Could Moore's law be purposely used for marketing purposes?

[jasong]

Check this out. The following is a list of past and future nanometre lengths for Intel and AMD:

130
90
65
45
32

Now check this out:130^2=16900, 90^2=8100, 65^2=4225, 45^2=2025, 32^2=1024. See the pattern? For each subsequent iteration, you can stuff approximately twice as many transistors in the same area. The thing is, if the pattern isn't forced in some way, doesn't it seem odd that each new fabrication allows close to twice as much stuff to be stuffed in? This leads me to believe that the two companies are intentionally forcing things to follow moore's law, possibly as a way to make more money.

Comments?

[TimSorbet]

http://en.wikipedia.org/wiki/Image:M...%282004%29.png
http://en.wikipedia.org/wiki/Moore%2...ndustry_driver
Of course they're trying to follow Moore's Law, is that a bad thing? It is a high goal to meet, so if they aim for it, computers (and therefore prime crunching ) will get faster than if they aren't following any rules and are just making chips without an actual goal.

[S485122]

Originally Posted by Mini-Geek

Of course they're trying to follow Moore's Law, is that a bad thing? It is a high goal to meet, so if they aim for it, computers (and therefore prime crunching ) will get faster than if they aren't following any rules and are just making chips without an actual goal.

The fact that the number of transistors doubles at each size decrease is caused by the huge cost a change in size implies : a whole new factory. It is not worthwhile to change for every percent in size change possible.

Only two communities of users want all that extra power : the gamers and the number crunchers. All other users do not profit from these newer machines, they have to buy costly new material every two or three years because they have bought costly updates to their programs, but overall their use of the computers does not change much : send and receive mail, chat, browse, a bit of word-processing, "spreadsheeting", viewing pictures and listening to music. All those tasks can be done as well with the hardware and software of 5 or even 10 years ago. Of course the new versions of programs add a lot of costly noise : interface that obfuscate the controls, distract from the main task of the program... Does anybody look at the "visualisation" of Windows Media Player while listening to music ?

With Vista most of the contemporary hardware becomes obsolete overnight. But was all power necessary for the most wanted improvement : security, I think not.

Since I am not a gamer I can not speak about their real needs, but I think that a lot of the pleasure is just in improving the hardware as a goal in it self.

This all corresponds to our consumerist society which is using the earth resources at an unsustainable rate and implies that poorer countries just stay poor so that we (from the rich countries) can go on. It also ignores tour responsibility to the next generations.

In conclusion part of Moore’s law is our greater technological ability, the other part is marketing.

[masser]

here's another link on the subject....

http://en.wikipedia.org/wiki/Planned_obsolescence

[TimSorbet]

Originally Posted by S485122

that poorer countries just stay poor so that we (from the rich countries) can go on.

Yes, we're making computers faster at an exponential rate, but that means that computers the speed of the ones we (the rich countries) bought a few years ago are now extremely cheap, so cheap that things such as the OLPC are possible.

[ewmayer]

Originally Posted by Mini-Geek

Yes, we're making computers faster at an exponential rate

Correct me if I'm missing something, but wouldn't Moore's "law" imply at best a geometric rate? (A fine distinction which is lost on most jounalists and Hollywood scriptwriters, who love to carelessly bandy about the term "exponential.")

My reply to the question asked by the thread title: what, you think it isn't already? By way of example, much of Intel's marketing (and related ad copy, e.g. the ad for the NASDAQ exchange a couple years back that used ML to predict that Intel's chips a few years down the road would execute "2 billion operations per second.")

In fact, with respect to Intel, it's an interesting cart-and-horse (or chicken-and-egg, or armadillo-and-road-crossing - take your pick) question as to what lengths Intel goes to to *make sure* their products meet periodic ML-based "performance" milestones *because* that is a crucial component of Intel's marketing strategy. For example, unlike Intel, most RISC-style chip manufacturers chose a path of increasing performance by being able to do more work per clock cycle, and only to a lesser extent by jacking up the clock rate of their chips. (Doing more work per cycle means throwing more transistors into the chip, which at a given process size increases its size and thus makes it harder to meet clock-rate targets.) Intel has instead settled for a relatively mediocre per-cycle performance curve but done a spectacular job at continually jacking up the raw operation frequency of their chips, apparently betting that raw MHz (or now GHz) is something their target audience grasps more readily than less-obvious performance measures such as "pipelined floating point operations per cycle". On that score, I'd say Intel was dead on the money.

[S485122]

Originally Posted by Mini-Geek

Yes, we're making computers faster at an exponential rate, but that means that computers the speed of the ones we (the rich countries) bought a few years ago are now extremely cheap, so cheap that things such as the OLPC are possible.

In my opinion it is not because we now can make much faster chips that the old ones have become cheap. Of course old investments have repayed themselves and do not need to be payed for by the consumers. It is also because the manufacturing price of top end products has little to do with the end price.

This is most obvious in the pharmatical industry where the lifecycle of drugs is : 1 one sell for a very high price to a very few, 2 sell for a high price to a few, 3 sell for a lower price to still more people, 4 sell for a low price to a lot of people, 5 stop selling the drug and create a new one. Ideally there should be one drug in each stage. It is even so that a company has refused to go public with new AIDS drugs, because the previous one still had some time to go on its marketting life-cycle.

[brunoparga]

Originally Posted by S485122

This all corresponds to our consumerist society which is using the earth resources at an unsustainable rate and implies that poorer countries just stay poor so that we (from the rich countries) can go on. It also ignores tour responsibility to the next generations.

(Just an off-topic comment, couldn't help it) This is why I fully support VHEMT. That's a nice way to deal with the next generations...

[S485122]

Originally Posted by brunoparga

(Just an off-topic comment, couldn't help it) This is why I fully support VHEMT. That's a nice way to deal with the next generations...

The url is Voluntary Human Extinction Movement.

[axn]

And the protocol is http (http://www.vhemt.org)

[S485122]

For once I retyped, used the tags, put a title. Instead of just being lazy and cutting and pasting the url from my browser. :-(

Thanks for the correction !