ARM (Finally) Adopts the 64-Bit Architecture
When ARM announced Cortex A15 last year, they said Cortex A15 is a high-performance chip that can be used not only in smartphones and tablets, but also in notebooks, PC’s and even servers. It was very clear from the beginning that they intended to make ARM popular in the server market with this powerful chip.
But one thing that was surprising and also a little disappointing was that they didn’t make this chip 64-bit. This means that any device can’t use more than 4 GB’s of RAM if the architecture only supports 32-bit. They did, however, add an extension for Cortex A15, that makes it 40-bit and can support up to 1 TB of RAM. But this should be mainly intended for servers.
I wouldn’t expect to see more than 4 GB of RAM tied to Cortex A15 on smartphones and tablets, because all the apps are written for 32-bit, and also there’s no point in doing that by 2014. In 2012 we’ll start seeing 1.5 GB and even 2 GB smartphones and tablets. Then 3 GB and 4 GB devices (think Transformer) in 2013 and early 2014.
And by the end of 2014, we should start seeing the next-gen chips that come after Cortex A15 and are based on the new ARMv8 architecture that is 64-bit.That’s when we’ll also start seeing 64-bit applications (think Photoshop, which is already arriving on Android this year), but the ARMv8 architecture will still have have support for 32-bit apps, and so will Android.
“ARMv8 will enable the development of ARM architecture compatible devices that can be designed to maximize the benefits across both 32-bit and 64-bit application areas,” said ARM CTO Mike Muller.
“This will bring the advantages of energy-efficient 64-bit computing to new applications such as high-end servers and computing, as well as offering backwards compatibility and migration for existing software through a consistent architecture.”
Nvidia has already been working on entering the server market with their own ARM-based Project Denver, that should arrive around 2013. Initially, I thought Nvidia was going to adopt Cortex A15 and therefore the Project Denver will be something like a 16 or 32 core with 2.5 Ghz per core, but now I’m not so sure anymore.
Nvidia might try to design their own ARMv7 chip to follow their Cortex A9-based chips, which would be similar to what Qualcomm is doing with Krait (custom design). That custom chip would come to servers, but stripped down versions would also arrive in smartphones and tablets. My guess is this is why Nvidia hasn’t said a word about their plans with Cortex A15 yet.
Now, Nvidia is announcing they are are going to support the ARMv8 architecture as well. It’s not surprising considering the ARMv8 architecture could support up to 128 cores per chip, which would help them a lot in the server market, but they also seem excited about it being in a whole range of computing devices:
“The combination of Nvidia’s leadership in energy-efficient, high-performance processing and the new ARMv8 architecture will enable game-shifting breakthroughs in devices across the full range of computing – from smartphones through to supercomputers.”
This was the announcement for the ARMv8 architecture, but ARM will also announce the first chip design based on it sometime next year, and it should arrive in devices in 2014 built at 20 nm. It will most likely be meant to replace Cortex A15, but we should hear about another one in 2013 that is meant to replace Cortex A7 in 2015.
ARM designs these chips for 2 years, so the timeline is not too hard to figure out. They switch the processing node every 2 years, too, and after 28 nm we should have 20 nm in 2014, and 14nm in 2016. Things will be getting very exciting for Android by then, and it should push it into new types of devices.
But before we go too far into the future, the chips we ought to be excited about right now are the ones that combine Cortex A7 and Cortex A15 in a big.Little configuration.