NVIDIA left the capital in recent years to buy two companies that are crucial for the future. Since the first, Mellanox, its technology has already been fully integrated into the Jen-Hsen Huang business portfolio. The second, which is ARM, has not yet been approved for purchase. On the other side of the ring, AMD acquired Xilinx and Intel putting all their meat on the grill with their dedicated GPUs.
In recent years, all of them have made a series of strategic movements which are moving towards a common architecture, but which for the moment is unprecedented. Which consists of combining CPUs and GPUs on the same chip or all of them with a SmartNIC in the central part.
What do we mean by processor architecture? Well, the way its internal elements are organized, communicate with each other and therefore it affects its function. And it must be said that we have several indications that the architectures of AMD and NVIDIA and Intel have several points which make us think that apart from the differences we will see a future common architecture. Let’s see which one and what differs from the current one.
The future of hardware for AMD, Intel and NVIDIA
The key to the future of hardware is the transition from SoCs to NoCs. Which we have already talked about many times, these are nothing more than SoCs or multichip modules. MCM, where the part responsible for managing the communication between the different elements is carried by an intelligent network controller or SmartNIC. Which is a processor responsible for moving data from one processor to another without them having to, which is a big paradigm shift.
The only time we saw this on home systems was on the PS3’s Cell Broadband Engine, where the different cores connected to each other using a mechanism called MFC in each of them. to communicate. The concept of the central SmartNIC, however, does not place this unit in each core, but in the center. The SmartNIC is designed to perform memory movements with a shorter delay and consuming less power. What is essential in the design of new architectures. The concept is therefore simple, in the future instead of being the processors that access the memory of the system, what they will do will be requests to the SmartNIC, which will be the one that will be responsible for making all these accesses for the processors. .
However, SmartNICs do not have to be programmed by software developers, as they will function as specific processing units to perform the task of internal and external intercommunication between the various elements that make up the SoC or the chip MCM. With the advantage of evolving much more easily. Because today, for the three companies, the construction of new processors under common architectures, but with different configurations, is a nightmare when it is necessary to design the infrastructure which interconnects the different elements.
NVIDIA DPU and Intel and AMD FPGAs
As we said before, the future of hardware is going through a series of moves made by the Big Three. From the takeover of Mellanox by NVIDIA, it aims to replace the central intercom of its GPUs to place a DPU, which is nothing more than a trade name for what Mellanox SmartNICs technology is. As for Intel and AMD, this involves the use of an FPGA, but we cannot forget the recent presentation by the blue giant of its IPU or Infrastructure Processing Unit. In the specific case of AMD, we have the case of Xilinx FPGA, which the company preconfigures to function as a SmartNIC.
While the one that is most clear about these changes is NVIDIA, their track record is clear in incorporating this technology into their GPUs. What we could see achieved for the first time in Lovelace and would be the reason why Nvidia would have succeeded in placing the large number of CUDA cores and therefore SM in its future RTX 40. Where it is said that the number of cores CUDA “” will become 18432, a number that if the number of them does not vary from the current RTX 30, we would be talking about a total of 144 CUDA cores, double that of the RTX 2080 Ti and with four times more calculation units.
The benefits of adopting this technology are clear. First, it allows us to control the quality of service when transmitting data to different client processors and allows us to create data management protocols for more advanced communication for the application. In addition to being more scalable than conventional intercom interfaces. Since the SmartNIC is responsible for managing the sending and receiving of all packages, ensuring that their transport and delivery are light and fast.
