AMD recently unveiled something truly remarkable today – a server rack that has the total processing power of 1 PetaFLOPs. That’ 10 to the power of 15 floating point operation per second or 20 to the power of 15 half precision FLOPs.
Here is kicker though: a decade ago in 2007 a computer of same power would have required roughly 6000 square feet of area and thousands of processors to power.
A decade ago, this would have been one of most powerful supercomputers on Earth, and today it is a server rack.
AMD’s Project 47 – 1 PetaFLOPs of Single Precision Compute at 30 GFLOPs Per Watt
The Project 47 is powered by 20x EPYC 7601 32 Core processors and around 80x Radeon Instinct GPUs. It supports 10 TB of Samsung Memory and 20x Mellanox 100G cards as well as 1 switch.
All of this is fitted into server rack that is roughly the height of 1.25 Lisa Su’s with an energy efficiency of 30 GFLOPs per watt.
The Project 47 super computer consumes around 33,333 watts of electricity. Project 47 will be available from Inventec and their principal distributor AMAX sometime in Q4 of following year.
Today at Capsaicin SIGGRAPH, AMD showcased what can be achieved when the world’s greatest server CPU is combined with the world’s greatest GPU, based on AMD’s revolutionary “Vega” architecture. Developed by AMD in collaboration with Inventec, Project 47 is based on Inventec’s P-series massively parallel computing platform and is a rack designed to excel in a range of tasks, from graphics virtualization to machine intelligence.
Back in 2007, there was a supercomputer called IBM Roadrunner. This was super computer project that was once most powerful, well top rated supercomputer of its time built by AMD and IBM for Los Alamos National Laboratory. The cluster had 696 racks spanning an area of 6000 square feet and consume 2,350,000 watts of electricity. The cluster consisted of 64,000 dual core Opteron CPUs and some accelerators.
After 10 years, AMD has managed to build a system that is 98% more energy efficient and 99.3% less space consumer. We are not yet sure how much this project will cost but we are pretty sure it would be way less than $100 million cost of original Roundrunner.
That exactly does AMD manage this feat? Well, the real power of AMD’s Project 47 is not CPU it is GPU. While AMD has progressed on CPU front has not been anywhere near as large to justify by gains seen here.
In fact, with 20 EPYC 7601 CPUs, you are looking at core count of 640 cores which simply is less than 128,000 core in original Roundrunner. Since we certainly did not see IPC increase of 20000% it is clear that the star of the Project 47 is the Radeon Instinct GPU.
Specifically, there are 20 2U P-47 systems, and each houses a single EPYC 7601 CPU and four Radeon Instinct MI25 GPU accelerators. Each machine also has a half-terabyte of memory, and a Mellanox 100Gb InfiniBand router is built into the rack.
With 80 Radeon Instincts inside the server rack, you can already account for roughly 960 TFLOPs (depending on the clock speed) already out of the 1000 TFLOPs that the P47 is rated at. With 128 PCIe lanes per CPU, the EPYC processors will act as the drivers of the Radeon Instinct and won’t actually handle the brunt of the load. So basically from an all-CPU based Roadrunner, we have come to P47, which is practically an all-GPU based show. It really speaks volumes for the bonkers growth in power we seen in the GPU department. The rapid scaling of core count, architectural gains, and node shrinks have really ushered in a new era of computational power.