Microsoft to Use ARM to Displace Intel in Cloud

[Arthur Hanson]

It looks like the forever alliance know as Wintel may be coming to an end. I don't expect Intel to give up without a fight, but this is the first very serious threat to the Wintel monopoly that was destroyed in mobile. Microsoft better look out also, for as the monopoly breaks up, their dominance could also break up as it has almost no footprint in mobile. The world is changing faster and faster and as Andy Grove said "Only the paranoid survive". The pace of change fully illustrates, we are truly in the age of the "Great Acceleration" and it's picking up speed.



Look out, Intel? Microsoft tries ARM processors in the cloud with Qualcomm partnership - GeekWire

 

[JeffreyHF]

If your concern is that Microsoft has "almost no footprint in mobile", they could solve that in one elegant move, by acquiring Qualcomm.

 

[Arthur Hanson]

Microsoft, Google and Apple have enough money to buy themselves into any market they like, but you do have to have the culture to go with it and integrate it. During this process, the whole game can change at any time in unintended ways.

 

[Bernard Murphy]

Perhaps as much a negotiating tactic as an anything else. No buyer likes to deal with a monopoly, people who buy servers are no different. This will put pressure on Intel to provide more for less. If they don't deliver, MS, Google and others can wave the threat of ARM-based servers at them. Unclear how quickly that threat would be realistic. Intel has huge investment, much more than core CPUs, in making their server platforms the best for server applications. Not that others couldn't catch up eventually but it will take a lot more than just showing off ARM-based servers.

 

[name99]

The message discipline ("This is all just a bargaining tactic") from the Intel ranks is quite astonishing. Not that it will change anything...

We have been through this before with mobile, with a multi-year-long "Intel will unleash their uber-superior chip design and fab skills any day now, just you wait". But the true believers don't want to larn anything from that episode either.

It seems to me much easier to simple accept that MS mean what they say. Intel provides a constrained set of products, optimized for Intel (short-term) revenue rather than technical excellence, with a substantial burden of overhead (primarily in design time and thus very slow readjustment to changing customer needs) to support functionality (x86 programming model) that MS does not need or care about in this space.

A primary flexibility that the ARM vendors will be willing to offer, and that Intel has not been willing to, will be over memory controllers. Intel is very big on segmenting by memory controller (thus forcing you to buy substantial compute performance even in the main thing you want is a large physical memory).

Likewise Intel's current solution if your problem is well-matched to HBM is to force you to buy a Xeon Phi, even if the computer aspect of a Xeon Phi poorly matches your needs.

Likewise Intel seems incapable of actually delivering persistent RAM. We've gone from Optane being about to sweep the world in nvRAM form to Optane being available (in limited amounts) ONLY for SSDs. And you just know that by the time Intel gets its act together for nvRAM, it will charge an arm and a leg for the Xeons that support it.
Meanwhile Optane is not the only nvRAM game in town, and it may well be that an ARM server vendor actually ships a complete, usable, nvRAM system before Intel does...

Essentially Intel remains stuck in a world where what it sells is LARGE steam engines that you're expected to treat like special snowflakes, and to which you're expected to adapt your business and programming model. Meanwhile ARM sells electric motors of all sizes, from tiny to huge.
If you want to build your car with 120 electric motors in it, from tiny to very large, ARM will happily supply you with all you need. Whereas Intel will ONLY supply you with a steam engine, all the while lecturing you that, with enough belts and gears you can easily run every window, the trunk, and the CD player all off that single engine, that you should be grateful to be part of the colossus that is the Intel family, and to get to do things the Intel way; and that your radical ideas about using lots of small separate motors are clearly stupid hippy thinking...

The amazing thing is that Intel LIVED THROUGH the world as it changed from the Steam Engines named System 360 to the rise of the PC and the Server. But Intel is a financial company these days, not an engineering company. Its finance-blinded management has no knowledge of, and no interest in, history as ancient as the 1980s, let alone the 1880s.

 

[lefty]

Interesting rumour from Digitimes says that Qualcomm's 10nm Centriq 2400 is made at TSMC, (not Samsung, like other people were speculating).
TSMC lands orders for HPC chips from Nvidia, Qualcomm, says paper

 

[hist78]

The amazing thing is that Intel LIVED THROUGH the world as it changed from the Steam Engines named System 360 to the rise of the PC and the Server. But Intel is a financial company these days, not an engineering company. Its finance-blinded management has no knowledge of, and no interest in, history as ancient as the 1980s, let alone the 1880s.

Is Intel just too big to compete in the fast changing semiconductors industry?

I think inside of Intel, there are way too many conflict of interest that can't be resolved under one roof.