An open source war
- 01 March, 2001 16:38
My first contestants, should they decide to accept the challenge, are Advanced Micro Devices and Intel. I love to watch these chipmeisters wage war. Thanks to the competitive challenges posed by AMD with its Athlon series of chips, the price/per-formance in CPU chips and supporting chipsets offered by both these companies is phenomenal. I picked up a 1GHz Athlon a couple months ago for well under $US300. I'm almost afraid to find out how much the faster chips are selling for now, lest the speed greed demon bite me yet again.
The catalyst for the fight is some advice I'd like to offer these two competitors on how each might leverage open source to attempt to beat the other. The operative word here is attempt, however. I don't want either company actually to subdue the other for any extended period of time, since the ongoing competition is what keeps prices low and performance high.
Enough beating around the bush. Here's the advice: go for broke this year and invest heavily in open source. Specifically, devote as many resources as possible for tuning optimisations in the GNU compilers to produce code that runs best with your chips. I'm not talking about bolstering generic 686 optimisations. I'm talking about chip-specific optimisations that a developer would activate with a compilation flag such as -O Athlon, -O Duron, -O Thunderbird, or -O Pentium Whatever. Hey, these companies should even add some new processor instructions unique to their individual CPUs, and then use the optimisation flags to make the compiler use these extra instructions. If it improves performance, why not?
Then AMD and/or Intel should post benchmarks demonstrating how much faster Linux and the various BSDs run when you rebuild them using these optimisations and run them on specific AMD or Intel hardware. Ideally, the benchmarks should be geared toward ISP, ASP and database performance.
Then these companies could convince one or more of the commercial Linux and BSD distributors to offer versions of their products that are optimised specifically for certain high-performance chips. Better yet, Intel and/or AMD could rebuild a handful of distributions and then offer the CD-ROM images free for the download. If the performance gains were significant, I guarantee you'd witness instant word-of-mouth advertising and a run on whatever motherboards were best for the chips in question. (A tip to these companies: if you do compile your own distributions, don't forget that you have to rebuild all of the applications, not just the kernel, in order to make this strategy worthwhile and attractive to potential customers.)
Most important is this advice: get started on the above strategies immediately, because all of the external factors are aligning perfectly for this move to produce great market gains.
In the first place, open source is about to get a very big boost because the dot-com economy is tanking. I know that sounds counter-intuitive, but open source guru Bruce Perens was right when he recently observed that a slow economy presents the best opportunities for open source. Free software is where you turn when you no longer have money to burn. That's exactly the position the dot-coms find themselves in right now. The dot-coms aren't floundering because the opportunities aren't there; they're floundering because they got spoiled on the irrational exuberance of investors and didn't learn how to spend their money wisely. Open source is in a perfect position to benefit from the lessons in frugality that these companies will now have to learn.
Second, Linux and possibly the BSDs are going to get a lot more attention and updates this year. The long-awaited Linux 2.4 kernel has been released, and minor kernel updates are sure to follow. Watch for a flood of new versions as the commercial distributors line up to ship their versions of Linux with the 2.4 kernel. Then expect a second round of distribution bug-fix updates later this year. Shoot for that round of updates.
Finally, do it now, because you can. And credit this fact to the nature of open source. Better yet, this contest doesn't even have to be limited to AMD and Intel, although that's where the fireworks would likely be prettiest. Regardless, AMD and Intel, admit it: it's a compelling strategy, isn't it? At least one of you could get a leg up on the other this way. Now that I've suggested a strategy, AMD and Intel, let's see you fight over which one does it first and best.