It looks like Microsoft is going to kick Wintel to the curb in a few weeks. SemiAccurate has gotten wind of a new initiative to revive WART once again.
You might recall WART aka Windows on ARM RT, the doomed attempt to try and stop Android from taking over tablets. Windows 8 was bad enough but locking it down to managed code from the Windows store, of which there was none at the, only added to the hilarity. The scheme was so well thought out and the tools were so good even Microsoft’s own Office RT broke all the rules just to run intolerably slowly. Others weren’t allowed to transgress in the same fashion so WART died a public and well-earned death.
Now with Windows 10 every is new and different. Except that you still have a unusably awful UI, it mostly still requires software from the Microsoft store of which there still is none, and it still runs on ARM. According to documents seen by SemiAccurate this time they can’t fail because, err, it didn’t really go into detail on that front other than to say it was going to be really good. One thing they did say is that a focus this time is cellular data with some variants including bundled data in the purchase price.
One other low-level feature is different, Microsoft is providing x86 emulation for WART, remember the Qualcomm announcement from a few months back? Better yet over the past 5-6 years, ARM CPUs have gotten vastly faster and more capable. We have gone from single or possibly dual 32-bit CPUs in the 1.x GHz range to 8x 64-bit cores running at well over 2GHz. Just looking at the memory bus speed and addressable space, it is a sea change.
Here is the part that SemiAccurate points out that emulating a more complex ISA from a vastly faster and wider CPU on a slower core and simpler ISA isn’t a good strategy for a tolerable software user experience. Luckily for the slides being shown off to OEMs, this isn’t a problem for Microsoft because, umm, did we mention it runs Windows 10? And. Umm. Hey look outside, isn’t the sky pretty today?
Back to reality Microsoft is said to be launching their ARM based Cloudbooks this week and a ton of WART devices at Computex. Unlike the last time the company tried this debacle, and the time before that, and the one before that, there is a real difference this time. Really, no joke. The difference is that Microsoft is opening up WART to PC form factors, think actual laptops, clamshells, and PC-like devices instead of the toys of the last few times. Microsoft is really opening up the full PC experience to ARM SoCs, things aren’t restricted to Intel anymore. Qualcomm is the only SoC vendor mentioned directly but SemiAccurate’s sources strongly suggest that they are the first, not the only one.
If you are wondering why SemiAccurate is saying this dead turkey is going to shatter WIntel, it isn’t because it will succeed. These new WARTbooks are doomed for the same reason the last few iterations were, mainly that they stink on ice. They will be slower than a PC, less flexible, far more locked down, but will run the software most people expect them to unlike the last few. But they will run it like a slow, sick slug. It is going to be a disaster, no doubt about it, and the meager price difference between an ARM cored slug and a crippled Intel cored slug won’t make the down sides worth it. Expect it to be the big winner in the “most returned product” category for the back to school season.
So why is WIntel dead? Because there was an unwritten rule only tread lightly upon with the last round of WARTbooks, the PC form factor is an Intel exclusive with AMD tolerated but only barely. ARM was for tablets and throwaway consumption device, real work was only done on x86. Microsoft didn’t upset this Apple, I mean apple cart, and Intel didn’t dally in non-Microsoft OSes for non-servers. That détente will be thrown out at Computex and the long simmering cold war will get awfully hot behind the scenes. In public things will be all happy and smiles as usual but this change can’t be undone. For the past 20 or so years, every time Intel has faced off with Microsoft, there has been one clear winner. This time is unlikely to be different but it is going to be a lot of fun to watch.S|A
Latest posts by Charlie Demerjian (see all)
- 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
- Intel can’t supply 14nm Xeons, HPE directly recommends AMD Epyc - Sep 7, 2018
- AMD reintroduces the Athlon name with two CPUs - Sep 6, 2018