New mobile applications, IoT, AI or HPC, increasingly demand and what can higher performance and lower energy consumption. Advances in semiconductor technology and new microarchitectures do not seem to satisfy this demand as they should. And that is why we must look for alternatives that accelerate these cases, as heterogeneous computing does . Heterogeneous computing refers to systems that use more. Than one type of processing unit, with capabilities to handle a task. That is, general-purpose computing is with computing (specific or accelerators) to achieve high performance and power consumption, especially in data transfer and during processing. This type of heterogeneous computing, although it seems quite new. The truth is that it has been with us for some time, although there are still great software and hardware challenges to solve. But we have had it since the launch of the first STI (Sony-Toshiba-IBM) Cell systems, to the latest heterogeneous multicore that are appearing today, through GPGPUs and other SoCs for mobile phones.
What is heterogeneous computing?
Traditionally, in homogeneous computing, a general-purpose CPU has been. However, in heterogeneous computing, other units have been given a leading place, such as the GPU, DSP, DPU, NPU, VPU, FPGA, and even some ASICs . And these other units can perform certain tasks in a more energy efficient way and with better performance. For example , while a CPU is made to process a few USA Email List complex tasks at a time, the GPU is made to process many simple tasks at once. Therefore, applications such as scientific ones where a large amount. Of mathematical operations or FLOPS are, the GPU will be superior to the CPU. So why not use it for this purpose? In recent years, the definition of heterogeneous. Computing has been to also encompass other processing units basd on other architectures. For example, processors on the Arm architecture can use what we know as, that is. Two types of CPU cores, some high-performance and others efficient to better adapt to the workloads of the moment. Adopting an alternative architecture can reveal smarter ways to handle existing workloads and computing tasks.
Why is heterogeneous computing important?
The famous Moore’s Law states that the number of transistors will double every months. On the other hand there is the Dennard scale , which says that as transistors become smaller, their power density remains constant. If we combine Brazil Phone Number List these two rules, we see that this has beenĀ in recent years. In the microprocessor industry. However, in recent years something has been “broken”, Moore’s Law has in force for more than years , but lately it has become a challenge to continue maintaining it. At the same time, Dennard’s scale during – , with the arrival of multicore processors as an alternative to improve performance. As the number of cores increases , power restrictions prevent all cores from operating at their maximum performance, causing some to have to be or reduce their performance to continue maintaining consumption and heat dissipation margins. Although we are not going to go into what dark silicone is.