Software has become a spoiled brat grown fat suckling on chip industry innovation

Software has become a spoiled brat grown fat suckling on chip industry innovation

Summary: Hardware technologies have provided an incredible innovation platform but it won't last forever. Where are the advances in software?

TOPICS: Software

Daniela Hernandez at Wired, interviewed the founders of DropCam, which uses cheap webcams for remote monitoring systems. For some reason, and a very wrong reason, Wired's editors decided on this headline:

Software Is Still King. Hardware Is Just Along for the Ride | Wired Business 

It's so very wrong and I'll explain why, as the chip industry's largest trade show, Semicon West opens this week in San Francisco.

The Wired article article talks about a hardware "renaissance" but that software is "still king."

That could not be further from the truth but it is a good example of how little understood are the incredible accomplishments of the chip industry. And the headline reflects a popular misconception about software's role in the story of our modern society and in the relentless march of technology.

Software is important, it instructs the hardware, which makes it seem as if it fuels the extraordinary engine of innovation that drives technological progress. But don't confuse the driver of a car with the car's performance -- this is not the Flintstones.

The plain fact of the matter is that it is software that gets the free ride, and hardware in the form of faster, smaller, cheaper microprocessors is what's responsible for computer technology's relentless leaps in performance.

This is the innovation platform of our age. And it's the chip industry, not software that has raced to meet the promises that tech companies made in transforming so many industries and businesses. 

What other set of innovations continually doubles performance and slashes its own prices with such clockwork reliability? Without that progress our technology revolution would have stalled out decades ago.

Yet there's very little celebration or acknowledgement of the incredible innovation that continues to be made by the chip industry. 

Slivers of condensed high technologies…

Microprocessor chips are magical, alchemical creations wrought from the base material of sand, and transformed into tiny, shiny slivers of condensed high technologies, and far more valuable than gold.

["Condensations of technologies"  is apt because the manufacturing process includes vaporizing advanced materials so that they can be condensed into layers, atom by atom.]

And every two years the chip industry manages to innovate around problem issues called "showstoppers" to deliver double the performance and lower prices. The software industry has never managed to do that.  

Chips are incredibly difficult to make and require the most advanced and expensive manufacturing systems of our age. Semiconductor fabs cost a few hundred million dollars in the 1970s, now its closer to $8 billion. 

Smaller is faster

The chip industry's innovation comes from continually being able to shrink wiring and transistors into ever smaller geometries. Shorter connections between transistors equals faster performance because the electrons travel shorter distances. And smaller chips use less power, making mobile devices possible, and making each new generation of chips a little bit greener. 

Shrinking the size of chips means more per wafer and that leads to lower prices. This continually opens new markets as less expensive chips enable new types of products.  Low prices for digital electronics is essential in distributing the information age into the hands of billions of people. Can software do that? I don't think so. 

This process of shrinking chip geometries is what Moore's Law refers to in its observation of a doubling in computer performance over a two year cycle. It has absolutely nothing to do with advances in software technologies. 

Tickets please…

Software gets a free ride, it runs faster, better, without having done anything. And software developers are happy to ride on the improved performance of their code for no effort at all. They all ride free on the hard-won innovations developed by chip engineers. 

How can software be king when there's been no new technologies? Coding is turning into scripting, an etch-a-sketch of connected software frameworks lumbered with huge amounts of redundant code. There are no lean, mean, double-speed software technologies on the horizon. In fact, software developers have gotten lazy, and lazier, their code grows ever more bloated and inefficient because it can, because the hardware makes even badly written software run like a champ.

Software gets a free ride but the ride will soon be over. Moore's Law is slowing and before the end of this decade it will largely end its phenomenal 60 year-long reign. Software will have to provide the improvements in computing performance that hardware once did. 

A despotic ruler?

Software will eventually become king. But it has to change radically, or it will be a morbidly obese monarch, demanding ever higher taxes from the spent chip technologies on whose groaning backs it tries to build ever larger monuments  of faux progress.  

So far, there's alarming little to report in the way of breakthrough software technologies -- there's not even a gym membership.

Topic: Software

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Hardware Versus Software

    In a CPU or memory chip, the fact that they can pack in more and more copies of the same old thing (multiple cores, multiple function units, multiple memory cells) still counts as an advance. In software, it doesn't; we actually have to develop new stuff, we can't count copies of old stuff.
  • That's more like it Tom!

    But let's name names, shall we?

    Why does Ed Bott not say "With Windows 8 MSFT 'reimagines' the operating system as a hideous chimera of the APPL ecosystem: the enforced thrusting towards the dysfunctional and incomplete METRO user interface; the imprisonment of the user in expensive hardware which cannot be reconfigured or extended; the insidious straighjacket of paltry METRO applications which must be delivered via the STORE after MSFT's parasitic 20-30% cut on developer revenues."

    I find the APPL and MSFT cut of developer sales disgusting and to your point. Those companies, having grown fat in collaboration with hardware advances in chip design, storage design, memory design, network design and infrastructure roll out, customer distribution chain ... have turned on their former allies and wish to bleed them dry.

    Why do ZDNET commentators not see the damaging strategic creep? In 2010 you could buy 3 decent mid range PC's and 3 Office licenses for £1,000 ... now you'll be lucky to get a Surface Pro and a subscription - A ONE YEAR SUBSCRIPTION mind you - for the same.
    "Will be a morbidly obese monarch?"

    What's all this 'will be'? Doesn't APPL have more reserve cash than the US friggin' Government?
    And by the way ... what about those taxes!

    And Larry Dignan thinks its a waste of time complaining. Loser.
    • Technically, complaining is a waste of time...

      It's a waste of time to complain unless you 1) have a bunch of people who feel the same way, and 2) are willing to actually do something about it that will affect the target of your complaints in the long term.

      You, on the other hand, have offered no such plans to take your business elsewhere, or even to consider doing so. Therefore, you would be better off printing out your complaints and sending them to Ballmer on a cardboard roll, so that he can give them all the attention your impotent whining ultimately deserves.

      Just like your incessant whining on the allegedly declining quality of this site--do you have another site to go to, or are you just doing it to stamp your foot and hold your breath? Because if it's the latter, then you're contributing to the decline with your own white-noise complaints.
      Third of Five
  • I Couldn't Agee More

    When I started in this industry I had to write business application programs that would run in 10K. Today, developers write programs that they hope will run at an acceptable speed, ON THE NEXT GENERATION OF HARDWARE.
    • Re: business application programs that would run in 10K

      How many months did that take you?
  • Software innovations

    There are software innovations which help the knowledgeable software engineer to develop (relatively easier) faster and more efficient software. Having taken a liking to optimization and parallelization techniques with C++ and .Net, I take some time to throw my code through the wringer and make the end result really good with fast and efficient software.

    .Net with WPF is/has a lot of innovation that makes it easy to write optimized software. However like a lot of tools, it is even easier to write really inefficient software.
  • Timely article

    As CTO of a fabless semiconductor company I can tell you that pretty much all hardware companies understand and value software as a key component of our solutions

    Unfortunately as the Wired article demonstrates the converse is not true

    This is going to become a real issue for the software industry in the coming decade as Moore's law starts to slow and the "build it and they'll come" philosophy of the SW industry will start to fail

    Power dissipation is a key concern from embedded applications through to HPC and people with vision at both ends are responding to challenges like the DARPA PERFECT initiative which seeks to achieve performance levels of 75GFLOPS/W for HPC (incumbents are currently delivering 10% of this). Large cloud or HPC installations which SW companies are using to deploy their services typically use 30-40% of their power just to cool their data-centres and for many of the cloud providers power is the single largest cost in their business

    The extensive work done by Bill Dally at Stanford (now CTO @ Nvidia) shows that huge power-savings (10x) can be realised by achieving better locality in both program and data in embedded systems. Now at Nvidia Bill is driving the next generation Echelon architecture using the techniques learned from the Stanford ELM project and Nvidia intends to roll out this architecture in their entire product range from mobile phones to HPC

    Here at Movidius we have just announced ( our new computer vision processor for mobile phones enabling computational vision applications from Lytro style refocusing to advanced AR, SLAM and robotics in a tiny low-cost device dissipating a few hundred mW

    This type of application typically required either a high performance desktop and GPU or cloud infrastructure using existing software paradigms (OpenCV)

    In summary for applications where power matters software engineers are going to have to think about data locality as a central part of their work and their future will contain bright shiny new hardware!
  • Flashbacks to 8-bit coding?

    If you've been around a while you'll remember what it was like trying to eek every bit of performance out of a 6502 processor with only 64k of ram (really, more like 32k after overhead).

    Will we start seeing coding efficiency become important as it was in the past? No more sloppy, bloated, interpreted code?

    I certainly hope so.
  • Some Software can be King

    There is a varying degrees and levels of software depending on the underlying ideas, code writing quality and overall execution. Some software, the Kings, will rise to the top. Our latest soft technology relieves much of the burden on the hardware.