Friday, August 23, 2013

Observing Microsoft, Part 4

This day is an interesting one for Microsoft. First, Ballmer sends out a letter to employees that states that he will resign within 12 months. Then it is announced that there is a committee on the Microsoft board, containing Bill Gates, of course, which has the responsibility of finding a new CEO. No, I suspect that Ballmer is not on that committee.

Some writers are saying that Microsoft is not forcing Ballmer out. But think about it. If you had to get rid of a failed CEO who owned 333 million shares of your company's stock, what would you do? It was most certainly a negotiated force-out. With a legal release. And probably some kind of honorary employment that requires Ballmer to only sell within certain windows of time and keeps him on a leash.

Welcome to the mobile revolution.

I must say that this change is way too late. After all, in 2010 people were already clamoring to fire Ballmer. And doesn't clean things up soon enough. Obviously Microsoft's board or directors should have been doing this for the last several years!

The reorganization that Ballmer has been accomplishing seems like a smart idea, except that it is trying to make a silk purse out of a sow's ear. It's made for the PC era which is slowly fading away. Still, the new organization is probably one less thing that a new CEO will have to worry about. That is: if he accepts this vision for the new Microsoft. A vision that depends upon Microsoft succeeding in the mobile revolution. Still with the reorg, Microsoft has a corporate culture that can't simply turn on a dime.

And Windows is exactly the problem.

Energy Efficiency

The mobile revolution has created two very interesting trends in the computing landscape. These are battery longevity and cloud computing. In order for batteries to last a long time, the products they power must be energy-efficient in a system-wide way. In order for cloud computing, with its massive compute farms, to be cost-effective, each server must be singularly power-efficient and generate as little heat as possible since cooling is a power consumption concern as well.

Of course battery longevity also affects electric cars like the Tesla. But, when it comes to computing, the battery longevity comes from three sources: more efficient batteries, hardware systems where power efficiency is an integral part of their design, and finally the economical use of resources in software. In the cloud computing arena, instead of more efficient batteries we are concerned with heat dissipation and cooling strategies.

More efficient batteries is a great thing, when you can get them. But advances in supercapacitors and carbon nanotube electrodes on various substrates is yet to pan out. This means that hardware systems such as SoC's (Systems on a Chip) must be designed with power efficiency in mind. Power management solutions that allow parts of a chip to turn themselves off on demand are one way to help.

Even at the chip level, you can send signals between various components of an SoC (System on a Chip) by using power-efficient transmission. For example, the MIPI M-PHY layer even enables lower power consumption by the transmission of the high-frequency data that usually chews up so much power. Consider using a camera and processing the data on-chip. Or using a scaler that operates from/to on-chip memory. These applications involve images, which are huge resource hogs and must be specially considered, in order to save significant amounts of power.

But there's more to this philosophy of power management, and this gets to the very heart of why SoC-based gadgets are so useful in this regard. General tasks that use power by processing large amounts of data are handled increasingly by specialized areas of the SoC. Like image scaling and resampling. Like encrypting and decrypting files. Like processing images from the onboard cameras. Like display processing and animation processing. Like movie codec processing. Each of these applications of modern gadgets are resource hogs. So they must be optimized for power efficiency at the very start or else batteries simply won't last as long.

Of course, you could simply user a bigger battery. Which makes the product larger. And less elegant!

Windows?

So what is the problem with Windows? The Wintel architecture wasn't built from the ground up for power-efficiency. Or distributed specialized computing, like so many gadgets are constructed these days. And now you can see what a daunting process this must be for Microsoft engineers that basicaly have to start over to get the job done. It will take quite a bit of time to get Windows to run on an SoC. Almost all implementations of Windows today are built to run on discrete CPUs. The Surface Pro appears to use a regular CPU board with a stock Intel part.

You see, power efficiency isn't just a hardware problem to solve. The software must also have this in mind with everything it does. The consumption of resources is a serious issue with any operating system, and affects the user experience in a huge way. I can't even begin to go into the legacy issues with the Windows operating system. The only way is to rewrite it. One piece at a time.

This problem has led many companies who lead the cloud computing initiatives to use Linux for their server operating systems. Mostly because it can easily be tailored for power efficiency. The server operating system share of Unix-based operating systems is 64%, compared to about 36% for Windows.

Servers are almost certainly going to go the way of the SoC also, with dedicated processors doing the expensive things like video codec processing, web page computation, image processing, etc. But I do see multiple cores and multithreading still being useful in the server market.

But not if they increase the power requirements of the system.

On mobile devices, Windows hasn't done so well either. Windows Phone probably has less than 3% of the mobile space, if that.

The Surface never clicked

Why didn't the Surface RT and the Surface Pro tablets succeed? First off, it's possible that they are simply yet to succeed. I just had to say that.

But more likely they will never succeed. It's hard to move into a market where your competitors have been working on the hardware solutions for years. And when hardware isn't your expertise.

At first, the Surface marketing campaign was all flash and no substance. A video of dancers clicking their tablet covers into their Surface tablets was certainly criticized by a few bloggers as vacuous. The main problem was it stressed the expensive keyboard cover, and skirted the issue that the cover is totally needed. With the cover, the Surface tablet becomes just a crappy laptop. That you can't really use on your lap, because of the kickstand. Their follow-up video was curt and to the point, but sounds a bit like propaganda. saying "Surface is yours. Your way of working. Your way of playing".

Yeah. Trying to get into the mind of their prospective users.

But it's clear that their strategies were simply not working, because they went to the old adage "if we don't look good, then maybe we should just make them look bad". And they started releasing anti-iPad ads. The first one used Siri's voice to sum it up "do you still think I'm pretty?". They compared the price of the legendary iPad to the Surface RT without a cover. I suspect that a Surface RT without a keyboard cover is pretty much useless. The next anti-iPad ad compared features in a less quirky way. But anybody using a Surface RT knew that it didn't support the apps that the iPad has, or really have any of the advanced iOS/iTMS ecosystem in place. And without the keyboard cover it was cheaper, certainly. But you really had to have the cover to get full functionality.

So Microsoft decided to drop the price. This was echoed in the nearly $1-billion charge they took that quarter. Then they followed up by dropping the price of the Surface Pro! It seems desperate to sell their inventory. Otherwise they will be taking another huge charge against Windows revenues like before.