Twenty years of Pentium 4 – an occasion to remember a processor and an architecture, at first much acclaimed, then later in the third generation Prescott much laughed at and ridiculed and finally mothballed in generation four (Cedar Mill). Intel had never seen anything like this before. How could this happen? A look back with a few personal experiences.
When, in the spring of 1999, John Barton, then head of testing and validation at Intel in Hillsboro (today Vice of the Digital Enterprise Group) at an event in Munich, carelessly invited us to come and see Hillsboro, he couldn’t suspect that one of the participants will take him at his word. Most of the time it’s just a nice phrase; but then I punched holes in my Intel PR Heiner Genzken, who actually made it possible for me to travel from tranquil Palm Springs to Portland / Hillsboro following the autumn IDF 1999.
Andreas Stiller, the longest-serving editor in the c’t and heise online editorial team, deals with processors, high performance computing, hardware-related programming, HPC programming and exciting scientific topics such as gravitational waves, CERN, etc. Even in well-deserved retirement, the He started at the end of 2017, of course, he can’t leave these topics.
–
–
–
–
“Real close contact”
So I sat there in the large conference room across from more than a dozen Intelians, who initially revealed to me that no journalist had ever been in their sacred halls. What I hadn’t really considered beforehand: There were top-secret systems everywhere in these halls, so the green light had to be given from the very top. Everywhere, black cloths were carefully wrapped over the special test boards, all kinds of pictures, posters, memos, etc. on the walls were also covered or hung – after all, the spy from Germany was evidently respected. Years later, an Intel employee told me how difficult it was to find enough black cloths in Portland and how much effort it had taken to get the halls journalist-proof. Photography was of course forbidden …
Sure, the wrapped test boards and systems were equipped with the Willamette, named after the great river through Portland. Willamette, that was the code name for the Pentium 4, which was officially released on November 20, 2000 over a year later. After all, Intel had already revealed its name in June (well, after Pentium III …). This time Intel also made sure that c’t got a test system in advance and that we didn’t have to run our benchmarks in a night-and-fog action at a CeBIT stand, as was the case with the Pentium P5, much to the displeasure of Intel. The online article with reference to our test was even a matter for the boss at c’t – but I couldn’t help but post an article on the bugs of the Pentium 4 on the same day, which of course also included someone with the great A20 -Gate was there.
The Pentium 4 Willamette was developed under the direction of Glenn Hinton, and I was very lucky to meet him a few times before and after the launch on the microprocessor forums, even sat with him when AMD’s Fred Weber introduced the sledgehammer. Hinton had received a clear order from the management to develop a processor for an “Industry leading clock rate”; He and his team had done that brilliantly. Intel was in a fierce GHz competition with AMD: GHz were the measure of all things back then, no matter what the real performance was. And so some processors shot up rather immature, for example the Pentium III with 1.13 GHz in the spring of 2000. Stupid, it was still quite unstable, it cost Intel’s reputation and the head of the microprocessor division Albert Yu the job.
– As high a cycle as possible, that was the target, and for that you needed a super pipeline with 20 or later 31 cycles
(Image: intel)
– .
