The Spec Sheet is a lie|Logic design

Marketing Opacity: Business V.2

Well, Dan. I’m not a super smart SC engineer like you are, but I do understand that some folks like what they like better than what they don’t like because of perceived importance. This singular ‘metric’ will then be used as the most important decision making factor. I would imagine that in this case your comment is predicated on the performance of the x86 Atom based SoC’s (System on a chip)
First, I would like to thank smart people like you for designing cool stuff. EE’s rank very high on my list of professions I respect, and I have EET’d for some really great embedded systems engineers.

That said, the performance of the computer system is but one factor in the over-all design of a product based solution. A product that utilizes a higher frequency microprocessor isn’t  going to be the fastest performing product in its class simply because of its processor clock speed. Your comment, when taken in context with the question, seems rational. As long as you don’t consider:

  • Why Atom was created
  • It’s age
  • The extra chip it requires
  • Thermodynamics

I’m sure we will agree it isn’t because of the C vs R ISC architecture), while the new (or proposed for 2016) Atom cores have a decent level of integration, the data sheet seems to stop at (somewhat weak) GPU and camera sensor. Neither of these deficiencies are because it’s a CISC) Some competing RISC based SoCs contain barometers, gyroscopes, accelerometers, RF radios, Xband receivers, non vote-shared cache, more switches, smaller interconnects, a smaller physical size, better heat coefficient and thermodynamics than Atom does (almost to the point where active cooling would be required) without clock throttling. These solutions that are designed for mobile products, didn’t need to be scaled from a microserver solution to do its best be wear a mobile hat simply because of revenue concerns. doesn’t worry about Google and Microsoft product manager relationships from which to hinge hopes for it’s market longevity.

That’s because ARM Licensed mobile SoC’s do what they were designed to do:
Power mobile technology.

Now, I understand what you have just read has nothing to do with processor architecture. Just like the RISC portion of my contribution to this question had nothing to do with the computational and process delta that each deploy. It was simply based on the flexibility and the ‘hard-point’ options you get when licensing with ARM. 
You get performance where it’s needed for the mobile industry because it’s a designed-for-mobile solution. 



Obviously, there are a couple of handsets that will utilize the Atom core based SoC. Why? Relationships, strategic marketing and one Tech Giant just couldn’t get their thrown together and patchwork compiler, some poor decision making, integration on expandable memory, doubled down on a philosophy of ‘security isn’t a priority unless security sells pay-per-click ads. Mix in a culture of, ‘If we can’t give them away fast enough…dog food them.’, market share strategy. 
The fact is, there are not any OEM’s who can run with the performance benchmarks that other fruity flavored OEM products despite the latter having a -400Mhz ‘disadvantage’ in clock speed, outperforming the entire industry and only expending %25 of the system memory to run an identical application.If Atom makes gains in mobile it won’t be because of its superior design. It will be because 2 Tech giants decided to enter a vertical market because neither one purposefully engineered a producer intended for the targeted market in which it peddles its wares, and relies on a vertical market strategy that was created by 2 mediocre solutions expending resources to overhaul their products.Their direct competitors simply engineer a solution with the purpose of properly bridging an identified gap, instead of looking for ‘good enough’ holes in which to dump last years overstock.
So…to wrap it all up, there may or may not be a technical impediment that keeps x86 CISC based SoC from performing as well as a RISC based uProcessor in a lab, or on the spec sheet. Keep in mind that the logic isn’t the only item on the BOM, 
Now, there is nothing wrong with making the decision to sell your micro-server processor design to a mobile handset OEM after you re-write the data sheet with a crayon. Nor is there anything wrong with giving away free code, free search apps, map apps and other free software and services to one person, just so you can use that free stuff to capture data points, trend app and search data and sell it to someone else. That person will then use that data generated by the user to sell stuff right back to the same person who generated that data in the first place. There might not be anything wrong with that either.
But is there anything right about it?
Performance of products, especially in mobile products isn’t based on bigger-faster-more. Sometimes the most important decisions are met with the greatest difficulty, not because you can’t figure out what else to put on a product. It’s because of the stuff you need to leave off.

Jason Yeaman

 

Advertisements
%d bloggers like this: