| Author |
Message |
David C. DiNucci
Guest
|
 Posted:
Tue Oct 04, 2005 12:15 am Post subject:
Not Moore's Law = House's Law? |
|
|
"Not Moore's Law" may be cleared up in an article on Merc's website
today about Computer History Museum talk on Thursday. The gist is that
Moore never predicted anything doubling every 18 months--originally
predicted number of transistors doubling every year, then changed to
doubling every two years--and that it was a marketing exec at Intel,
Dave House, who noted that computing performance was doubling every 18
months.
-Dave |
|
| Back to top |
|
 |
Anton Ertl
Guest
|
| Posted:
Tue Oct 04, 2005 12:15 am Post subject:
Re: Not Moore's Law = House's Law? |
|
|
"David C. DiNucci" <dave@elepar.com> writes:
| Quote: | "Not Moore's Law" may be cleared up in an article on Merc's website
today about Computer History Museum talk on Thursday. The gist is that
Moore never predicted anything doubling every 18 months--originally
predicted number of transistors doubling every year, then changed to
doubling every two years
|
According to
<http://research.microsoft.com/users/Gray/Moore_Law.html>, Moore
predicted 18 months per doubling of the number of transistors per IC
for ten years in 1975 (after the 12-month prediction for ten years in
1965), and reaffirmed this for another ten years in 1995. The
bibliography does not contain the '75 paper, however; it does contain
a reference to a paper by Moore in 1995.
In <2002Dec22.163332@a0.complang.tuwien.ac.at> I computed the
price/bit of DRAM, and got an average halving in 17 months, pretty
close to Moore's prediction (if we assume that the price/chip stayed
the same and that the transistors/bit have not changed much).
| Quote: | --and that it was a marketing exec at Intel,
Dave House, who noted that computing performance was doubling every 18
months.
|
That seems somewhat extravagant, but of course, one might define
"computing performance" to fit this claim (e.g., by counting the
number of independent FLOPS that can be performed by all processors in
the world). In any case, in
<2003Sep27.120214@a0.complang.tuwien.ac.at> I looked at a more
conservative version of not-Moore's law.
- anton
--
M. Anton Ertl Some things have to be seen to be believed
anton@mips.complang.tuwien.ac.at Most things have to be believed to be seen
http://www.complang.tuwien.ac.at/anton/home.html |
|
| Back to top |
|
|
David C. DiNucci
Guest
|
| Posted:
Tue Oct 04, 2005 12:15 am Post subject:
Re: Not Moore's Law = House's Law? |
|
|
Anton Ertl wrote:
| Quote: | "David C. DiNucci" <dave@elepar.com> writes:
"Not Moore's Law" may be cleared up in an article on Merc's website
today...
According to
http://research.microsoft.com/users/Gray/Moore_Law.html>, Moore
predicted 18 months per doubling of the number of transistors per IC
for ten years in 1975 (after the 12-month prediction for ten years in
1965), and reaffirmed this for another ten years in 1995. The
bibliography does not contain the '75 paper, however; it does contain
a reference to a paper by Moore in 1995.
|
Well, then, let me revise that intro:
"Not Moore's Law" may have been thrown into even more confusion in an
article on Merc's website today...
http://www.siliconvalley.com/mld/siliconvalley/12806765.htm
So nobody at the talk asked him of the apparent discrepancy between his
current statements and his published papers? Or was the Merc (or I)
simply misreporting his current statements? (Eugene, were you there?)
-Dave |
|
| Back to top |
|
|
Brig Campbell
Guest
|
|
| Back to top |
|
|
rhnlogic
Guest
|
| Posted:
Tue Oct 04, 2005 5:55 am Post subject:
Re: Not Moore's Law = House's Law? |
|
|
Anton Ertl wrote:
| Quote: | "David C. DiNucci" <dave@elepar.com> writes:
"Not Moore's Law" may be cleared up in an article on Merc's website
today about Computer History Museum talk on Thursday. The gist is that
Moore never predicted anything doubling every 18 months--originally
predicted number of transistors doubling every year, then changed to
doubling every two years
According to
http://research.microsoft.com/users/Gray/Moore_Law.html>, Moore
predicted 18 months per doubling of the number of transistors per IC
for ten years in 1975 (after the 12-month prediction for ten years in
1965), and reaffirmed this for another ten years in 1995.
|
I attended the Computer History Museum talk, and heard Gordon Moore
mention only his one year and two year transistor doubling predictions.
He did mention other time periods in relation to testing his doubling
predictions: after the first 10 years, he was off by one doubling;
and that sometime after the later two year doubling estimate he
measuring a doubling rate closer to 21 months.
Only Dave House mentioned the 18 month number in relation to CPU
performance, and Gordon didn't try to correct him.
Of course, neither my hearing nor my memory are perfect. YMMV.
--
Ron
http://www.nicholson.com/rhn/home.html |
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in