Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
I, Cringely - The Survival of the Nerdiest with Robert X. Cringely
Search I,Cringely:

The Pulpit
The Pulpit

<< [ Do What I Mean ]   |  Speed Bump  |   [ PayAcquaintance ] >>

Weekly Column

Speed Bump: While Moore's Law Isn't About to be Repealed Soon, We Might See It Slowing Down a Little

Status: [CLOSED]
By Robert X. Cringely

The only certainty in the computer industry for the last 30 years has been Moore's Law, which says that computing power doubles every 18 months. From time to time, it looks like Gordon Moore is going to be repealed by some technical limitation, then clever engineers think of a dodge, and we're safe for a few more years. While Moore's Law probably won't go on forever, we are certainly safe through at least the end of this decade. But that doesn't mean even Moore's Law isn't subject to outside influences that might put a crimp in performance growth of economically-viable computing. That appears to be the case right now as two different effects are just starting to be felt in the industry -- effects that will tend to drive prices up, not down, and drive smaller manufacturers completely out of business. These effects are, not surprisingly, almost unknown to the people who will be hurt most by them.

Scientists at Stanford University have discovered that there is a finite limit to the speed at which you can apply a magnetic charge to recording media. We won't hit this limit for several more years, and there may be a way to get around it using magnetism and heat together, but either way, there is a huge and very expensive RR&D mountain to be climbed in the storage industry. Similarly, IBM has had a very difficult and expensive time getting its new 90 nanometer CMOS fab running in New York -- an experience shared by ALL the leading-edge fabs as they come online. It is becoming so expensive to build these plants that there are likely to eventually be no more than a dozen or so around the world by the end of this decade. They will be built either by wealthy companies like IBM and Intel that can't afford not to, or by national consortia in countries like Japan and Taiwan that view this industry as a strategic national priority.

What this means to you and me is that in both disk storage and semiconductors, the big outfits get bigger and the small outfits consolidate or die. It is true that smaller feature size and bigger wafers ideally mean that semiconductor per-part costs drop along with energy consumption, but there are limits to this effect. While Moore's law can continue to march on for a while longer, the cost to make each step is becoming more expensive. At some point, perhaps soon, the cost to increase capacity could exceed the market value of doing so. The price, performance, and replacement march may be coming to an end, which is not all bad unless you make your living building this stuff.

But wait, there's more! Rapidly approaching from the opposite direction is EU Directive 2002/95/EC, "the restriction of the use of certain hazardous substances in electrical and electronic equipment," also called RoHS, or more commonly, Pb-Free. While this initial legislation is from Europe, it is being adopted globally (though most of the manufacturers affected don't yet seem to know that).

Pb-Free is primarily aimed at reducing the amount of lead dumped in landfills, but includes five other chemicals commonly found in electronic components, cases and cabling. Electronics contain lead because connections are made with solder that is a 63/37 mix of tin and lead, which melts at significantly lower temperatures than the same metals used independently. Pb-Free primarily means replacing solder with something else that usually doesn't work as well.

After July 1, 2006, you can't sell your product within the EU if it contains more than 0.1 percent lead, and adding ballast weight is specifically not allowed.

The key replacement for tin/lead solder is a tin/silver/copper alloy available exclusively from Japan. While first appearing to ignore the initiative, Japanese companies quietly developed and patented most of the 20-plus alloys needed to implement it. Ironically, silver in the water supply is much more toxic than lead to marine life, something we'll probably get around to addressing in later series of laws and regulations.

But going lead-free means much more than swapping one solder for another. Circuit board plating has to be changed to work with the new alloys. Soldering equipment has to be changed, too. And because of process temperature differences, you can't easily combine old and new soldering technologies on a single board, so there can be no easy phase-in of production. It has to be done cold turkey, which means inevitable problems and supply disruptions.

For extremely dense assemblies like mobile phones or graphics cards that use ball grid mounting, another alloy adding bismuth to the mix is required to get the proper flow-out. The problem there is that the presence of lead in any quantity destabilizes bismuth alloys with contaminated assemblies literally falling apart. This might not be a problem if there was a standardized marking scheme to identify Pb-Free components, but at present, there isn't one. That isn't really surprising, though, given that there is also no Pb-Free qualification or standard for companies to build against. So far there is no place to take your product to get it certified lead-free.

This doesn't mean that there aren't already Pb-Free products. Sony's PlayStation 2 is Pb-Free. So are some Nokia phones, though they also have half the life expectancy of Nokia phones containing lead.

Conventional wisdom in Pb-Free circles says that the new devices will cost 15 to 25 percent more, though that couldn't be wildly off in any direction. Power supply manufacturers seem to be especially behind in implementing Pb-Free and there will be shortages.

Again, this is a trend that will hurt smaller players. The Tier I (extremely high volume) manufacturers already produce some amount of Pb-Free, and their practices, standardized or not, will be entrenched by the deadline. But none of the Tier II manufacturers are ready, and Tier III hasn't even started looking at it yet. Due to the scope of the problem and the limited funds for Tier IIs (slim) and IIIs (zilch) to do R&D, I see both relying on published standards to get themselves in line. Only problem is THERE ARE NO STANDARDS. So some little companies will die and prices may be pushed up more as a result.

The great unanswered question is whether any of this will be noticed two years from now. There is certainly a speed bump of sorts coming, though I've given it some thought and the results are far from clear. Of course, it is a great time to be in the Pb-Free equipment business, or shortly will be. And procrastination is nothing new when it comes to environmental standards. If President Bush is still in office two years from now, he may well just push back the deadline, maybe forever.

If the deadline is firm, then the big will get bigger, but there will be other effects as well. If you have to replace all that wave soldering equipment anyway, what else logically falls out of such a decision? Given the current view that Pb-Free connections are less reliable, I'd see an even stronger move toward Systems On Chip, which minimize the number of soldered connections. New environmental regulations have long been the impetus for technological development and is one of the few ways short of starting wars that governments can actually effect such change.

So I think this is going to be ugly and expensive, and there is going to be a lot of head-scratching and doom-saying rather like Y2K, but in the end, I'm just not sure how much it is going to add to the cost of goods. If the cost without Pb-Free would have been N and the cost to be Pb-free cost is 2N, but some of that cost is payment to accelerate from 90 nanometers down to 65 nanometers, which itself drops the real estate cost by half and the power cost by even more, maybe we're lots better off Pb-Free. If probable connection failure drops the MTBF from 100,000 hours to 50,000 hours, will we really be inconvenienced that much?

What I find most disturbing about this is that hardly anyone is thinking about it, but then there is nothing new in that, is there? I first wrote about Y2K in 1990, the world didn't go berserk about it until 1998, but that was enough time. The saving grace here is people like the reader who brought this issue to my attention DO worry about these things. Sharing those concerns make it much more likely that we'll weather this storm, too.

The problem I find much more interesting is the one we started with, diminishing rates of return on capital investments in storage and semiconductors. Now this seems to present a wonderfully opportunity. You see when the hardware vendors stop doubling capacity every 18 months or so, they are going to have to squeeze out value some other way. In the past several years, there has been little done to improve the performance of software. So there is now a lot of room for improvement by just cleaning up and speeding up the existing code. That is, of course, if you haven't offshored all your best coding talent.

Comments from the Tribe

Status: [CLOSED] read all comments (0)