The ghost of Netburst still roams at large at Intel

The chipmaker has to let go of that failure, says Tom Yager

Intel’s first server-targeted core microarchitecture CPU, the dual-core Xeon Processor 5100 (alias Woodcrest), has made its debut. A client CPU line-branded Core 2 Duo was also launched, so in servers, desktops and notebooks, Pentium 4 (alias Netburst) is officially off the roadmap.

The Pentium 4’s life began in 2001. AMD founder Jerry Sanders referred to Pentium 4 as “a step back in innovation”. Intel fired back that the Netburst architecture “has the ability to scale to 10GHz in its lifetime”. Pride goeth before the fall.

Shortly after Pentium 4’s introduction, I dubbed it “the world’s fastest microprocessor for programs that fit entirely in cache”. When the chip was working from instructions and associated data already stored in Level 2 cache and queued up in Netburst’s ocean-deep pipeline, it screamed. But when Netburst had to access main memory or conduct I/O discussions with off-chip peripherals, its shared, unidirectional bus instantly downshifted from fifth gear to first. Netburst was a jet-powered sled slogging through rush-hour traffic.

Intel knew that Pentium 4 was wallward-bound well before it shared that knowledge with us. Pentium M, released in 2003, grew out of an epiphany within Intel that Netburst was a wrong-headed design. Intel could wedge a Pentium 4 into a notebook PC, but it created a hot and heavy machine with a battery life matching a toddler’s attention span. Intel pulled together an independent engineering team and told them that “everything is on the table”. The first thing that team did was sweep Pentium 4 into the dustbin and pull Pentium III’s blueprints out of the archives. Intel owed its domination of the notebook market to a design that completely disavowed Netburst, aided by some fast-and-loose play described in AMD’s filing against Intel.

The qualities that made Pentium 4 unsuitable for notebooks made it unsuitable, period. Intel attempted to foist its amendments to core computing principles on the market by executive order. It failed. I do not patronise when I say that I’m proud of buyers for seeing through the ploy and placing Netburst among the shortest-lived CPU architectures in history.

Now it falls to buyers to maintain vigilant scepticism. While Intel seems to have come to its technological senses, it has not mended its marketecture. The “Core” brand confusingly refers to two unrelated, incompatible architectures — Core Duo and Core Solo are not 64-bit Core microarchitecture CPUs, but are fairly minor enhancements to the 32-bit Pentium M. Core microarchitecture marketing is reaching for revisionist continuity by crediting the contributions that Netburst made to Core, attempting to paint over Core architecture’s design goal of repeating none of Netburst’s mistakes. Intel also managed to lure at least two journalists to its site to perform benchmarks on unreleased, overclocked Core 2 Duo systems, the results of which were published alongside benchmarks of AMD’s Athlon 64 FX-62 , a shipping CPU that AMD sent directly to reviewers so they could test them without vendor interference. The market still has plenty of reason to remain sceptical.

And, where technology is concerned, turning back the clock, as it were, does not restore Intel’s competitive edge. The fastest notebook PC in my lab is also the lightest and coolest. It has the longest battery life and is the only notebook capable of running 64-bit Windows XP, 2003 Server and other 64-bit OSes. Core microarchitecture? No. AMD Turion 64.

Join the newsletter!

Error: Please check your email address.

Tags technologynetburstintel

More about AMDIntel

Show Comments

Market Place

[]