Nvidia registers the Hopper brand. The after Ampere getting closer?

Nvidia confirmed, albeit indirectly, that he is working on a project called Hopper. The company has filed a trademark registration application with the US Patent and Trademark Office. The name Hopper had emerged in mid-November and allegedly conceals rumors a future computing architecture.

According to the Twitter user “Kopite7kimi”, leaker of the rather well-known hardware sector, among the many new features Hopper would also introduce an unprecedented MCM-type project (multi-chip module), where there are multiple dies on a package to offer overall larger and more powerful GPUs.

Another advantage would relate to yields, because it is easier to make problem-free smaller chips than large monolithic processors. Think, for example, of what AMD does with the Ryzen 3000.

As we have seen in the past, Nvidia has been experimenting in the field of multi-chip modules for some time. The project we know, called RC 18, however, looks to the world of artificial intelligence.

AMD is experimenting in this direction and Intel, which is preparing to land in 2020 in the world of GPUs, has several technologies to carry out completely similar projects.

The future is therefore traced, but to say when we will see the first MCM solution is difficult. As for Nvidia, it is possible that the company initially adopts an MCM project in the Tesla accelerator range, and then expands it to other areas over time, as needed.

As for “Hopper”, the name would be congruous with what has been seen in recent years: Nvidia's computational architectures are called with the name of eminent scientists, from Fermi to Kepler, to move on to Maxwell and Turing, just to give some examples .

Hopper would call Grace Murray Hopper, an American mathematician, computer science and military believed by many to be a pioneer of computer programming and famous for his work on the Navy's first digital computer, the Harvard Mark I, but also for the COBOL language.