Is ARM Changing Its Licensing Model? Apparently, And According To Qualcomm, It’s Not Good
03-11-2022 | By Robin Mitchell
While nothing has been officially confirmed, Qualcomm has accused ARM of unfair practices in its desire to change licensing terms that would have a major effect on OEMs. What exactly has Qualcomm accused ARM of, how would this affect OEM manufacturers, and is this evidence for switching over to RISC-V?
What exactly has Qualcomm accused ARM of?
It’s common knowledge that ARM is currently suing Qualcomm and its Nuvia subsidiary for a supposed breach of license agreements after Qualcomm absorbed Nuvia’s designs to create server products that supposedly bypass ARM licensing. However, Qualcomm recently filed a counterclaim whereby it accuses ARM of unfair trading practices with a supposed desire to change its licensing agreements that affects OEMs and consumers of ARM technologies.
Specifically, what Qualcomm has accused ARM of planning to do is to make changes to licensing agreements that prohibit companies from customising some aspects of the ARM chipsets and instead force customers to either use or obtain licenses for other ARM proprietary IPs. Simply put, these changes to the license agreement are expected to make it impossible for customers to use the ARM core but create their own modified chipsets that support advanced functions such as GPUs and TPUs. For example, Qualcomm utilises ARM cores but has developed its own customised chipset for graphics (the Adreno GPU).
In response to these claims, ARM has stated that the counter-filing is filled with inaccuracies but has failed to officially respond to the allegations, which could suggest that Qualcomm’s claim is based on truth. Furthermore, it should be noted that Qualcomm has only heard of interactions between its OEM partners and ARM, whereby ARM has supposedly threatened OEMs about changing licenses.
How would such a change in license agreement hit OEMs and the engineering community?
It is hard to say if the allegations by Qualcomm have any basis in truth, but if they do, then it could spell trouble for those using ARM cores. As stated earlier, preventing OEMs from developing their own chipsets in favour of ARM IP would seriously disrupt the ability of engineers to craft unique solutions (such as hardware accelerators). For example, Google has developed its own custom chipsets for Tensor processors, resulting in a dramatic increase in performance. Preventing Google from using integrated TPUs would force Google to look at alternative measures that would unlikely be able to provide the same level of performance.
One potential solution is that customers have to utilise external modules and chips that sit outside of an ARM processor, but while this would allow for custom hardware to be used, it would see significant performance impacts. Furthermore, Qualcomm also alleged that it would no longer provide technology license agreements to any semiconductor company after 2024 and instead will be changing its business model to provide licenses to device-maker specifically. If true, this would allow manufacturers of devices to use ARM products and obtain licenses, but semiconductor companies (such as Qualcomm, Ti, and Xilinx), would not be allowed to use ARM CPUs. As such, semiconductors may only be able to get an Architecture License Agreement (ALA) which requires the semiconductor company to design their own processor from scratch (and it is improbable that this will be allowed to be resold to third parties).
Is this a sign to move towards RISC-V?
If there is one thing I am personally getting extremely annoyed at, it’s the constant bickering of companies such as IBM, Qualcomm, Google, ARM, and Apple, who are always suing each other while trying to create entirely unfair license agreements. But thankfully, the wheel of progress continues to turn, and open-source solutions are becoming increasingly more of a reality.
One processor that doesn’t have to worry about licensing agreements, chipsets, or architecture is RISC-V. Anyone can make it, use it, customise it, and no one can say a darn thing about it. So long as an engineer makes their RISC-V platform conform to the basic RISC-V instruction set, then it can be labelled as a RISC-V compatible device, and better, they can integrate any hardware inside or outside of the chip they want.
After reading the quarrels of Qualcomm (should be Quarrelcomm), it would seem that RISC-V has positioned itself perfectly in the engineering community. If ARM continues to poke the hornet’s nest, major semiconductor designers could move away from ARM entirely. What economic benefit is there from using ARM cores if RISC-V can be made to have equal performance? Why use an ARM core when a small fabless company down the road offers RISC-V cores with fair licensing agreements?