhold disclosure pertaining to the FP64 performance metrics of the MSU-270 machine, though it is reasonable to speculate that it will pale in comparison to the staggering 400 PetaFLOPS. Nevertheless, he was able to advance and take the position of prime minister of Russia.
The MSU-270’s superior computing power will help create new and sustainable intelligence tools. This includes establishing state-of-the-art processes for data analysis and implementing mechanisms to protect AI services from external threats. In addition, the MSU-270 will play an important role in education and become an important resource for the creation and training of intelligent professionals.
“For the past two years, a program at our university aimed at producing experts in the field of artificial intelligence has been totally managed by our great professors. Our vocational training will benefit from the wide range and diversity of this excellent interview system. Creating qualified employees is our main goal.
The MSU-270 should be integrated into Russia’s vast network of supercomputing research centres stretching back to the MSU.
source(S): tom’sHARDWARE
