AMD has been working on server parts based on 64-bit ARM cores for a while now. The goal is to offer a small server SoC with 4 to 8 Cortex A57 cores in 28nm and the first chip should roll out next year. It is called Hierofalcon and we mentioned it last month.
Hierofalcon is designed with networking and communications in mind, hence it features integrated 10Gb network support. Now there is talk of a consumer chip that should target the tablet market. SweClockers reports AMD will indeed introduce a tablet chip, wither with Cortex A57 or Cortex A53 cores (or both). The chip is also said to feature GCN graphics.
Although SweClockers is a reputable site with a pretty good track record, we’re not sure about this.
Right off the bat one would assume that the consumer chip is a Hierofalcon derivative, but that isn’t the case. Hierofalcon is designed with an entirely different market in mind, it has a few bits that consumer chips don’t need and its TDP is 15W to 30W. AMD’s Seattle parts will come with 8 to 16 ARM cores, they are not consumerish by any means.
In other words, it would be easier to design a consumer chip from the ground up. The other question is – why should it? There are already plenty of highly competitive ARM SoCs out there and companies like Qualcomm, Samsung and Nvidia are throwing tons of money at new designs as we speak. Just because it can be done doesn’t mean that it has to be done.
In addition, a hypothetical AMD ARM consumer part would practically have to compete with something from AMD’s own stable, Jaguar-based chips and future x86 small cores. There is a caveat though. AMD might be interested in the consumer ARM space as a way of bringing GCN to an entirely new market, but it could do that without its own chips.
There’s already been a lot of AMD - ARM speculation. Many people expected AMD to enter the ARM space after Nvidia announced Tegra. Now many observers believe AMD will eventually start licensing GPU IP to ARM outfits, and that appears to be Nvidia’s long-term plan as well. With Mantle in the mix, sticking GCN into mobile parts, consoles and PCs starts to make a lot more sense. However, this doesn’t mean AMD will try to directly compete with the likes of Qualcomm in the consumer chip space.
If it is indeed working on a consumer ARM chip, it might be a technology demonstrator rather than a proper product that will end up with actual design wins. Investors would probably love it though – and it wouldn’t take much effort to create plenty of positive buzz. AMD spinners would have a field day - GCN and Mantle in tablets, smartphones, handheld consoles, PCs and big consoles.