Around the beginning of the last decade, Intel had the most efficient processor and was trying to argue that competing designs—which had faster clock speeds but did less actual work—were in fact inferior parts. They tried to get the industry to abandon MHz as a measure of performance and failed. At the end of the decade, Apple (Motorola) and AMD had more efficient processors and tried to get the industry to move away from MHz. They, too, failed. Now, the pendulum has started to swing back, and we have another chance to move away from this antiquated measure of performance.
What is driving this change from Intels side isnt so much competition with AMDs and Apples processors, which clearly are not running at higher clock speeds, but the increasing enticement of Intels lower-cost parts. These low-cost parts—Pentium 4s without HyperThreading and Celerons—have similar clock speeds to Intels high-end offerings but do substantially less work. In other words, they are much less efficient. An interesting way to look at this is that Intel is competing with itself and losing.
The perpetual problem is that if people are locked into MHz, Intel has no easy way to show the relative value between a high-cost Pentium 4 Extreme Edition with HyperThreading (one of the few product names that is almost a sentence) at 3.1GHz, which sells for around $1,000, and a 3GHz Celeron, which sells for a fraction of that. Realize there are a number of variables that make up the Extreme Edition part: massive cache, the HyperThreading component and much faster front-side bus speed. However, few people who buy computers have the skills (or want to have the skills, for that matter) needed to understand the differences these technologies bring.
To address this, Intel is attempting to apply a number to their products so that a buyer can determine which is better. Now, this isnt the industry standard that we really need, but it is at least an admission that MHz isnt an industry standard—which opens the door for firms like Futuremark, which make benchmarks, to fill this gap.
To understand the new numbers, you first need to realize that mobile and desktop can be confusing if you compare one with the other. The simple advice is, dont. Youll know which type of product you want early on, and your comparisons will be between the different notebooks and desktops. You will probably never actually compare a notebook with a desktop.
Next page: How to group product lines.
How to group product
lines”> For the desktop lines, there are three groupings. The 300 series is for the low-end Celeron line, the 500 for the mainstream Pentium 4 with HyperThreading line and the 700 for the Pentium 4 Extreme Edition parts, which are targeted at power gamers.
On laptops, like desktops, the 300 is the low-end (Celeron) line. However, 500 designates the P4 M (M for mobile) line, which is targeted at desktop-replacement station-to-station users. The 700 line encompasses the new Pentium M, which is targeted for folks who are highly mobile.
A station-to-station user is one whos always plugged in and simply relocates a lot. Such a user therefore needs something more portable than a typical desktop machine, but he or she doesnt really travel all that much. Someone who works from home two days and is in the office three days would be an ideal user for a desktop replacement box.
Going forward, the relative capability of each part will increment based on changes to a number of factors. If there is a significant—as defined by Intel—change in the instructions set (such as SSE2), capability (such as HyperThreading), hardware (such as increased cache) or speed (MHz is still in there somewhere), the processor will be incremented. That means a 725 will be better than a 720. If both are at the same price, the 725 will be a better deal. Trying to figure out how much better will be an exercise in frustration. If you are setting a spec, set minimums to establish class, and then move on. If you try to become too granular, you will probably just confuse the situation, and the vendors will be driven to the newer processors anyway as Intel moves from processor to processor.
Over the next few weeks, vendors will be working on how to respond to this Intel move. You can, at least for now, ignore this part-number change, as MHz will still be around, and you can build that into a spec. However, given that the real measure of value is likely system performance (vendors have been putting fast chips in slow systems for years), maybe its time to toss out vendor-sourced metrics and set a specification—as the military often does—off independent benchmarks or ones you do yourself. Granted, benchmarks are harder to compile than processor speeds or arbitrary numbers, but they generally help you get the best actual deal, if done properly. They also help you work through what product you actually need. Being able to demonstrate that you are doing the most with the funds for which you are responsible is generally considered to be a good career move and one you should consider here.
While the industry still needs a metric similar to horsepower, in the end this move by Intel both increases the chance that we will have one and further exemplifies the need for a solid benchmarking process that includes all of the features, since we really do need to know that we are getting what we paid for.
Rob Enderle is the principal analyst for the Enderle Group, a company specializing in emerging personal technology.