ARM vs. Intel – Silver bullet?

Reply to this post

RFM AvatarSmall

 

 

 

 

 

Last time it was software. This time its emulators 

  • Qualcomm and Microsoft have announced that Windows is once again coming to the ARM processor but this time the approach is completely different to the disaster that was Windows RT.
  • In Windows RT, Microsoft modified Windows 8 such that it would work on an ARM processor and in the process killed flexibility and backwards compatibility to legacy software.
  • The result was a platform that was shunned by both developers and users, completely killing any hope that ARM would gain penetration in Intel’s home turf of PCs.
  • The fact that Intel has cut its lower end Atom line of products that aimed to compete with ARM in Android tablets has left space in the market for these products to grow into.
  • This time the approach is completely different as Qualcomm and Microsoft have produced an x86 emulator that fools the software into thinking that there is an x86 chip present.
  • The net result is that any Win32 and universal Windows app will run on the device with no modifications being required by the developer.
  • The net result is hoped to be cheaper, fan-less, always-on, mid to low end PCs that have longer battery life than their counterparts powered by Intel.
  • Qualcomm and Microsoft have also promised that Adobe Photoshop, Microsoft Office and Windows 10 games will all run on these products and it is here that I find the big caveat in this strategy.
  • This caveat is performance.
  • Intel processors may be power hogs but they offer blistering performance in real world devices as well as in benchmark tests.
  • ARM has been able to match some of the benchmarks but has never been able to come close to Intel in real devices.
  • This is why the mention of Photoshop, Office and games is so important as these three are well known to be very processor intensive.
  • Their requirements are so high that the software is written directly to the processor (written to the metal) to avoid any lags created by going to the processor via the operating system.
  • This is where the problem will occur as processor heavy apps will no longer be written directly to the metal but instead will be going through the emulator.
  • The emulator process is as follows:
    1. Translate requests from the x86 programs sitting on top of it into the RISC instruction set that ARM understands.
    2. Execute the request on the ARM processor.
    3. Translate the results back into the x86 instruction set so that the app can run.
  • Consequently, the emulator will incur additional processing overhead as well as consume power.
  • The big questions are how much will it consume and will it have an impact on the overall user experience?
  • For Intel, this is a critical question because if there is no impact it could see its market share in the mid-range PC market (most of the volume) come under serious threat.
  • In Q3 16A Intel reported non-GAAP gross margin of 64.8% compared to Qualcomm at 58.9% but if I remove the profits from licencing, I estimate that Qualcomm’s chip gross margin is around 40%.
  • Consequently, if Qualcomm’s Snapdragon chipset plus the emulator can match Intel’s performance, Intel will have to cut its prices to stay in contention.
  • This could see its gross margin come under sustained pressure as the first real challenge to its monopoly finally hits home.
  • History is on Intel’s side as emulators on battery powered devices have always impacted the user experience so much that the experience failed to win over users.
  • In order to put pressure on Intel, the Qualcomm powered Windows 10 devices will have offer the same level of functionality and performance, better battery life as well as a cheaper price.
  • These are my three criteria for Qualcomm to really challenge Intel and success will come down to the quality of the emulator that it has created.
  • Qualcomm will also need to work closely with the device makers as there are endless hardware configurations for Windows 10 PCs and clumsy integration could easily make a complete mess of the elegant product that Qualcomm and Microsoft have created.
  • The first devices will be available early 2017 (launch at CES 2017 looks likely) and it is by these that Intel’s outlook will be judged.
  • This is obviously negative for Intel but it is worth remembering that every attempt to dislodge Intel to date has been a miserable failure.