Rick Merritt wrote: "This is not because ARM processors are weaker than Intel per se, but because the implementation of x86-based software on ARM involves a translation step to convert x86 instructions to ARM instructions. This translation step adds complexity to an implementation meaning that substantial performance overheads often result leading to a poor user experience"
MicroSoft compiles to an intermediate language(MSIL/CIL) that runs on any CPU/RISC that runs as a stack machine. It is a matter of compiling the CIL Byte Code to the ARM ISA rather than emulating x86 ISA.
X86 ISA may be better than ARM as a stack machine and could be the reason for its success.
There is a compiler option to compile to x86 specifically or to "any cpu".
My FPGA stack machine is better. It only takes 3 dual port memory blocks, a few hundred LUTs, and the performance is on par with an FPGA accelerator.
I am using the new compiler Roslyn Syntax API to program a stack machine implemented on an FPGA. Using Roslyn/C# takes care of compilation issues and the Syntax Walker shows the execution sequencing.
Intel/Altera could do an ASIC with a faster clock and who knows?
Microsoft was already moving towards a more ISA agnostic position in consumer too, they don't want to be behind the curve again. They also need Windows Server on ARM , they can't just leave that market to Linux. With HoloLens they can't exactly use x86 in the long run and that's the next form factor. ARM is a must for Microsoft. Good for them that they know it and acting accordingly.
The bigger threat to Intel is AMD and they can also hurt Intel in PC quite a lot - Intel makes a lot more in PC than in server while having 97-ish % revenue share. In PC you combine the market's decline with share losses, lower ASPs and lower fab utilization and it gets extra complicated. AMD's core might be a bit slower overall but it does seem efficient and they might get some large server customers. AMD should also have the ARM version of their core ready to go when necessary. Overall AMD is a huge threat short and mid term, ARM is still mid and long.
Richard Windsor, Analyst at Edison Investment Research:
"Intel's 60%+ gross margins are a juicy target. To date, Intel has been able to brush aside any threat to its dominance in processors for the data centre, but with Microsoft deciding to port Windows Server and Azure to ARM, the threat is back. The data centre has long been the saviour of Intel's financial performance as it has been supporting the company as the legacy PC business has been going through its rough patch. ARM has taken pot shots at the data centre before with semiconductor makers announcing chips, but this has never gotten off the ground.
We have long been convinced that the main reason for this is software as many servers run huge amounts of legacy software that will need to be ported in order for an ARM based server to work. Until now, no one has been willing to do this. This is why the demonstration of Windows Server running on the ARM based Qualcomm Centriq 2400 is so significant. If this can show to run with similar performance characteristics to Intel, then it would make a lot of sense for Microsoft to begin migrating its servers over.
There are some signs of this already and Hewlett Packard mentioned weakness in a large customer on its most recent earnings call which is widely believed to have been Microsoft.
However, this solution is still internal only to Microsoft and we doubt that it will be willing to take any risks until it is sure that it can work just as well as Intel. The key to this is performance and here Intel has historically beaten ARM-based processors hands down time and again. This is not because ARM processors are weaker than Intel per se, but because the implementation of x86-based software on ARM involves a translation step to convert x86 instructions to ARM instructions. This translation step adds complexity to an implementation meaning that substantial performance overheads often result leading to a poor user experience. In the data centre, performance is critical and so it is a very big stretch to say that Intel is beaten. In fact, there is a very long road ahead for ARM to make a real impact in servers but with the company now privately owned, there is plenty of money for investment.
The real risk for Intel here is not so much market share but margins.Group gross margins are still way above the industry average at over 60% and if real competition comes into the server market, these are going to come under real pressure.Qualcomm is highly motivated to make this work as its core markets are now saturated leaving it needing other avenues with which to pursue further growth.This is what lies behind its acquisition of NXP as well as its aggressive push into other areas such as automotive.
As always, the outcome of this battle for the data centre will be determined by execution and whilst we see no immediate threat to Intel, Qualcomm and Microsoft represent its greatest threat to date."