Yousef, link back to the post where I said "Since power consumption generally increases as the square of frequency"
You will find the following immediately after that:
As for Intel, they are planning a 300 MHz Deschutes notebook chip,
which must have < 10w dissipation or it couldn't be a notebook chip.
Since power consumption generally increases as the square of
frequency, this means they could get to 424 MHz with < 20w
dissipation.
Clearly, in this context, I was comparing a higher Vcc (and correspondingly higher Idsat) desktop chip to the 300 MHz Deschutes notebook chip. Since both AMD and Intel have produced notebook chips as their first 0.25 um products, we can to some extent forecast the maximum operating frequency of the future desktop chips based on the 0.25 um process.
The rest of your post is also highly misleading:
The reason that the AMD process produces smaller die is due to tighter design rules (spacings between poly lines and contacts, contact enclosures ...).
These design rules only affect density and not FET size/performance.
In going from the 0.35 process to the 0.25 process, the transistors per square mm increased by a factor of 2.38. AMD already had tighter design rules and higher transistor density then Intel in the 0.35 process. From what I've read, Intel's transistor density increased by only a 1.5 factor or a 1.26 factor (depending on whose estimate of Tillamook die size you use) going from 0.35 to 0.25 um. The big improvement in AMD transistor density could not be explained by even a proportional 0.25/0.35 reduction in FET dimensions and in design rule spacings -- also, the higher operating freq of the AMD notebook chip (initial intro at 233 vs 200 for Tillamook, projected max clock speed of 300, vs 266 for Tillamook) argue that FET dimensions were reduced by *more* than a 0.25/0.35 ratio.
BTW, AMD's .25um process is optimized for 2.2V supply voltages (as reported in the literature)
Sorry, the desktop 266 and 300 use a 2.1v supply on my Shuttle 603 motherboard. 2.2v is not even available.
Notebook K6's in .25um process at 1.8V will not be running at 300mhz. They will be lucky to hit 233mhz.
Apparently, Compaq bought a truckful of these "lucky" 233mhz!
Make it Better,
Petz |
