Fresh on the heels of AMD's release of its Milan server CPUs, Chinese cloud computing giant Tencent Cloud has unveiled a new generation of servers powered by the latest processors, saying it brings a multi-fold performance boost.
The cloud server, called Star Lake SA3, offers up to 232 cores of single-node computing power, with application scenarios covering video codec, offline rendering, database, and AI.
Tencent Cloud said the server delivers more than 220% performance improvement for AI scenarios, for example.
The server has added SEV-SNP features for more comprehensive data security protection. Its air-cooled cooling capacity has been improved by 22%, and the compatibility of the liquid-cooled design has reduced carbon emissions by 8%, Tencent Cloud said.
The server adopts Tencent Cloud's latest-generation CBS block storage engine, which enables the entire storage capacity to reach 8GB/s and shortens IO latency by 63%, which can better support low-latency and high-stability applications.
The server's bottom layer is based on the new 100G network architecture, which can provide up to 100G of ultra-large network bandwidth.
Its single-node network bandwidth has increased by up to 400% compared to the previous generation, and network packet forwarding performance has increased by more than 260%, so it can also perform well in high-traffic, high-forwarding network-based application scenarios.
Star Lake is Tencent Cloud's in-house developed server brand, and after a year of research and development in 2019, Tencent Cloud has released Star Lake's first server hardware.
Tencent Cloud said that in the past year, the application scale of SA3's predecessor, Star Lake SA2, grew 30 times and the customer scale grew 50 times.
During last year's November 11 shopping festival, the largest shopping day in the Chinese e-commerce field, Star Lake SA2 became the computing platform of choice for many e-commerce companies, supporting 90% of the computing power demand in the scenario, the company said.
Tencent is now running more than one million servers across the network, the first company in China and the fifth in the world to operate more than one million servers.
AMD hopes to grab more share from Intel
The Star Lake SA3 is powered by a server chip called Milan that AMD announced on Monday, which aims to take more market share from rival Intel.
The chip processes faster than the best data center chips available today, AMD said.
AMD completed the chip's design and commissioned TSMC to use a 7-nanometer chip manufacturing process to achieve mass production.
AMD currently cooperates with third-party chip manufacturers to mass production. Intel uses a different strategy, manufacturing its own chips, but the delay in the latest generation of chip manufacturing technology has plagued it.
In comparison, the Milan chip and its predecessor have outperformed Intel's chips, helping AMD to grab more market share and win over customers including Google.
But Intel is about to fight back. Analysts expect Intel to launch its latest Ice Lake server chip in the coming weeks, which will be the first server chip to be mass-produced using Intel's 10-nanometer manufacturing process.
Intel's 10-nanometer manufacturing process produces chips with performance equivalent to the 7-nanometer manufacturing process of Milan chips.
(Source: Tencent Cloud)