Apple has patented a new method of using memory chips in its System-on-Chip (SoC) for computers. Recall that Apple abandoned the use of Intel processors in its Mac computers and decided to develop its own Apple Silicon processors, the first of which was the M1 chip.
Apple’s new processor architecture combines an ARM-based compute processor with a proprietary GPU. Both require fast access to system memory to work effectively. Apple has found a way to solve this problem. It gave the CPU and GPU the ability to access the same pool of physical memory.
The disadvantage of the so-called Unified Memory Access (UMA), which is now used in M1 chips, is that both processors have to share the total memory bus bandwidth, as well as its physical capabilities. Thus, in some scenarios, one of the processors may feel the lack of one or the other. Apple has patented a method of using memory that can solve this problem by combining high-bandwidth DRAM cache with high-capacity DRAM main memory.
“Both types of DRAM memory form a memory system. One type of memory is optimized for high bandwidth, and the other is optimized for physical volume. The use of both types of memory in one platform allows, on the one hand, to increase the energy efficiency of such solutions, and on the other hand, to take advantage of both high-speed memory and memory with a large physical volume “
The patent was filed back in 2016, however, as indicated by the resource