The Zen 2 architecture in 7 nanometers jumps over the efficiency of Intel products, and that has already brought AMD its first major project against Intel.
This week, the annual Supercomputing Conference will be held in Dallas, where experts and large companies will take the opportunity to showcase their progress in this field. During the event, the new Top500 list was released, listing the 500 fastest supercomputers in the world. There are no AMD processors among them, but that will change soon.
AMD presence at annual Supercomputing Conference
As we saw a few days ago, AMD has made good use of the time in which Intel has been stuck in 14-nanometer designs. Its Zen 2 architecture in 7 nanometers has finally turned the cake around, and now, yes, it’s achieving higher performance per watt. This is perhaps not so crucial for users who build a home PC, but of course in systems that move at maximum power in tens of thousands of kilowatts.
Hawk, the first major project with Zen 2
Rome is the first chip to bring Zen 2 to 7 nanometers. Although TSMC will start manufacturing in early 2019, the company already has the first 10,000 units. It is the High-Performance Computing Centre (HLRS) of the University of Stuttgart that is interested and has signed a contract for the manufacture of Hawk, which will be one of the three most powerful supercomputers in Europe and one of the fifteen most powerful in the world.
Hawk will multiply the power of the current data center by 3.5 as compared with the Intel’s Xeon chip line. This is by no means an insignificant leap for the current thirtieth (30th) supercomputer in the world in this ranking by performance. This new machine will have over 5,000 nodes, each with two Rom processors. Each of these processors has 64 physical cores and 128 logical processing threads. These will operate at a frequency of 2.35 GHz, which is not very high for consumer processors but is not bad at all for this type of system.
AMD Hawk offers billions of operations per second
As a result, they will add up to 640,000 physical cores in a system with no less than 665 TB of RAM and a further 26,000 TB of memory. The maximum theoretical yield expected from this supercomputer is 24.06 TFlops or trillions of floating point operations per second. The assembly of Hawk will be carried out by Hewlett Packard in a project costing an estimated 38 million euros.
According to HLRS itself, it will be the world’s fastest supercomputer for industrial production with applications in computer technology and scientific and industrial research. The falcon is expected to be useful in researching applications in the fields of energy, climate, mobility and health.
AMD opens door to the supercomputing technology
This is the first major project that brings AMD to a starting point in the field of supercomputing, and from the brand data, it does not look like it will be the only one. Since the new solutions are now more efficient than Intel’s, more data centers are likely to make the leap once and for all.