![](https://zareason.com/wp-content/uploads/2024/12/russia-chip-IA.jpg)
With the aim of becoming independent from the United States, Europe and its allies, Russia is increasing its investments in the semiconductor industry. Since the invasion of Ukraine, sanctions imposed on the country have hampered access to cutting-edge chips and the local development of critical technologies such as photolithography machines. To overcome this situation, Vladimir Putin’s government announced an investment of $2.54 billion by 2030 to boost its semiconductor industry.
First Russian server for AI and high-performance computing
The effort is beginning to show results, albeit timid. THE GravityRussian computer and server manufacturer, launched its first server focused on artificial intelligence (AI) and high-performance computing applications. Called S2124Bthe equipment is equipped with two 48-core CPUs designed and manufactured in Russia.
“The Graviton S2124B server is ideal for AI tasks (in particular, machine learning, language models) as well as for high-performance computing in a wide range of applications. It is capable of delivering up to 480 Tflops of processing power in FP64 format for high-performance computing (HPC) workloads and up to 26,726 Tflops in FP8 and INT8 formats. This makes the server applicable in scenarios that require the processing of large amounts of data,” highlights Graviton on its website.
Full specifications of the CPUs have not been released, but they are believed to be BE-S1000 chips from Baikal Electronics. These processors are based on the ARM Cortex-A75 architecture, operating at up to 2 GHz, and support DDR4-3200 memories. Although technically functional, these chips do not impress compared to cutting-edge solutions on the international market.
Dependence on foreign GPUs
Despite advances in CPU development, Russia has not yet managed to create high-performance GPUs. The S2124B server supports up to eight GPU cards for AI, but the units used are not of Russian origin. It’s likely that Graviton is using NVIDIA H100 GPUs, as the performance numbers released by the company match the capabilities of this AI-focused chip.
This dependence on foreign GPUs highlights an important bottleneck: international sanctions restrict access to components from companies such as NVIDIA and AMD, forcing Russian manufacturers to resort to intermediaries and parallel channels to obtain these technologies.
Source: https://www.hardware.com.br/noticias/cpu-putin-mercado-ia.html