ARM has today announced their latest chip for smartphones, tablets and other embedded products, the ARM Cortex-A17. This chip is ARM’s latest attempt to go after the middle of the market and to build a chip that their licensees will want to buy from them as physical IP rather than an architectural license (see all of the ARM-v7 licensees). ARM wants to build cores that are attractive to their customers and cores that are likely to be considered good enough to implement in products cheaply and effectively. What’s interesting is that the A17 is not actually supposed to replace the Cortex-A15 chip, ARM’s most powerful chip. The A17 is actually supposed to be a replacement for the much-used ARM Cortex-A9 chip which has quite a long run in terms of mobile design cycles. Do keep in mind, however, that ARM has already released a chip that was supposed to be the successor to the A9, the A12, but that chip has seen little to no companies actually adopting it in their chips/products.
As you can tell, ARM sees a huge opportunity for the $150-$350 price segment for mobile device shipments, which will likely make up a huge chunk of the company’s revenue by 2014. Especially when you consider that most high-end chips are custom designed architecture licenses, while the mid-range and low-end chips are mostly physical IP licenses. A good example is Qualcomm’s first 64-bit SoC, the Snapdragon 410, which is actually based on ARM’s A53 SoC.
The A17 is also specifically designed to make the best use of the fact that there is so much 28nm capacity in the world’s fabs right now and 28nm, according to them, provides the most transistors per dollar. Based on the current state of 28nm, they are probably very right about this fact as most 28nm processes have already been figured out and yields improved. However, do keep in mind that this would be a 28nm low power process, which I believe is one of the most mature and prevalent processes in the various fabs around the world (TSMC, Globalfoundries, UMC, etc.)
They also prided themselves in delivering a much better overall product to OEMs and consumers by creating a nice suite of technologies that easily complement each other. So, in addition to the Cortex-A17 chip, ARM is also recommending a Mali T720 GPU, V500 Video engine and DP5500 display processor. They also have ARM’s POP IP for CPU and GPU at 28nm.
But if you look at the fundamental designs themselves, ARM is looking to significantly improve the performance of the A17 over the A9 and do it in a much lower power envelope as well.
The Cortex-17 is slated to deliver 60% performance over the Cortex-A9 AND reduce the power consumption by 20%. The claim that these performance improvements and power consumption reductions are measured using the SpecInt benchmark. So, not only are you getting a huge improvement in performance, but you’re also saving a significant amount of power, which translates to better products and designs. The A17 also allows for a big.LITTLE-type architecture which would see the A17 chips as the big chips and the ARM Cortex-A7 as the little chip. And since these chips are still 32-bit, there isn’t much that hasn’t already been done to optimize for these new chips. ARM’s goal is to make the A17 a frictionless addition to their lineup and to make it attractive for companies to implement in 2015, to further elevate the level of mid-range smartphones and tablets.
The truth is that the Cortex-A17 is poised to become the new standard in mid-range smartphone processors and it will probably become a core portion of many SoCs in that price range for 2015. What really remains to be seen is whether or not ARM can avoid the same situation that they had with the A12 where virtually nobody licensed the chip and it was effectively rendered a dud. 2015 is already looking like it will be an interesting year, especially if we can really elevate the quality of smartphone processors by such large percentages. Obviously, the ability to evaluate ARM’s claims will take some time to do, but we’ll do our best when phones finally do start to come out.