08-04-2021 | | By Robin Mitchell
Recently, VLSI (unit of a hedge fund) filed a patent infringement case against Intel claiming that Intel used their dynamic power and performance technology in Speed Shift. Why are dynamic power states used, what is Speed Shift, and why does VLSI think that the patents in question have been breached?
Since the Intel 4004, the development of CPUs has dramatically changed. The introduction of pipe-lining made better use of CPU hardware, the introduction of cache sped up operations on small areas of memory, and additional cores allowed for multiple processes to execute concurrently.
In the past, CPUs would be fixed at a specific clock rate which would result in a specific amount of power consumption. However, as the power consumption of processors increased, the need for better thermal management solutions also increased, and this introduced the use of fans on heat-sinks which can be noisy. To make matters worse, the development of mobile devices saw a need to try and reduce power consumption as much as possible to conserve battery life.
The solution to these problems is Dynamic Frequency Scaling and this technique involves changing the frequency at which a processor operates while in operation. Due to switching losses in CMOS technology, as the frequency of a CMOS-based processor increases, the power consumption also increases. Therefore, if the frequency is lowered, the total amount of power consumed decreases.
Many different CPU developers use different techniques, but the technology specifically used by Intel is called Speed Step. In this technology, the OS decides what power state the CPU should be running at, and will instruct the processor what state to enter. P states indicate what performance state the CPU should be at while running, and C states indicate what the performance state the CPU should do when idle.
The current solution for power management is to let the OS decide what the processor should be doing. However, this introduces latency as it takes time for the OS to interact with the processor. Furthermore, the OS has to keep an accurate record of all processes running and make determinations as to the current load.
To solve this issue, Intel has developed a new system that has now been introduced into their Skylake architecture called Speed Shift. In this system, the processor decides what P state it should be in without any OS interference. In essence, the processor can see what is being demanded of it, and then it can ramp up its voltage and frequency to improve performance or drop these to reduce power consumption. According to Intel, the new system not only reduces latency from 100ms to 30ms but also provides better battery performance.
In 2019, VLSI (a unit held by a hedge fund group), is suing Intel for breaching two patents held by them; US6366522B1 and US6633187B1. According to VLSI, Intel’s Speed Shift technology clearly violates the stated patents which describe processors self-monitoring their use and scaling their clocks and power consumption as needed. On March 2nd, a jury awarded VLSI with $2.18 billion in damages, and the next patent infringement cases are soon to begin.
The two patents featured by VLSI are awfully similar to Intel’s Speed Shift technology. Furthermore, it is also interesting to note that the term “Speed Shift” has not been trademarked whereas Speed Step has. However, we have recently stated in previous articles about how patents can hinder technological development, and the patents held by VLSI are somewhat obvious.
To start, both patents were filed in 2020 by SigmaTel (there’s a name you don’t hear every day), but the patent rights have since been transferred to VLSI Technology LLC. The patents were granted between two and three years after filing, and one of the patents, US6366522B1, has now expired.
What makes the case interesting is that the concept of power and frequency throttling is well known and widespread, but moving the control from an OS (software) to the CPU (hardware) suddenly causes a patent infringement. Furthermore, it should be asked if the patent holders had capitalised on the patent. As odd as it may seem, capitalising on the patent actually benefits the industry as the technology is allowed to be used. If, however, it was essentially buried, then either the technology was denied for others to use, was so obvious that no one even knew it was patented, or the technology was simply unneeded and essentially useless. Either way, the large billion-dollar award could be argued to be unfair, especially considering that one of the patents has now expired.