Last week ARM announced a new process for its Artisan POP, UMC’s 28HPC. This isn’t a major change in the way of the force but it does open up some avenues for fast but cheap silicon.
The new process is called 28HPC, an update to UMC’s current 28nm processes with higher performance. UMC lists it as a gate last process, something you tend not to find much of at other places, likely for a reason. That said it is out there, working, and now you can get ARM POP products on it. The main design listed in the release the Cortex-A53 core but that is just the highlight, the announcement likely heralds tons of other ‘lesser’ products in the portfolio as well.
If you are wondering why anyone would care about 28nm products, especially 28nm products from UMC, you probably don’t understand the market. Enthusiasts consider the 28nm process ‘yestertech’ but it is more than adequate for most users in most use cases. It may not be as energy efficient or as fast as 14/16nm processes but it blows them away in one deadly important category, cost.
Looking at the cell phone market, the number of high-end/halo devices being sold with silicon from the bleeding edge node is rapidly declining as a percentage of the market. The sales volume leaders all use older but still quite adequate SoCs for their devices, and most users don’t know or care about the difference. Their devices just work, do what they want, are better than their last one, are wholly adequate for their needs, and cost 1/3 or so what the Galaxy iToys do.
This is a market of billions and 28nm SoCs are just fine there, and will be for a while. Until you can make a <$8 quad-core 64-bit SoC on 14/16nm, 28nm will probably rule the market mainly on cost and adequacy grounds. So ARM announcing a new variant on a 28nm geometry node is important in a serious way, it probably shaves a few tens of cents off each SoC made on it. That is deadly important for servicing a mere few billion customers.S|A
Latest posts by Charlie Demerjian (see all)
- HyperX ships it’s 60 millionth enthusiast memory module - Oct 15, 2018
- Bittware/Nallatech water cools 300W of Xilinx FPGA - Oct 12, 2018
- More on Intel’s 10nm process problems - Sep 17, 2018
- Intel puts out another 14nm 2020 server platform - Sep 11, 2018
- Why Can’t Intel Supply Enough 14nm Xeons? - Sep 10, 2018