backtop


Print 57 comment(s) - last by JumpingJack.. on Oct 12 at 10:50 PM


A young Gordon Moore circa 1975  (Source: Reuters)
Gordon Moore took part in the afternoon keynote during IDF 2007

The fall edition of the 2007 Intel Developer Forum (IDF) is officially underway from the San Francisco Moscone Center.  The Tuesday morning keynote featured more details about the Nehalem architecture as one of the main points of the discussion.

In a later session, Dr. Moira Gunn, host of NPR Tech Nation, hosted a fireside chat with Gordon Moore, Intel co-founder and creator of Moore's Law.  Moore received a well-deserved standing ovation from the crowded conference hall packed with thousands of attendees more than willing to respect a Silicon Valley legend.  

Of course, the question on everyone's mind was the validity of Moore's Law. Specifically, whether or not it holds up today the same way it did when Moore first documented his observations almost forty years ago.

Moore's Law -- actually more of a conjecture -- essentially states the number of transistors placed on an integrated circuit doubles every two years.  His observation helped outline trends the semiconductor industry for more than 40 years. 

"We have another decade, a decade and a half, before we hit something that is fairly fundamental," Moore said during the session.  That something "fundamental" is material science.  Even the most advanced lithography conceivable today can't eliminate the brick wall that is the nanoscale. 

Even at some point, lining up individual atoms no longer becomes feasible for transistor design.  Researchers from Intel are already easing into the field of using carbon nanotubes for processor interconnects; a team from the University of Pennsylvania just announced a new method for storing data via phase-changing nanowires.

"It's an exciting time," he said.  "I'd love to come back in 100 years and see what happened in the meantime."

Of course, even Moore's understanding of transistor trends is no match for the prowess of ambitious engineers. Conventional computing principles go out the window with the advent of quantum computing, for example.  Other types of alternative computing, including biological-based neural-computing, does not readily translate to transistor-count -- but that hasn't stopped researchers from making enormous progress in the last few years.

The death of Moore's Law is imminent, but new research and new materials assure that its successor will pack the same punch.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled

RE: Already dead
By KristopherKubicki (blog) on 9/19/2007 12:12:50 PM , Rating: 4
There's two conjectures to Moore's "Law" -- storage density and transistor density. I think if you look at the research chips, you'll see Moore's Law is holding up fine. The 32nm test shuttle that Intel announced yesterday was 1.9 billion transistors per chip. Then again, Moore's Law doesn't say anything about "production" level hardware.

However, you'll see that storage densities doubling every 12 months to two years is "sort of" dead. It's clear that hard drive densities certainly aren't increasing that fast. NAND "sort of" made huge progress switching from SLC to MLC -- but again that's not really production yet anyway.

The part that irks me a little is Moore's Law clearly originated around transistor count for logic. in the 1980s, those insane leaps of transistor count were all core logic. Now we have chips that are a billion or so transistors, but 800 million of which are cache.

It's hard to discount this, given that the original conjecture revolved around transistor count AND storage density, but its also easy to say the "spirit" of the conjecture is not what it used to be either.


RE: Already dead
By TomZ on 9/19/2007 12:17:57 PM , Rating: 2
Well the difference is explainable in terms of the complexity of the design effort required. Moore's law seems to mainly relate to manufacturing technology, but our ability to tackle really large engineering problems like chip design is not going to be able to grow that fast. It's just not possible.


RE: Already dead
By LogicallyGenius on 9/20/07, Rating: -1
RE: Already dead
By rdeegvainl on 9/20/2007 2:02:38 AM , Rating: 2
what does having bigger processors have to do with fitting with doubling density?


RE: Already dead
By jajig on 9/20/2007 4:45:25 AM , Rating: 2
Could you please restate that none of it makes sense to me.

If you're being sarcastic, I don't get it.


RE: Already dead
By TomZ on 9/20/2007 2:10:18 PM , Rating: 2
I think you missed my point, which is that manufacturing capability seems to be outstripping our ability to design chips that complicated. As the OP pointed out, a large portion of the transitors are RAM cache already.

4D chips - that's funny! You may as well go ahead and patent that idea.


RE: Already dead
By masher2 (blog) on 9/19/2007 12:48:48 PM , Rating: 3
> "The 32nm test shuttle that Intel announced yesterday was 1.9 billion transistors per chip"

Ah, but that's 2 years away at least. The Itanium 2 held 600M transistors all the way back in 2004. Doubling every year would mean a 2009 chip should hold 3.4 billion transistors-- over twice as many as that test shuttle. Doubling every year (Moore's original law), the chip should hold 20 billion transistors.

The original conjecture was transistor count at equal manufacturing cost. That implies roughly equal die sizes, which is why raw counts are a poor metric. Comparing the time to move to new process nodes is a much better indicator. Right now, Intel calls it 2 years...but that interval has been slowly growing for decades, and will continue to do so.


RE: Already dead
By masher2 (blog) on 9/19/2007 12:50:38 PM , Rating: 2
Edit: doubling every TWO years would mean a 3.4B transistor chip. Doubling every year would yield the 20B version.


RE: Already dead
By KristopherKubicki (blog) on 9/19/2007 12:57:01 PM , Rating: 3
I agree completely. However I think you have to look at the density function over some actual density (mm^2 or what-have-you). I'm sure several of those 32nm shuttles fit on a single Itanium die.


RE: Already dead
By masher2 (blog) on 9/19/2007 1:04:43 PM , Rating: 3
> "However I think you have to look at the density function over some actual density (mm^2 or what-have-you). "

That's what the process node gives you. Average feature size yields a linear metric for transistor density.

Of course, the process node gives you minimum feature size, not mean size, but that's exactly what Moore's Law intends. Every new nodes halves the minimum feature size, giving us the potential to double transistor counts, whether or not its actually done.


RE: Already dead
By JarredWalton on 9/19/2007 6:10:14 PM , Rating: 2
Wasn't the original "Moore's Law" statement more an analysis of manufacturing ideals as opposed to even transistor counts? I believe it was from Moore pointing out that the "sweet spot" in terms of minimum cost involved related to transistor count (taking into account manufacturing and other influences) was doubling every ~12 months, but that the rate would likely slow down. So if we call it every 18-24 months over the long haul, and the first ICs in 1965 were around 60 transistors... well, we're doing okay, but all good guesses come to an end at some point. :)


RE: Already dead
By JumpingJack on 9/20/2007 9:42:29 PM , Rating: 2
quote:
That's what the process node gives you. Average feature size yields a linear metric for transistor density.


NO.... each node 1/2 the area or double the density.... this is not rocket science.

A process node, 90 nm .. 65 nm what have you... reverse to a single dimension, it is defined in the industry as the distance that would measure 1/2 the pitch between the two closest parallel metal lines in the first metallization layer.

Again... the node choice through time are made by taking a scaling factor of 0.7 65nm is 0.7 of 90 nm, 45 nm is 0.7 of 65 nm.... that is in ONE dimension. Area is 2 dimension.

Area = width x height scale width by 0.7, and height by 0.7 so
NewArea = 0.7 x width x 0.7 x height = (0.7x0.7)width x height
or
NewArea = 0.49x(width x height) = 0.49 X Area

So each scaling to the next node 1/2 the area or doubles the density. This is Moores law...

This is too funny.

quote:
"The 32nm test shuttle that Intel announced yesterday was 1.9 billion transistors per chip"

Ah, but that's 2 years away at least.


Do you not see the irony here... first you say Moore's law is dead, then you basically state Moore's law in your rebuttal to Kris.

45 nm will launch Nov. 12th, 2 years later in 2009 32 nm will launch... Moore's Law as defined.


RE: Already dead
By Oregonian2 on 9/19/2007 1:54:30 PM , Rating: 2
The "law" as usually presented (without regard to the more historically correct conjecture not presented as a "law") isn't that trannie density doubles every two years, but the total number on the die does. So even if they use the same process, doubling the die's area would satisfy the law's continuance. Instead it's usually been a combination of the two. Its likewise a negative if dies are made smaller (which they want to do, counter to continuing the law). One of the limiting factors is 'what the heck to do with the trannies even if we can make it'. Multi-cores seems to be the answer short term.


RE: Already dead
By FITCamaro on 9/20/2007 6:54:33 AM , Rating: 2
quote:
Now we have chips that are a billion or so transistors, but 800 million of which are cache.


This is certainly true for Intel, but AMD doesn't have nearly as much cache as Intel.


RE: Already dead
By coldpower27 on 9/20/2007 9:36:38 AM , Rating: 2
AMD doesn't have chips that large though, they are at about 1/2 a Billion or so right now. Though their core logic to cache ratio will be going towards more cache with their Shanghai refresh on the 45nm. SO AMD's is slowly headed towards that.


RE: Already dead
By Eskimo on 9/20/2007 1:09:20 PM , Rating: 2
quote:
NAND "sort of" made huge progress switching from SLC to MLC -- but again that's not really production yet anyway.


Kristopher, not sure where you are getting this from. MLC has been shipping since ~2001 and currently about 80% of all NAND shipments worldwide are MLC.

DRAM and NAND are actually the drivers in the semiconductor industry for Moore's law. While micrprocessor transistor density doubling has slowed to 24 months, DRAM and NAND are averaging about 18 months and 12 months respectively for doubling.


RE: Already dead
By KristopherKubicki (blog) on 9/20/2007 11:15:47 PM , Rating: 2
Hi Eskimo,

I admittedly did not phrase that well. SLC densities are larger right now, though MLC will bring the most capacity. However, you don't see MLC SSDs or the like right now because its still prohibitively expensive. In a sense, MLC is just coming online for those type of devices.

Kristopher


RE: Already dead
By rogard on 9/21/2007 1:21:23 AM , Rating: 2
Good point, that nowadays 80% of the transistors are caches, But as far as I can remember, Moore's law did not specify the purpose of the transistors. Apart from that, in the early 1970ies...did Moore even THINK of caches, or even on-die caches, let alone the huge L1/L2/L3(!) caches we are seeing now? I doubt it.
When I started my studies of informatics/computer science in 1990, desktop CPUs used to have something like 256 BYTES of L1 instruction cache....those were the times :-)


"I'm an Internet expert too. It's all right to wire the industrial zone only, but there are many problems if other regions of the North are wired." -- North Korean Supreme Commander Kim Jong-il














botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki