Home / Opinions / Testing Times

Testing Times

by

Some time before Christmas Intel held a two hour briefing about benchmarking which was run by Shervin Kheradpir, Director of Intel's Performance, Benchmarking and Competitive Analysis Group.

I’ve been to plenty of briefings for new products and they are unremittingly upbeat with PowerPoint slide after PowerPoint slide telling you how great the New Shiny Thing is going to be. You generally get a fair dose of technical background that helps you to understand the key features of The New Thing but you can usually guarantee that some two thirds of the presentation will be a complete waste of time.

The Intel briefing was unusual because it wasn’t related to a specific product so it instead followed the form that you get during an internal company meeting when you want to give the sales force a gee-up. First they set the scene, then they knock you down by telling you how rubbish your performance is, and then they give you a direction for the future and send you on your way singing the company song, determined to be a better person in future.

The first hour of the briefing was spent telling us that game time demos and benchmarks such as 3DMark are a very bad tool for testing processors. This came as a surprise as I have never read a processor review that stated that the new Pentium Whatever is a waste of time as it doesn’t increase your Quake Time demo frame rate. Intel’s point is that time demos don’t stress the processor as they don’t use the game’s Artificial Intelligence but instead stress the graphics and memory I/O sub-systems.

Intel also showed a plethora of charts and graphs that compared two unspecified PCs showing frame rates recorded with FRAPS. The axes of the graphs were vague in the extreme and the only conclusion we could draw was that System A had a period of zero fps that lasted for eleven seconds and was therefore worse than System B, which kept up a steady frame rate throughout. This suggested that System A had shut down for a time while it updated its anti-virus software.

The inferences were clear; though the integrated memory controller on the Athlon 64 helps it to kick the living daylights out of Pentium 4/Pentium D and therefore will turn in a better 3DMark score than an Intel PC, there were far greater benefits to be had by using a dual-core Intel processor.

All true, but we already knew this. And Intel must surely know that we know, but it served to paint a gloomy picture of the current state of affairs and set us up for the second part of the briefing - the Big Whoop.

Intel is desperate to shake off the old, outdated concept that clock cycles matter, and instead we should look at what a processor can actually do for the user. Better yet let’s think of it as a platform, and hey, let’s call it Viiv.

Despite the hype, Viiv is nothing more than a processor and chipset so while you can keep to the Viiv specs and build a new Media Center with a Core Duo and an Intel 945G motherboard with integrated graphics you can also take the non-Viiv approach and build one using a Pentium Extreme Edition on an i975X motherboard with four cores hammering away behind a PCI Express graphics card. One approach uses a small amount of power and generates minimal heat and noise while the second approach is rather less power frugal. This presents a problem as the Extreme Edition/i975X will trounce the benchmark figures of a Core Duo yet the latter system will likely turn out to be a more practical proposition for most living rooms.

comments powered by Disqus