Nvidia Hopper next architecture after Ampere

Every major launch from AMD, Intel and Nvidia is in most cases caused by all sorts of leaks. The most common is that reports on new products or even entire series begin with a code name, where for example the Nvidia Geforce RTX 2000 series is called “Turing” and the AMD Radeon RX 5000 family is “Navi”.

Next in line from Nvidia is “Ampere” which is expected to be introduced in 2020 and take the step to 7 nanometers. It now appears that the architecture is succeeded by “Hopper”. It writes the Twitter user @ kopite7kimi, which on several previous occasions revealed detailed information about upcoming news from Nvidia. The post was removed shortly after publication, but was intercepted by German 3Dcenter.

Interesting beyond the name itself is the information that Hopper means a radical design change. The architecture is said to take the step to the Multi-Chip Module (MCM). This means that several separate, usually smaller, circuits are manufactured and then joined together on a substrate to create a more potent solution. It follows the same trend as AMD with the Ryzen 3000 series, where several chiplet is used to create processors with more than eight cores.

This would allow Nvidia to create more high-performance graphics chips than in the conventional way. As Moore’s team has slowed down, Nvidia has found it increasingly difficult to raise the performance bar and the company’s most high-performance solutions have been around 700-800 mm² in size, which is close to the limit of what is possible to manufacture with today’s lithographic processes (reticle limit) .

Grace Hopper, who is said to give its name to architecture, was an American computer scientist who, among other things, created the first compiler. Together with IBM, she assisted in the completion of the electromagnetic computer Harvard Mark I in 1944, which was used by the US military for ballistic calculations. One of the computer’s very first programs to run figured out the implosion concept for the Manhattan project.

Read more about Nvidia’s upcoming architectures: