January 04, 2001, 11:31 AM — Unfortunately, this is not a hands-on column. I haven't had the opportunity to try one of the new Pentium 4 chips. The chip, announced with the usual fanfare amidst an ongoing ad campaign for the Pentium III, is a 1.5-GHz speed demon. News flash: for most of the world, the speed will be meaningless. Why?
If you've followed my columns on ITworld.com, you know my thoughts on this matter. Intel and AMD keep eliminating the bottleneck that doesn't exist for the great majority of the computers out there. Most of us are limited by disk speed, network speed, Internet pipe speed, and so on; the CPU has long ago stopped being the bottleneck. It's the breakaway champion of the speed demons. When will the rest of the system catch up?
Since I've last addressed this issue, there has been little public notice of any significant enhancements in the biggest bottleneck: the disk subsystem. Sure, disk rotational speed has not "jumped" from 10K to 15K rpm. And guess what? As I stated months ago, that kind of speed increase will only provide marginal improvements in disk speed. Based on the numbers posted on one disk vendor's Website, a 15K rpm disk can be expected to have an average read time of about 6 ms. This is accomplished by moving the disk faster (15K rpm becomes 250 revolutions per second or a 4 ms cycle, making the average latency 2 ms) and by going to a high speed SCSI (160 Mbs) interface. Compare this to a 10K rpm disk with the same interface and the speed difference is 1 ms. Compare it to a 7,200 rpm drive and the difference is a "staggering" 2 ms.
All of this is impressive, but it still does not get us to the sub-millisecond, large-capacity, nonvolatile storage that we need.
When will mass storage catch up? The answer is still unknown.
Nonvolatile RAM drives are an interim solution at best. As the volume of data stored increases, the cost of RAM drives quickly becomes prohibitive. I know of one large company that installed a RAM drive to ease its system response time problems. This step got the company through a crunch, but now it's looking for other solutions: the gains afforded by the RAM drive will become insignificant when the company database grows to 10 times its current size.
It is comforting, though, to see other authors pointing out the lack of a need for a 1.5-GHz chip. Maybe we are maturing -- it's not the horsepower of the engine that counts but how fast the car can get around the track. For that you need the engine, yes, but you also need steering, suspension, tires, fuel tank. The entire system must move forward. Oh, well. I guess we'll just keep waiting.