In an announcement today, Arm unveiled its roadmap of CPUs until 2020. Arm's new “laptop-class” chips aim to beat Intel’s Core i5-U CPUs.
Arm and Arm chip makers such as Qualcomm have been taking aim at the laptop market for the past few years. This is why recently we’ve been seeing not just Chromebooks with Arm chips, but also Windows 10-based laptops.
Starting this year and until 2020, we’re going to see three laptop-class performance CPU cores from Arm . The first, which Arm already announced previously and should come out this year on a 7nm process, is Cortex-A76. The Cortex-A76 promises a 35% performance uplift compared to the previous generation, the Cortex-A75, along with four times the machine learning inference performance.
For 2019, Arm has planned the “Deimos” core, while in 2020 we should see the “Hercules” core on both 7nm and 5nm processes. All three chips utilize the newer DynamiQ technology that will replace big.LITTLE.
Beating Intel’s Core i5
Arm is not shy about stating its ultimate goal with these upcoming chips, and that’s to beat the Intel Core i5 U-line, while using significantly less power too. According to Arm's own tests, Cortex-A76 is already expected to be on par with Intel’s Core i5-7300U CPU, using less than 5W of power, compared to 15W for the 7300U.
Arm also says that these chips’ performance and efficiency have been progressing at a faster rate than the industry average. At Cortex-A76’s launch, Arm claimed an average of 20% performance gains year over year for the past few years. Now the firm is also claiming that the chips coming after Cortex-A76 will see an approximately 15% performance gain year over year.
Arm also showed how its 2020 Hercules CPU should have 2.5 times the compute performance of the 2016 Cortex A73. Much of that performance gain should be from the transition from the 16nm process to the 5nm process, but Arm still promises much larger compute performance improvements compared to the rest of the industry.
Targeting the Laptop Market
Although Arm's chips may have potential to reach parity with Intel’s mainstream laptop CPUs, that’s not the only hurdle Arm will face in the laptop market. Arm and its partner chip makers will still have to convince consumers that Arm -powered Windows 10 or Chromebook laptops will be just as good software-wise as the Intel or AMD counterparts.
Arm believes that always-on laptops with 5G connectivity will be one of the things that will set the Arm -based laptops apart from the competition. The company seems to believe that both because Arm -based laptops will have significantly better battery life that can last for days and because Qualcomm will likely be the leading provider of 5G modems, giving Arm chip makers an edge in the always-on laptop market.
Even if their chips are more modern and more efficient, if they attempt to compete with a u series i5 on Windows, they're going to have to use 10 to 15W ARM parts to compete with 15W Intel chips to overcome the x86 emulation, so they really need Microsoft to do a much better job supporting apps designed and optimized for x86 on ARM platforms. But really I think that means they need all app developers to release builds for Windows on ARM and I don't see that happening until there is a huge demand for it.
Now assuming they do get that support, there is the pricing issue. The high performance ARM platforms released and announced so far that can run Windows 10 are flagship SOCs with flagship pricing. But they have to compete with Intel's main stream mid range chips that cost half what these flagship ARM SOCs cost. Actually, right now Snapdragon 835 is almost competitive with quad core Braswell (right up until x86 emulation comes into play), but a Braswell SOC costs a third what a Snapdragon 835 costs.
I think the only way they can penetrate the Windows laptop market is if they can both improve performance enough to be competitive (which right now they are no where close to) and cut the cost substantially to give people some incentive to be an early adopter and try out an ARM laptop even though the ecosystem of ARM optimized applications doesn't exist yet. They have to do both because no one wants a slow, crappy laptop at any price, and most people won't pay a premium for an acceptably quick laptop that can't run the software they need.
Now if they could just get away from Windows and take the x86 emulation out of the picture, they'd have a pretty good product. These would probably make great Chromebooks or Linux laptops. But again, pricing would be critical.
It'll be nice if these can approach the performance of the x86--and Windows and Linux support them without end-user issues like the limitations of ARM Win10.
But when did people decide the perfectly serviceable "increase" needed to be replaced with the goofy-sounding "uplift"? It's like that time in the 90s when TV Guide started calling the broadcast TV networks "webs".
I don't miss the days of "make sure you download the right executable for your CPU architecture, or build from source yourself", personally.
The same things that would make them good Chromebooks or Linux boxes--not running superfluous x86 emulation--would probably make them cromulent under Windows, too (albeit at the cost mentioned in my first paragraph.)
If running x86 code is what matters to you, why would you run Windows 10 on ARM? Someone who buys a laptop that is powered by a 5W CPU will only be interested in non-CPU-intensive applications, which will surely have native ARM versions.
Probably slower since it will have to emulate x86 instead of natively run it. It will be similar to Intels Itanium uArch in that respect.
I do like how they tout large performance gains. Of course when performance is so much lower than the competition the performance gains will be better.
I only wonder if ARMs efficiency will still exist as their CPU becomes more complex or if it will do like all things and probably match up to Intel/AMD.
Seems legit brah.