22-09-2022 | By Robin Mitchell
Recently, Intel announced that one of its next generation processors will operate at 6GHz out-of-the-box, and therefore igniting a new arms race in CPU design. What challenges have CPUs faced over the past decade, what has Intel announced, and how could this fuel a new race?
Ever since the first microprocessors, the computing power of devices has exponentially grown year on year, and devices that used to take up an entire floor of a building now comfortably fit on a tiny silicon die the size of a grain of rice. Despite numerous mobile and network technological advances, computers have stagnated over the last decade, with operating systems taking just as long to boot as they did 10 years ago (if not longer in some cases). Additionally, long-term stability still seems to be problematic (a Windows 10 system after a year or two of operation will show a significant performance decrease).
Now, there is no doubt that CPU performances have increased over time, but the performance of CPUs can be measured in numerous ways. For example, efficiency performance can indicate how energy efficient a processor is, while multi-tasking performance shows how good a CPU is at executing multiple tasks simultaneously. In the case of modern computing, the performance of single-threaded tasks has barely increased over the last decade, and it is this lack of improvement that sees operating systems slow to boot and respond.
Fundamentally, operating systems and user applications rely on some singular process that uses sequential instructions, and it is not possible to execute future instructions before the previous instructions have been executed. Thus, it doesn’t matter how many processors a CPU has nor the amount of RAM available; a single task will run on a 10-core CPU just as fast as it would on a single-core CPU.
It is possible to split up the tasks of an operating system into different processes that can each run on their own specific core (thus, allowing for parallel execution), which can significantly help with improved performance. However, many operating system services rely on the kernel for operation, making it difficult to separate these additional services from the main operating system thread.
Over the past decade, increasing core frequency would be the simplest solution to improving single-thread performance, but considering that going above 5GHz results in massive heat dissipation (not available to the average consumer), processors have had to rely on pipelining and superscalar designs to try and improve performance.
Recently, Intel announced that it will release a new CPU, the Raptor Lake, that will provide users with a CPU that can natively operate at 6GHz. While pricing and availability have not been made public, it is expected that the new processor may be a KS-branded version of the upcoming 13900K chip. At the same time, the announcement from Intel hits hard at AMD, who recently announced that their next-generation CPU will operate at 5.7GHz.
While a 300MHz difference may seem marginal, it can provide massive improvements in single-threaded performance. Additionally, Intel announced that the 6GHz speed would be available on all cores simultaneously and not just for individual cores running under a performance mode. According to Intel, the new processor speeds will see a single-threaded performance increase of 15%, while multi-threaded performance will increase by 41%.
However, Intel further announced that the new architecture will support overclocking to 8GHz, double the speed of typical computer cores. Considering that such speeds have only been historically achievable with the use of liquid nitrogen and other exotic cooling methods, this announcement indicates that a new CPU race may be on the horizon.
For the longest time, computer manufacturers used MHz as the defining characteristic of their machines. One computer operating at 8MHz would be considered inferior to those operating at 10MHz, which existed well into the early 2000s. However, it became clear that frequency had little effect on multicore designs, so the battle shifted from GHz to core count, FLOPS, and memory bandwidth.
However, the announcement from AMD and Intel on their next generation of CPUs could ignite a new race for faster single-threaded machines. While a 15% increase may not seem major, it is undoubtedly a step in the right direction for improving the performance of applications through increased responsiveness and faster booting times. Furthermore, the increased performance across multicore designs will also benefit multi-threaded applications as each thread will see accelerated execution.
But while faster CPUs may help to improve general computing performance, it is certainly not the best option for numerous modern applications, especially in the mobile field. By far, the most essential feature in low-end computational tasks and those dependent on portable power supplies is energy efficiency; a mobile device that runs out of power in 20 minutes because of an 8GHz core CPU is not a practical device. Since the vast majority of tech users are now switching to tablets and mobiles using cloud-based applications, high-performance CPUs may be more of a niche product.
Overall, the increase in CPU speeds from Intel and AMD is certainly welcoming as having many cores can only get you so far, but while it may be great for tech-savvy individuals and supercomputers, it may not be the best option for large datacenters or mobile users.