Paul DeMone (pdemone@igs.net) on 6/18/03 wrote:
---------------------------
>Bill Todd (billtodd@metrocast.net) on 6/18/03 wrote:
>---------------------------
>[...]
>>And I never suggested in the least that it did. Though it is interesting to note
>>that it doesn't seem to clock all that much *faster* than 1.5 GHz even if you raise
>>the voltage all the way up to 1.5v (where they'd seem hard-pressed to get more than
>>about 1.6 GHz out of it with a decent margin of safety, if I understand the chart correctly).
>
>Intel is free to chose the sample it schmoo'd from anywhere in the process. Can
>you think of good strategic reasons for and against choosing parts in the slow vs
>typical vs fast processing range for public disclosure prior to product introduction?
>Is Intel more interested in exciting customers who rarely get excited about clock
>rates anyway or avoiding tipping its hand to its competitors?

So when you don't care for what *Intel* says about its own product you implicitly try to discredit it? One might almost think you're getting a bit defensive about some of your earlier statements that just don't seem to be supported by the emerging evidence.

>
>>
>>However, that interesting .pdf (thanks for the pointer) does support my understanding
>>that Madison indeed consumes 130 W at 1.3v (and 1.5 GHz).
>
>Explain how.

It's not all that hard to devine had you bothered to read the cited reference, Paul: it's right there in black and white on page 5. Is there anything else you'd like to have me read for you?

>
>> So how the power consumption
>>could be halved by reducing the voltage only 15% and the frequency only 13% (to
>>1.1v and 1.3 GHz, as you suggest below) remains unclear to me (though I don't have
>>any problem believing that a 1 GHz part - possibly at even lower voltage - might
>>halve the power requirement).
>
>If you start with faulty assumptions things only go down hill.

A statement you might want to take to heart yourself. Meanwhile, please identify any faulty assumptions I might have made.

A shmoo plot tells
>you absolutely nothing about power

Duh. Since my reference was to the .pdf rather than specifically to the schmoo plot, that observation seems rather pointless.

and is only a single sample anecdotal hint
>about the frequency yield. We don't know anything about which stepping this is
>compared to the production device. Similar faulty assumptions were the basis
>of apparent good sounding arguments in the past why Willamette would never
>scale above 1.7 GHz and why Northwood would never get above 2.5 GHz.

That's right. When it really comes right down to it, we just don't *know* anything (well, it appears that this was an ISSCC 2003 presentation, for what that may be worth). For all I know you could be Rob Young masquerading as someone more competent while talking up Itanic: you certainly sound that way right now.

- bill