HN Explorer

Poll: Will this recession affect Moore's Law?

7 points by geuis ↗ HN
A posting earlier about slowing desktop sales made me think. If the recession gets much worse, or if it's already that bad and my head is in the sand, could this slow down Moore's Law?

We like to think ML is a fact of nature, but it's entirely a human construct. If the big companies like Intel cut their R&D or production schedules it could slow down the predicted doubling of processor transistor count.

14 comments

[ 1.8 ms ] story [ 61.4 ms ] thread
tokenadult
I don't think the recession will affect Moore's law.
lacker
The bigger story is how Moore's law has switched from "faster and faster" to "more and more cores". Many people (for example Rich Hickey the guy behind Clojure) point to this as evidence that parallel programming will become more important soon.
madcaptenor
I don't know. Did previous recessions affect Moore's Law?
TrevorJ
No, there will always be some facet of the world that will apply pressure to continue progress on new technology, be it corporate, private or defense. I do think however, that the possibility exists for the available CONSUMER products to slow in their development cycles somewhat.
russell
Faster and faster was never part of Moore's law. It was all about transistor density. Faster and faster stopped in part because of heat problems. Feature size will continue to get smaller for a while, but the schedule will be subject to economic conditions. A new state of the art fab costs in the billions and the price goes up as the feature size goes down. At some point feature size reduction will seriously slow down because of physical constraints.

Intel and others switched from bigger CPUs to more cores because the were hitting the wall in design complexity of the pipelines. More cores is a more efficient use of the real estate.

There are still tricks to be tried like 3D chip designs.

geuis
No one said anything about speed. Moore's Law has nothing to do with transistor speed, only with transistor density.
ramchip
We like to think ML is a fact of nature, but it's entirely a human construct.

It's a nitpick, but I get a bit annoyed when people speak for 'us' and then give their personal opinion. Moore's Law is just an estimation of progress speed that happens to fit fairly well so far, and I haven't seen anyone taking it like a fact of nature.

joshu
I wonder if you could rephrase it: Transistor density as a function of dollars spent on fabs.
skmurphy
A new level of density requires invention and experimentation. You probably can't do much more than double density without actually building a prototype. It's also a function of cumulative spending at an industry level as no one is vertically integrated anymore, not even IBM or Intel: fab and test equipment separated from design from fab operator from EDA software etc... many layers.
anamax
Which Moore's Law? The one about the cost of a fab increasing pretty much in proportion to feature size shrinks implies that we'd eventually spend the world's GDP on a single plant. It's hard to see how that wouldn't be affected by a recession.
xiaoma
Please forgive my ignorance about this, but how has Moore's Law been doing even in this decade? I realize the law was about transistor count rather than speed. However the subjective performance I get for the money doesn't feel like it's increased sixteen-fold since 2001.

Has the transistor count been doubling every other year? If so, what does that mean from the perspective of the user if performance isn't also improving at a similar rate?

skmurphy
Look at memory chips and FPGAs.
crpatino
Exponential growth cannot go on for ever; at least not in an Universe with hard limits, like our own. My personal opinion is that in Nature, every time you find data with an exponential growth trend, it means you have found the slope of a sigmoid trend and you have not observed long enough.

Moore's Law is nothing but a myth now. In its original form, it was a forecast for the next 10 years... in the sixties!!!! Now we have people talking about infinite growth based in a prediction expired before they were born! Not even its creator thinks it's valid anymore. http://www.techworld.com/opsys/news/index.cfm?newsid=3477