When Nvidia revealed the news about its upcoming powerful Grace CPU, a lot of information was kept sealed, with a few referring to it as a monster Superchip when news broke out that it was expected to have 144-cores.
Now, more information has come to light with Tech Radar sharing more deets about Nvidia’s upcoming server CPU that is meant to power and accelerate AI and HPC workloads.
Nvidia revealed that the 144-core server processor will be manufactured on a specialized version of TSMC’s 4nm process node that’s tuned for the specific characteristics of its products.
The company also published power efficiency figures for its NVLink-C2C interconnect (a technology that makes the dual-CPU superchip possible), which is said to consume 5 times less power than the PCIe 5.0 interface while delivering up to 900 GB/s of throughput.
The Grace CPU Superchip, which was revealed earlier this year is comprised of two Grace CPUs linked up via a high-speed NVLink interconnect, which is a similar technique used in Apple’s M1 Ultra processor unveiled this year.
Nvidia has used this technology to create a 144-core CPU with 1TB/s of memory bandwidth and 396MB on-chip cache that itself claims will be the fastest processor on the market for workloads ranging from AI to HPC and more.
“A new type of data centre has emerged – AI factories that process and refine mountains of data to produce intelligence,” said Nvidia CEO Jensen Huang, when the new chip was first announced.
“The Grace CPU Superchip offers the highest performance, memory bandwidth and Nvidia software platforms in one chip and will shine as the CPU of the world’s AI infrastructure,” Huang added.
The Grace CPU Superchip is expected to offer the highest performance and memory bandwidth on the market
The Grace CPU Superchip which is intended for the server market clearly shows how serious Nvidia is about entering the server market so as to capitalise on the opportunities the future holds in both the AI and cloud computing infrastructure.
Aside from that, Nvidia’s entry into the server market will accelerate the advancement of Arm-based chips, given the fact the x86 architecture is highly dominated by the Intel Xeon and AMD EPYC chips on which they are based.
However, it still remains unclear when the new Nvidia superchip will be widely available. The company recently revealed it will feature in a range of new pre-built servers launching in the first half of 2023. The new systems – from the likes of Asus, Gigabyte, Supermicro and others – will be based on four new 2U reference designs teased by Nvidia at Computex 2022.
The four designs are each specced out to cater for specific use cases, from cloud gaming to digital twins, HPC and AI. Nvidia says the designs can be modified easily by partners to “quickly spin up motherboards leveraging their existing system architectures”.
Also, read: The Blackview Tab 13 Can Use Internet Without WiFi or SIM Card
