What The Tech 333: x86 Emulation on ARM

Share post

Conversation 2 comments

• 538

  23 November, 2016 - 4:18 pm

  <p>I’m getting a similar sort of feeling with this as I did with the iOS / Android mobile bridges that were announced a while back. It’d certainly be nice to have, but it seems like a hefty technological challenge (to do right) so it’s one thing to say your working on it and another to have it working well, and I can’t see if being the saviour of WP as some people seem to be hoping. Still, something interesting to look out for in the medium term. &nbsp;</p>

  Log in to Reply

• 8870

  05 December, 2016 - 4:39 pm

  <p>Don’t get me wrong here, I like you guys. But when you venture into topics for CPU technologies you have less expertise. This podcasts discussion on ARM versus Intel is a good example.</p>
  <p>Intel processors are and have always been built for maximum computational type of jobs. You touched on that idea in this same show, Intel CPUs are less interesting because they are measured on gen to gen enhancements and everyone knows that yardstick. ARM is not judged in the same way&nbsp;and if they were let’s be honest they would look pretty bad by comparison. If ARM was more competitive, their power and cost advantage would quickly disappear. Talking about replacing Intel with ARM is baloney (or baloney-ish) in the context you used in this show.</p>
  <p>ARM has&nbsp;done really well for phones because the computational needs are less and the degree of integration required for a phone favors their ecosystem, that&nbsp;is their real advantage. Integration is the key that helped them win phones.</p>
  <p>Ironically Intel’s big weakness in phones is related to their greatest strength, they rely on their own fabs. They can make a competitive CPU for phones in capabilities and power, but the problem is that for Intel to be competitive in the phone market they have to right size not just the CPU and chipset logic but they also need to build a silicon ecosystem that offers competitive silicon logic blocks for everything else a phone needs. And that logic does not all run on Intel fab processes. That’s why they failed.</p>
  <p>Paring down the x86 CPU for phones is straightforward and very doable. The problem that in order to be successful they have to integrate everything else into their unique fab ecosystem and that is a huge problem. Intel fabs are world class for digital logic, power, reliability, and quality. They are not so good with mixed-signal applications (think radios and tricky IO).&nbsp;Even Intel makes Ethernet and radio chips at external Fabs, why is that? The&nbsp;key problem of mixed signal cell libraries is proving to be too difficult and too expensive (even for Intel).&nbsp;</p>
  <p>The better question to ask is how well ARM will perform in areas&nbsp;such as servers and consumer PCs, how difficult it will be to migrate the massive investment of software, and will the end products&nbsp;be as flexible running apps? Windows RT shows the OS can run on ARM, but migration is a lot harder than it looks and you don’t get the benefits of the Intel ecosystem. &nbsp;That’s the trade-off, not the CPU.</p>
  <p>P.S. Can you increase the amount of space for comments? Your web tool only shows 4-lines at a time… blech!</p>

  Log in to Reply