De mannen van Sharky Extreme hebben wat interessante info los weten te peuteren van 'sources close to Intel'. Het blijkt dat de Pentium 4 niet gedelayed wordt door marketing technische redenen zoals eerder door The Register gesuggereerd werd, maar vanwege de ICH2 chip die een klein probleem bleek te hebben zodra hij werd gekoppeld aan de i850 of i860:
The Pentium 4 (see our preview right here) is now targeted to launch in week 48 of this year. If our math is correct, that should put the launch around the end of November or beginning of December. This is a bit later than we previously expected, but we now know why the product will be delayed.
It turns out that ICH2 has an issue that only manifests with the i850 and i860 MCHs. Certain PCI cards run into a problem with certain commands that pre-fetch data from memory. Sometimes ICH2 will not invalidate pre-fetched data when it should. The issue is due to the interaction between the MCH and ICH2. From what we hear, Intel fully understands the problem, has solved it, and is doing a "low risk stepping" of ICH2 with a minor metal layer fix in order to get rid of the problem. Once again, this problem only occurs when ICH2 is matched to the i850 and i860 MCHs. Current ICH2 will not encounter this problem.
Named B1 prime, the new samples are going into production in week 41 and should be available in week 42. Intel's customers are being told to continue their validation testing using the B1 stepping of ICH2, and that B1 prime will not require any BIOS or motherboard design changes.[break]Sharky Extreme heeft ook nog wat achtergrond informatie over het verdwijnen van de Timna CPU weten te bemachtigen. Intel heeft ingezien dat het product te laat op de markt zou komen, en dat het daarom overbodig is geworden:[/break]Our sources say Intel has said that their manufacturing scale as well as cost reductions in the general PC industry have, among other things, driven down the overall value platform BOM (bill of materials, a.k.a. the part count and cost). The lowering part counts and therefore costs that high-integration was intended to yield have been seen without Timna. This means that, if Intel were to push Timna, they would be pushing an unneeded part at costs similar to the regular parts.