Intel's Timna has dead duck look'n'feel
The system on a chip deal that's a canard
Behind the scenes at the Computex show a fortnight ago, some folk were shown Intel's not-so-up-and-coming Timna system on a chip device.
Gigabyte was less cautious, and showed a Timna motherboard in its suite in the Hyatt, although a suit brought his hand up in front of our digital camera just before we caught it for posterity.
Because the word on the Computex street is that Timna, now officially delayed until Q1 next year, is a duck that may not hatch, never mind get round to flying.
And, those observers point out, there's more than one reason why Intel may well draw a veil over the whole Timna exercise.
First, competitive forces from both inside and outside Intel militate against Timna, in the shape of AMD's Duron and Via's bastard-creation, the Cyrix III. Although the latter is currently vastly overpriced, most think that if Via continues to hawk this CPU, which has, of course, an IDT WinChip core, rather than a Cyrix core, it will be sold into cheapo-cheapo markets at prices closer to red cents than to greenbacks.
The Timna Duck also competes against Intel's traditional Celeron family inside. Pressure from AMD in the shape of the Duron will result in Intel continuing to use its marketing muscle to drive down the price of Celerons and increase the clock speed, as roadmaps we have seen already demonstrate.
However, there are also grave technical issues with the Timna microprocessor which mean that it may never be served up as Chipzilla's duck l'orange.
As Gigabyte's mobo clearly showed, Timna arrives in an S-370S configuration, and sources close to Intel's plans say that engineers are finding it extremely hard to build the dense mass of graphics and audio transistors into a package that is both cheap and efficient.
There are more serious technical problems for Timna too, and those are related to the now infamous memory translator hub (MTH) technology that has caused Intel so much grief over the last 18 months.
Documents shown to us by an engineer close to Intel's plans in Taiwan, reveal that the firm was told by a third party consultancy as long ago as autumn last year that building the MTH technology into chipsets was technically unviable and would inevitably have a serious effect on performance. So it proved.
Nevertheless, and despite these clear technical difficulties, Intel, for reasons which must be to do with the chipocracy currently dogging the firm, decided to ignore the warnings and press ahead with the plans.
The technical difficulties to overcome with the MTH clearly remain, according to the documents we saw, but Intel, on the face of it at least, appears to want to ignore their conclusions. The most recent roadmaps we have seen show not only Timna slated for Q1 next year, but, later in the year, a Timna Plus.
Don't think that Intel cares that major mobo customers like Gigabyte may have pressed ahead with R&D on Timna developments - many different Taiwanese firms we talked to were better about the experience they had had not only with the i820 chipset, but with the i840, Willamette's Tehama, and the i810 and i815 too. Those bitter experiences, as we have pointed out here previously, have simply pushed them into alternative chipset camps, including those of ServerWorks and Via, and to experiment ever more intensively with the Leader of the Chip Opposition, AMD.
It is entirely possible that Intel may press ahead with its Timna microprocessor but common sense, now seemingly in short supply in some divisions of the firm, suggests that rather than hatch this candidate for duckdom, the engineers should just boil it lightly for three minutes, put some rock salt on the yolk, and consume it.
If the Timna Duck hatches, it may well just cheep softly and keel over before it even gets a chance to quack its quack. ®