"Max Eskin" <maxeskin_at_hotmail.com> wrote:
> tried. For example, it would be possible to place transistors at
> 1 per cubic inch, and have the cool adequately, but it's clearly
> never been attempted, because if it were, there would be boxes
> to plug into a PC XT to get Pentium II speeds.
What on earth makes you think that you could build an x86 CPU with
PII performance out of discrete transistors? It would be very difficult
to get even the performance of a 286 out of discrete transistors.
The discrete stuff tends to be *MUCH* slower than ICs.
For reference, the IBM 7030 (Stretch) had one of the most complex CPUs
ever constructed out of discrete transistors; it contained about 169,000
of them. Mind you, that's *only* the CPU; it doesn't include the memory
controller, Exchange (I/O channels), Disk Synchronizer, or anything
else. That's within an order of magnitude of most of the microprocessors
used in today's desktop computers, and even closer if you make it a fair
comparison by not counting the transistors that implement the on-chip
cache memory.
In fact, that's more transistors than a 286 CPU had (*), but AFAIK
the 286 had several times more raw CPU performance than the Stretch.
The overall packaging density of the Stretch CPU was approximately
0.06 transistors per cubic inch. The power supplies, fans, mounting
frames, backplanes, cables, cabinets, etc. take up a lot more room
than the transistors.
Cheers,
Eric
(*) Intel's web site claims that the 286 had 134,000 transistors:
http://www.intel.com/pressroom/archive/backgrnd/CN71898A.HTM
They also make the completely ridiculous and untrue claim that the 286
was the first microprocessor that offered software compatability with its
predecessors. To the best of my knowledge, the Intel 4040 was the first
microprocessor to do so, and many others did this as well before the 286
was introduced, such as the Z-80, 8085 and 6802.
Perhaps next year the government will declassify the 2nd-generation version
of the F-14 computer, so that we can argue about whether it was the first
backward-compatible microprocessor. :-)
Received on Mon Oct 19 1998 - 00:39:38 BST