2020 will be marked by the advent of processors, especially in the field of servers and in the broader sense of computing power. ARM To exit Intel’s pedestal.
Continuation of the article below
A year ago, ARM processors were for mobile devices and Intel processors were for servers. There is not much credible reason to change it: no matter how hard the ARM environment tries to improve the design and functionality of its chips, they are relying on cores and gions in an unreliable way. ‘Intel. Intel, moreover, was more concerned with recalling AMD, which, along with its Raison processors on PCs and Epic on servers, demonstrated such excellent performance for very little money, which would discourage users from using Zion. This is what happened even though the transformation into companies was very slow, as it was tied to a period of contracts spanning three to five years.
But there. In a very short time, the world of supercomputers, but also the cloud giants, even the European Union, announced that their most powerful designs are now based on ARM processors. The coupe de Grace was introduced by Apple at the end of 2020 with the introduction of the new generation Max without Intel processors.
The reasons for changing the situation are linked to the industrial skills of various players, which is standard in the world of processors. Gordon Moore, co-founder of Intel, preached with his celebrities Named law, Processor founders have every interest in doubling the power of their chips every two years. To increase the power of a processor, it is necessary to increase the number of its transistors, which means that they have to be engraved very neatly each time, so that they fit on a square silicon, which means it is more expensive to invest in factories. It is clear that we need to sell more and more processors to raise enough funds to invest in the next generation plant.
Basically, ARM processors were much less efficient than Intel processors: their silicon square was smaller, with very few transistors, to use as little power as possible to maximize the autonomy of mobile devices. With many times more mobile devices being sold than PCs or servers, the ARM ecosystem is raising funds much faster than Intel to build the next generation. Intel once made the choice to own its own factories; He can only trust himself to earn the return to invest in it. Manufacturers of ARM-licensed processors, including Apple, Samsung, Qualcomm and others, are all third-party players, primarily Taiwanese DSMC.
In 2020, Intel finally, at the expense of considerable effort, brought in factories producing chips with 10 nm line carving elegance. Again, its Zion processors for servers are still manufactured in 14nm generation chains. In the ARM world, it takes us two years to switch to 7 nm. With its new Mac, Apple is already launching DSMC’s new 5nm channels.
Taking advantage of such a dynamic DSMC, AMD decided to make its x86 processors at home. Just like Nvidia does for its GPUs now.
You have it: By 2020, ARM processors will have more transistors than Intel, despite their lower consumption, and therefore compete for power by spending less power on data centers.
Among the cloud giants, AWS has already developed its own ARM processor to power its compute servers. By the end of this year, we know what Microsoft will do The same, For its Azure cloud servers, but like Apple, for the new generation client devices from its surface brand. In the same phenomenon in supercomputers, the EU saw the rise of ARM processors as the most favorable as it sought solutions to free itself from dependence on American suppliers, controlling its production itself. .
The rise of Nvidia
Among the players in computing power, there is one who has embraced the ARM adventure in particular: Nvidia. Nvidia, popular for its GPU-based graphics cards in the PC and workstation world, initially realized that its development would be found in driver data centers, where its components make it possible to speed up much-needed applications. He still had to find a place between Intel’s ten years of dominance and the arrival of AMD.
Nvidia was one of the first to promote the use of ARM processors on the most powerful computers late last year. Truth be told, at the time this event passed like an announcement effect, almost an event. Because, at the same time, Nvidia was creating a new life in data centers with the most sophisticated strategies. To strengthen its presence on the servers, we saw the acquisition of Melanox, a supplier of control cards for the network and storage. We also saw the introduction of the A100, its most powerful GPU built-in to calculate industrial simulations, stimulate decision-making applications, or support artificial intelligence algorithms.
On the other hand, last September, the surprise was total when he announced his acquisition of ARM.
Nvidia abruptly designed this new competitor party to eliminate x86 processors, namely Intel, but not AMD. Both responded. First, you need to become an AI expert by purchasing a Chicop. To learn how to make accelerator chips, second by purchasing Jillinks. We can bet that this war has just begun.