Considerations about "Moore's Law"

[shadow_2609]

Hi everyone!
I wanted to open this topic because... you know, Moore's Law! I don't get it!
Other reasons because I wanted to open this topic is because I would like to know the prospective of everyone (or almost):
Let's start by saying... Why it has to be a law? Mr. Moore (hats off), just made a prediction, NOT  a rule to follow or otherwise you will sent to the last level of the hell where you walks bare feet on upside down chips; here it is my "consideration":
First of all, why it is not a law? I answer this thing mentioning real LAWS like Ohm's Law, Lentz's Law, Kirchhoff's Law... this are LAWS, it defines something, it describes a phenomenon, it gives you a set of rules to determine other tangible things... Moore's LAW it is just a set of 2 predictions (or proven theories, by now) made from just an analysis, and maybe to enhance competitivity for whatever reasons.
Second of all, everyone is throwing this USELESS terror of "OOOH NOOO! Moore's Law is reaching an end, what now?" ... this prove again that this is just a theory, if we will not reach x nanometer in x time the world will not collapse, so there is 3 options:
1)this so-called-law it was formulated in an erroneous way (if it is a REAL LAW)
2)we are so great that we are bending a law
3)this is not a law at all
Now, I said what i think and I hope I didn't create a flame post, I say that this is jus my simple, and maybe ignorant, opinion, and with the help of this community, we may all learn something.
Let's start now! 
What are your opinions? Should we consider more things? Let me know!
Have a nice day
shadow_2609

[Kilrah]

It just happens to have been correct for so long that it was called a "law" as a recognition to the accuracy of the prediction...

[shadow_2609]

It just happens to have been correct for so long that it was called a "law" as a recognition to the accuracy of the prediction...

But so, why we keep calling it a Law and keep freaking out when we are reaching the boundary of this prediction? I don't get it  btw thanks for the reply

[ataradov]

Who is freaking out? Who cares, really?

[Kilrah]

Who cares about what it's called?

It's not a name that will change the fact that something that was true for 40 years and isn't anymore is interesting and with no doubt revealing of a change in the industry.

[apis]

I'm guessing it's just modern media that are generating their usual click bait kind of news on a minimal budget.

[ataradov]

Who cares about what it's called?

About the whole thing in general. It is a nifty observation, noting more.

revealing of a change in the industry.

I don't see it. The fact that we are approaching physical limits will lead to change, but the limits were known 40 year ago as well.

And that's actually good that we've reached this limit, now we can stop worrying about improving the process and start working on architectural improvements.

[shadow_2609]

Who is freaking out? Who cares, really?

Because I keep reading about the end of Moore's Law like if it is the end of the world and, if I am not wrong, I think I read something on EETimes about this, but like months ago so don't take my words as 100% sure.
Thanks for the reply

[shadow_2609]

Who cares about what it's called?

It's not a name that will change the fact that something that was true for 40 years and isn't anymore is interesting and with no doubt revealing of a change in the industry.

Great point! I think is more of a change more than an end of something.
Thanks for the reply 

[ataradov]

EETimes

Well, yes, useless clickbait media. Be glad they don't write about Yellowstone exploding yet. News are of similar quality and value.

[shadow_2609]

I'm guessing it's just modern media that are generating their usual click bait kind of news on a minimal budget.

Yes it could be as simple as that, in fact I don't find much point of view from blogs or things not managed by a "company"
Thanks for the reply

[shadow_2609]

EETimes

Well, yes, useless clickbait media. Be glad they don't write about Yellowstone exploding yet. News are of similar quality and value.

...and I erroneous thought it was a place to find other information beyond company buyouts and Friday mini quiz ahahahaha 
Joking obviously, they also make interesting article about microcontroller (which I absolutely love) and some analog stuffs.
Thanks for replying

[Kilrah]

Because I keep reading about the end of Moore's Law like if it is the end of the world

Well humans don't like any kind of change, that's well known.

I don't see it. The fact that we are approaching physical limits will lead to change, but the limits were known 40 year ago as well.

It's not the engineering field that "freaks out", they obviously know how things are and wouldn't be surprised the least.
It's the sales/marketing depts who have been able to live the easy life for decades and had nothing more to do than shove the "better and faster" stuff that regularly and "naturally" came out of the labs every 2 years out of the door with the ever repeating same speech... These are the ones who will now have to go out of their comfort zone and actually start working to figure out what to do next.

[MK14]

Moore's Law may be carrying on just fine. The new advances in 3D stacked, integrated circuits, such as flash memory, are examples of this. Maybe cpus will go the same way, and we will have cpus with a huge number of cores. Cooling them will be one of the on-going challenges, but having a huge number of cores, running at lower frequencies to keep the power dissipation down, is also viable. E.g. Intel's Knights Landing.

But software is still struggling to keep up with usefully being able to use a huge number of cores. But it is quite easy, with some types of software, much harder with others.
There may also be ways of making things considerably smaller than they are now, but which have not been invented yet.

tl;dr
There was the age of mechanical computers, relay computers, valves/tubes, discrete transistors, then Silicon Chips. Hopefully there will be another such technology, sometime in the future.

EDIT:
Where Moore's Law seems to have considerably slowed down is in terms of the main single execution (thread) core speed of a modern day cpu, at about 4 GHz (currently, http://ark.intel.com/products/88195/Intel-Core-i7-6700K-Processor-8M-Cache-up-to-4_20-GHz, before overclocking), with slightly higher turbo clocks. We seem to have reached a point of very diminishing returns, and great expense, to go below about 10 nm. We are at 14 nm feature size. Next is 10 nm. Even reaching 10 nm (Intel), is taking a rather long time.

[all_repair]

Self fulfilling prophecy that makes their own living so tough while other people are working on anti-Moore law.  Choose your side carefully.

[Brumby]

I agree with what's been said above... It was a neat observation that just held true for so long, it was given the honorary title of 'Law'.

Any thinking person with an appreciation of the physical challenges will have expected a point where this "Law" will falter at some time.  They may not agree when this would happen, but they would certainly not expect it to continue for a thousand years, let alone indefinitely - something which would be mandatory for a formal Law.

Another problem with thinking about Moore's Law as something invariant (as a true Law should be) is that the maintenance of the fundamental of the 'Law' is driven by human endeavour.  As such, there is no inherent mechanism where the Universe presents conditions that drive or even support such a 'Law'.  If researchers were to 'down tools' for whatever reason, then the formal Law attribute utterly collapses.

As for the Media - yes, some of their material will be clickbait ... possibly a lot of it ... but what such stories also do is draw public attention to the subject, reinforcing the desire for Moore's Law to continue and adding incentive, if not pressure, on the research labs to dig deeper.

What I believe will happen is that researchers will develop alternate technologies to the point of commercial production that will reinvigorate processor development and give Moore's Law a new lease on life.  It might come from photonics, quantum computing, a radical development in silicon or something we haven't even imagined yet. 

It should also be noted that the definition of Moore's Law had changed at one point to reflect a more appropriate measurement - and that this might be something that could be subject to change again, should the technology involved at the time be unable to provide a compatible metric.


Nevertheless, consider 'Moore's Law' as having an honorary title - and you'll be fine.

If the Media want to jump around with the subject, then let them.  There's bound to be an EE or researcher that will enjoy the noise when they put in for funding.

[Assafl]

Moore's law came from Intel so traditionally it is about microprocessor density speed and price.

But it is a technology observation. So to me it reflects the entire state of the semiconductor industry. So while high stability and high accuracy voltage references do not go down in price (perhaps even the opposite) - embedded systems with on board AD DA BT Wifi USB etc keep getting more and more integrated and more and more capable.

Today's programmers "waste" resources using high level interpretive languages that are quicker in time to market but make use of the extra computational power of devices. This it true not just to embedded but also to desktops, laptops and mobile as well.

[shadow_2609]

Moore's Law may be carrying on just fine. The new advances in 3D stacked, integrated circuits, such as flash memory, are examples of this. Maybe cpus will go the same way, and we will have cpus with a huge number of cores. Cooling them will be one of the on-going challenges, but having a huge number of cores, running at lower frequencies to keep the power dissipation down, is also viable. E.g. Intel's Knights Landing.

But software is still struggling to keep up with usefully being able to use a huge number of cores. But it is quite easy, with some types of software, much harder with others.
There may also be ways of making things considerably smaller than they are now, but which have not been invented yet.

tl;dr
There was the age of mechanical computers, relay computers, valves/tubes, discrete transistors, then Silicon Chips. Hopefully there will be another such technology, sometime in the future.

EDIT:
Where Moore's Law seems to have considerably slowed down is in terms of the main single execution (thread) core speed of a modern day cpu, at about 4 GHz (currently, http://ark.intel.com/products/88195/Intel-Core-i7-6700K-Processor-8M-Cache-up-to-4_20-GHz, before overclocking), with slightly higher turbo clocks. We seem to have reached a point of very diminishing returns, and great expense, to go below about 10 nm. We are at 14 nm feature size. Next is 10 nm. Even reaching 10 nm (Intel), is taking a rather long time.

So it is possible to this "LAW" to keep being true right? I think the possibility is like what you said, 3d chips structure, and for the temperature I heard graphite is better then silicon, as I remember clearly, I read that oppose less to the flow of electrons, so it produce less heat and can allow for faster speed (and maybe even smaller integration).
Might me quantistic cpu the next step? I read that are not faster then normal cpu but it can process more data, in fact, the 1 bit base can represent more than 1 or 0.
Thanks for the reply 

[shadow_2609]

Moore's law came from Intel so traditionally it is about microprocessor density speed and price.

But it is a technology observation. So to me it reflects the entire state of the semiconductor industry. So while high stability and high accuracy voltage references do not go down in price (perhaps even the opposite) - embedded systems with on board AD DA BT Wifi USB etc keep getting more and more integrated and more and more capable.

Today's programmers "waste" resources using high level interpretive languages that are quicker in time to market but make use of the extra computational power of devices. This it true not just to embedded but also to desktops, laptops and mobile as well.

So you think it is more vast then just microprocessors? That is a great point! I neve thought about that.
I can agree with you on the wasting of power computation using high level language, if you are talking about pic (with his wasteful c++ compiler), atmel (here just 1 word... ARDUINO! really a WASTE of "1 instruction 1 cycle" propriety), ARM... the only exception is when you are programming something like a computer program or a smartphone app, it is a great thing if you can make a routine in plain assembly, but it is a lot more difficult to make a GUI or a graphical function (or method).
Thanks for the reply

[tatus1969]

I think it can at least be extended to things like RAM, hard drives, flash memories, computer mainboards, smart phones. Wait for two years, and you get at least double performance in these areas for the same price.

Maybe it is just subjective impression, but to me it looks like the curves are getting flatter recently. Just last week I bought two new hard drives, that 2x6TB cost me 350€. The 2x3TB two years ago were at 150€ if I remember correctly.

[shadow_2609]

I agree with what's been said above... It was a neat observation that just held true for so long, it was given the honorary title of 'Law'.

Any thinking person with an appreciation of the physical challenges will have expected a point where this "Law" will falter at some time.  They may not agree when this would happen, but they would certainly not expect it to continue for a thousand years, let alone indefinitely - something which would be mandatory for a formal Law.

Another problem with thinking about Moore's Law as something invariant (as a true Law should be) is that the maintenance of the fundamental of the 'Law' is driven by human endeavour.  As such, there is no inherent mechanism where the Universe presents conditions that drive or even support such a 'Law'.  If researchers were to 'down tools' for whatever reason, then the formal Law attribute utterly collapses.

As for the Media - yes, some of their material will be clickbait ... possibly a lot of it ... but what such stories also do is draw public attention to the subject, reinforcing the desire for Moore's Law to continue and adding incentive, if not pressure, on the research labs to dig deeper.

What I believe will happen is that researchers will develop alternate technologies to the point of commercial production that will reinvigorate processor development and give Moore's Law a new lease on life.  It might come from photonics, quantum computing, a radical development in silicon or something we haven't even imagined yet. 

It should also be noted that the definition of Moore's Law had changed at one point to reflect a more appropriate measurement - and that this might be something that could be subject to change again, should the technology involved at the time be unable to provide a compatible metric.


Nevertheless, consider 'Moore's Law' as having an honorary title - and you'll be fine.

If the Media want to jump around with the subject, then let them.  There's bound to be an EE or researcher that will enjoy the noise when they put in for funding.

I didn't know it changed, but it is exactly what I thought, it is not something mandatory, and, by the way, media makes our life difficult sometimes because they are biased of visualization/share things sooo... And I think I will consider it as an honoring title like you said to have less problem with the question haha
Thanks again!!

[shadow_2609]

I think it can at least be extended to things like RAM, hard drives, flash memories, computer mainboards, smart phones. Wait for two years, and you get at least double performance in these areas for the same price.

Maybe it is just subjective impression, but to me it looks like the curves are getting flatter recently. Just last week I bought two new hard drives, that 2x6TB cost me 350€. The 2x3TB two years ago were at 150€ if I remember correctly.

This is interesting, I am not aware of cost of computer parts, mainly because i don't have any money to spend and secondly because luckly nothing is gone wrong (just wait and see, 7 days from now my pc will have something), but I noticed it in external hdd, they don't go down so easily expecialy SSD.
Thanks for the new point of view 

[Richard Crowley]

Moore's law came from Intel ...

Moore's editorial was published on April 19, 1965 when Moore worked at Fairchild.
Intel did not exist at the time. It was founded on July 18, 1968

[CatalinaWOW]

Moores law was an observation that turned into a plan.  The entire semi industry guides it's R&D, marketing and whatever by what is required to make this true.  It is a convenient form of collaberation.  Calling it a law is convenient, but has the psychological benefit of making it easier to justify sales plans, R&D budgets and the like.  Unfortunately it is a plan that is harder and more expensive to follow each year.  The death of Moores law has been projected since the late 1980s.  There is no real telling when it when we will finally depart from Moores law, but it does seem likely in the next decade or so.

There is nothing wrong with calling it a law.  Ohms law is an observation that has been found to be true in many circumstances.  It is not universally true.  Try operating your SMD resistor at 1 million volts and comparing the results to an ohms law prediction.

[Kilrah]

I think it can at least be extended to things like RAM, hard drives, flash memories, computer mainboards, smart phones. Wait for two years, and you get at least double performance in these areas for the same price.

Maybe it is just subjective impression, but to me it looks like the curves are getting flatter recently.

It has flattened like crazy, mor than that it's pretty much hit a brick wall around 2010, and pretty much instantly switched from a 2-3 year cycle to I'd say a 6+ year one.

That sudden halt is what drives people to wonder what to do to and talk about Moore's law... even if it doesn't apply to that. Chips may actually still become more complicated but the problem is that we aren't really getting much extra performance out of them..

[Brumby]

Ohms law is an observation that has been found to be true in many circumstances.  It is not universally true.  Try operating your SMD resistor at 1 million volts and comparing the results to an ohms law prediction.

Last I knew, Ohms Law doesn't cover vaporisation.

[shadow_2609]

Moores law was an observation that turned into a plan.  The entire semi industry guides it's R&D, marketing and whatever by what is required to make this true.  It is a convenient form of collaberation.  Calling it a law is convenient, but has the psychological benefit of making it easier to justify sales plans, R&D budgets and the like.  Unfortunately it is a plan that is harder and more expensive to follow each year.  The death of Moores law has been projected since the late 1980s.  There is no real telling when it when we will finally depart from Moores law, but it does seem likely in the next decade or so.

There is nothing wrong with calling it a law.  Ohms law is an observation that has been found to be true in many circumstances.  It is not universally true.  Try operating your SMD resistor at 1 million volts and comparing the results to an ohms law prediction.

So is more convenient to say that is a law for marketing puropise? I don't get it  but maybe is because I'm not into electronic marketing hehehe.
One thing, may I ask where Ohm's Law fail? Cause I don't really think that 1000000 V vs 1 ohm resistor will be a wrong calculation... obviously, in high current you have to take care of copper resistance and internal ESR of the power supply... but I cannot see where it could fail  am I ignoring something?
Thanks for the reply

[shadow_2609]

Last I knew, Ohms Law doesn't cover vaporisation.

Tell this to my 1 ohm resistor erroneously taken from the bin of 1k resistor 

[Richard Crowley]

So is more convenient to say that is a law for marketing puropise? I don't get it  but maybe is because I'm not into electronic marketing hehehe.

No.  The term was popularized by Carver Mead, a professor at Caltech. 
Hardly seems like anything to do with marketing, electronic or otherwise.

[MK14]

So is more convenient to say that is a law for marketing puropise? I don't get it  but maybe is because I'm not into electronic marketing hehehe.
One thing, may I ask where Ohm's Law fail? Cause I don't really think that 1000000 V vs 1 ohm resistor will be a wrong calculation... obviously, in high current you have to take care of copper resistance and internal ESR of the power supply... but I cannot see where it could fail  am I ignoring something?
Thanks for the reply

To be more accurate, it would be called, something like,

Moore's Approximation or Moore's Estimation.

Ohms Law does not always work.

E.g. What is the resistance of a normal diode at 0.2 Volts ?
0.5 Volts ?
1 Volt ?
Can you even get to 2 Volts with one ? (please don't ask Mike'ElectricStuff, PhotoInduction or even someone who works at an electricity generating plant that question though). (You can get to 2 Volts but the current needed, does not follow Ohms law, and the diode might explode).

Consider CDs as being similar to Moore's law, which (in simple terms) use various light frequencies to store data on them.
We had CDs approx 0.65 GB
Later DVDs approx 4.7 GB
Later Blu-Rays which can go up to 50 GB, maybe more now or later.

But eventually these CDs/DVDs/Blu-Rays run out of light frequency and/or practical limit on multiple layers, to reach a limit for affordable discs.
I.e. The Laws of Physics, can't be defeated, as far as we know, at the moment.

Moore's Law has reached a huge slow down, with current Silicon Chips, especially cpus.

But sooner or later, technologies will hopefully be invented which will further progress the Moore's Law for computer circuitry.

Another problem/limiting factor of Moore's Law is that it needs the market to continually spend a huge amount of money, for buying the devices. Unfortunately the PC/Chips market has been declining over recent years, as people moved over to other formats. Such as Tablets and mobile phones.
So with less and less money available to create ever faster cpu chips, Intel has had to lay off staff, and spend less and less on developing future cpu chips. That also slows down Moore's Law.

tl;dr
If someone with $100,000,000,000,000,000,000,000,000,000,000,000 said give me a faster cpu, and they were given a few years, we could probably make faster chips. But at the moment ever decreasing money resources are available.

[CatalinaWOW]

Moores law was an observation that turned into a plan.  The entire semi industry guides it's R&D, marketing and whatever by what is required to make this true.  It is a convenient form of collaberation.  Calling it a law is convenient, but has the psychological benefit of making it easier to justify sales plans, R&D budgets and the like.  Unfortunately it is a plan that is harder and more expensive to follow each year.  The death of Moores law has been projected since the late 1980s.  There is no real telling when it when we will finally depart from Moores law, but it does seem likely in the next decade or so.

There is nothing wrong with calling it a law.  Ohms law is an observation that has been found to be true in many circumstances.  It is not universally true.  Try operating your SMD resistor at 1 million volts and comparing the results to an ohms law prediction.

So is more convenient to say that is a law for marketing puropise? I don't get it  but maybe is because I'm not into electronic marketing hehehe.
One thing, may I ask where Ohm's Law fail? Cause I don't really think that 1000000 V vs 1 ohm resistor will be a wrong calculation... obviously, in high current you have to take care of copper resistance and internal ESR of the power supply... but I cannot see where it could fail  am I ignoring something?
Thanks for the reply

Ohm's law assumes/postulates linear behavior of materials.  It literally says that the ratio of volts to amps for a segment of conducter is constant.  All materials behave non-linearly when you are far enough from room ambient conditions.  The causes of the non-linearity vary.  It may be simple self heating which changes the temperature.  It may be a phase change.  It may be strange quantum things that happen at low temperature.   Some materials depart from linearity very close to standard operating conditions.  These properties are exploited to make devices like surgistors.

So ohms law fails whenever you get out of the linear region.  A significant part of engineering is making sure that the assumptions apply, which is why we select resistors with a high enough watt rating to stay in the linear operating region.  My example of a SMD resistor with a million volts across it was supposed to convey the idea of a material operating outside the linear region.  There obviously was not enough depth in the explanation.

It is over simplification to think of Moore's law as just marketing.  Just about every segment of the industry for several decades used it as a roadmap.  Mask makers knew what feature size they needed to be able to make over the next few years.  Resist makers knew how they would have to modify their formulations to pattern at finer and finer sizes.  Clean room vendors knew the particle sizes that would need to be controlled.  The marketers knew what they would be selling.  There are an unbelievable number of pieces to the puzzle, and they all used Moore's law to guide when they achieved goals in their own piece of the technology.  They still do to some extent.

[hamster_nz]

One thing, may I ask where Ohm's Law fail?

Basically anywhere where the thing under test is non linear. A simple example could be an light bulb where the resistance changes as the filament gets hot. If you halve the voltage across a glowing bulb you will not half the current, which is against the prediction of Ohm's Law....

[tatus1969]

Ohms Law does not always work.

It's somehow getting philosophical here, and I will be happy if someone can negate my position here:

Everything created by us humans is an approximation based on observation. The more through the observation the better the approximation. It is our human arrogance giving them the suffix "law". For example: Newton found that a mass accelerates linearly under constant force. Einstein then improved this formula for high velocities. As our methods of observing the universe constantly evolve, maybe we will find a more complete formula in an even greater context, who knows?

In a way we can then compare Ohms Law with Newton, and quantum mechanical electron gas theory would at least go in the direction of Einstein's general theory of relativity. I prefer using Ohms Law whenever it gives me an acceptable approximation :-)

[shadow_2609]

So is more convenient to say that is a law for marketing puropise? I don't get it  but maybe is because I'm not into electronic marketing hehehe.
One thing, may I ask where Ohm's Law fail? Cause I don't really think that 1000000 V vs 1 ohm resistor will be a wrong calculation... obviously, in high current you have to take care of copper resistance and internal ESR of the power supply... but I cannot see where it could fail  am I ignoring something?
Thanks for the reply

To be more accurate, it would be called, something like,

Moore's Approximation or Moore's Estimation.

Ohms Law does not always work.

E.g. What is the resistance of a normal diode at 0.2 Volts ?
0.5 Volts ?
1 Volt ?
Can you even get to 2 Volts with one ? (please don't ask Mike'ElectricStuff, PhotoInduction or even someone who works at an electricity generating plant that question though). (You can get to 2 Volts but the current needed, does not follow Ohms law, and the diode might explode).

Consider CDs as being similar to Moore's law, which (in simple terms) use various light frequencies to store data on them.
We had CDs approx 0.65 GB
Later DVDs approx 4.7 GB
Later Blu-Rays which can go up to 50 GB, maybe more now or later.

But eventually these CDs/DVDs/Blu-Rays run out of light frequency and/or practical limit on multiple layers, to reach a limit for affordable discs.
I.e. The Laws of Physics, can't be defeated, as far as we know, at the moment.

Moore's Law has reached a huge slow down, with current Silicon Chips, especially cpus.

But sooner or later, technologies will hopefully be invented which will further progress the Moore's Law for computer circuitry.

Another problem/limiting factor of Moore's Law is that it needs the market to continually spend a huge amount of money, for buying the devices. Unfortunately the PC/Chips market has been declining over recent years, as people moved over to other formats. Such as Tablets and mobile phones.
So with less and less money available to create ever faster cpu chips, Intel has had to lay off staff, and spend less and less on developing future cpu chips. That also slows down Moore's Law.

tl;dr
If someone with $100,000,000,000,000,000,000,000,000,000,000,000 said give me a faster cpu, and they were given a few years, we could probably make faster chips. But at the moment ever decreasing money resources are available.

Let's keep this example about the diode, ohm's law doesn't define that a resistance is constant for types of devices, it gives just relationship between current, voltage and resistance, in fact we could calculate the resistance of that diode at 0,2 volts if we know the current, for ohm's law R=V/I, let's say 200nA, R=0.5/0.0000002= 2.5Mohm, so it does respect the law.
And, like we said, it will take a lot of times to reach to an end, it's just slowed down
Thanks for the reply!

[shadow_2609]

Moores law was an observation that turned into a plan.  The entire semi industry guides it's R&D, marketing and whatever by what is required to make this true.  It is a convenient form of collaberation.  Calling it a law is convenient, but has the psychological benefit of making it easier to justify sales plans, R&D budgets and the like.  Unfortunately it is a plan that is harder and more expensive to follow each year.  The death of Moores law has been projected since the late 1980s.  There is no real telling when it when we will finally depart from Moores law, but it does seem likely in the next decade or so.

There is nothing wrong with calling it a law.  Ohms law is an observation that has been found to be true in many circumstances.  It is not universally true.  Try operating your SMD resistor at 1 million volts and comparing the results to an ohms law prediction.

So is more convenient to say that is a law for marketing puropise? I don't get it  but maybe is because I'm not into electronic marketing hehehe.
One thing, may I ask where Ohm's Law fail? Cause I don't really think that 1000000 V vs 1 ohm resistor will be a wrong calculation... obviously, in high current you have to take care of copper resistance and internal ESR of the power supply... but I cannot see where it could fail  am I ignoring something?
Thanks for the reply

Ohm's law assumes/postulates linear behavior of materials.  It literally says that the ratio of volts to amps for a segment of conducter is constant.  All materials behave non-linearly when you are far enough from room ambient conditions.  The causes of the non-linearity vary.  It may be simple self heating which changes the temperature.  It may be a phase change.  It may be strange quantum things that happen at low temperature.   Some materials depart from linearity very close to standard operating conditions.  These properties are exploited to make devices like surgistors.

So ohms law fails whenever you get out of the linear region.  A significant part of engineering is making sure that the assumptions apply, which is why we select resistors with a high enough watt rating to stay in the linear operating region.  My example of a SMD resistor with a million volts across it was supposed to convey the idea of a material operating outside the linear region.  There obviously was not enough depth in the explanation.

It is over simplification to think of Moore's law as just marketing.  Just about every segment of the industry for several decades used it as a roadmap.  Mask makers knew what feature size they needed to be able to make over the next few years.  Resist makers knew how they would have to modify their formulations to pattern at finer and finer sizes.  Clean room vendors knew the particle sizes that would need to be controlled.  The marketers knew what they would be selling.  There are an unbelievable number of pieces to the puzzle, and they all used Moore's law to guide when they achieved goals in their own piece of the technology.  They still do to some extent.

Well, even if we experience non-linearity or resistance drift, ohm's law always apply, an example could be the varistor, it changes his resistance so can drop the current surge, so the voltage don't spike and the load is safe... but ohm's law always apply, that is why this varistor works, more resistance, less current, and dhte voltage can be determined by this 2 other parameter.
It will be impossible in NEVER that 1v applied to 1 ohm will generate 2 amps or 0.5 amps... ohm's law.
Thanks for the reply

[shadow_2609]

One thing, may I ask where Ohm's Law fail?

Basically anywhere where the thing under test is non linear. A simple example could be an light bulb where the resistance changes as the filament gets hot. If you halve the voltage across a glowing bulb you will not half the current, which is against the prediction of Ohm's Law....

Like I said before, it will be impossible to get 2 amps with 1 volts applied to 1 ohm... this will always apply, even in inductors: at high frequency we have a high resistance (or impedence) which limits the current through the inductor.
As you said the resistance change, and if we measure the current at a given times we can see this resistance rising, inducing more voltage drop on the component for the increased current.
Thanks for the reply

[MK14]

Let's keep this example about the diode, ohm's law doesn't define that a resistance is constant for types of devices, it gives just relationship between current, voltage and resistance, in fact we could calculate the resistance of that diode at 0,2 volts if we know the current, for ohm's law R=V/I, let's say 200nA, R=0.5/0.0000002= 2.5Mohm, so it does respect the law.
And, like we said, it will take a lot of times to reach to an end, it's just slowed down
Thanks for the reply!

We seem to have gone off topic. But to answer, anyway.

Ohms Law falls to pieces, because it can no longer predict the answer.

If I give you a specific FET or bipolar transistor (a diode was used above, because it is a semiconductor), and there is precisely 10 volts between the collector and emitter (Bjt), but I refuse to tell you what the base/gate's voltage/current etc are. Tell me what current will flow (between C and E), using Ohms law ?

You probably CAN'T because the Fet/Bjt is not an ohmic device. (With base/gate voltage/current unknown, you don't have enough information to work it out).

Moore's Law has similar problems with it. It basically falls to pieces, once the fundamental laws of Physics, limits have been reached, and/or it is no longer affordable to develop new and better devices.

Toilets kind of stopped developing about 100 years or more ago. I think it is a similar concept.

The other problem/fault with Moore's Law. Is that for many people. We have reached the "Good Enough" as regards performance/speed for a modern day PC. Since many people are only browsing the internet, watching Youtube etc etc. Creating Emails.
Modern days PCs are "Good Enough", so why go and buy a better/faster PC, as it already does everything they want.
Hence lack of funding/market for future ever faster cpus.

tl;dr
We may need killer applications, which could be future VR (virtual Reality) and/or gaming systems (but much realistic), or maybe even very smart AI.
But currently, modern PCs are good enough for most people, most of the time. If it ain't broke why fix it ?
Which sadly (if you like super fast cpus), damages Moore's Law, at least as regards cpus/PCs.

[hamster_nz]

But currently, modern PCs are good enough for most people, most of the time. If it ain't broke why fix it ?

But the current CPUs are only just good enough... everybody wants more cores, more memory bandwidth and greater consolidation for less space and power.

Currently we have 2xCPUs, each with  12 cores, 512GB RAM, and would like more per blade.

[MK14]

But currently, modern PCs are good enough for most people, most of the time. If it ain't broke why fix it ?

But the current CPUs are only just good enough... everybody wants more cores, more memory bandwidth and greater consolidation for less space and power.

Currently we have 2xCPUs, each with  12 cores, 512GB RAM, and would like more per blade.

But that is the thing. Some people (especially me, typing this on my 6 core computer), want the latest/fastest/greatest cpu ever. But sadly these days, we are more of a minority, rather than the main people who go out and buy $299 PCs.

Look at the success of the latest Intel Atom (and similar) cpus, with terrible performance. But which sell like hot cakes, because they are priced at $249 each, rather than $999 or $1,999, for the latest overclockable, decent graphics carded PC.

The market has shifted, which is another significant reason why Moore's Law is not performing well these days.

Also Intel seem to have a somewhat crazy price to the highly prized, 10 core Broadwell-E processor, at almost £1500 (i7 -6950X), just for the cpu. Not many enthusiasts would be willing to pay that much.

EDIT:
I was mainly talking about the consumer market.
With businesses (servers), they often want much more capabilities, every year. So as regards that I fully agree. More cores and memory, etc etc.
But even in business, some server use (e.g. a gentle web server), is fine, even on a slower, older cpu. So again, they may leave upgrading for a long time. Until it starts failing and/or the repair contracts are too expensive or something.

[tatus1969]

Look at the success of the latest Intel Atom (and similar) cpus, with terrible performance. But which sell like hot cakes, because they are priced at $249 each, rather than $999 or $1,999, for the latest overclockable, decent graphics carded PC.

The market has shifted, which is another significant reason why Moore's Law is not performing well these days.

You somewhat contradict yourself here. Moores Law predicts the ratio between performance and price to increase squared. This can be achieved by making faster and faster CPUs, but also by making them cheaper and cheaper. So even if we are about to reach the physical limits there is still way to keep that development (although as I think at a reduced rate).

[MK14]

Look at the success of the latest Intel Atom (and similar) cpus, with terrible performance. But which sell like hot cakes, because they are priced at $249 each, rather than $999 or $1,999, for the latest overclockable, decent graphics carded PC.

The market has shifted, which is another significant reason why Moore's Law is not performing well these days.

You somewhat contradict yourself here. Moores Law predicts the ratio between performance and price to increase squared. This can be achieved by making faster and faster CPUs, but also by making them cheaper and cheaper. So even if we are about to reach the physical limits there is still way to keep that development (although as I think at a reduced rate).

I covered that in my earlier post (before the one you quoted). When I said that Moore's Law was continuing, via things like 3D stacked dies, and very large number of cored cpus. But it remains to be seen whether such technologies can keep the price the same or less, while simultaneously increasing the number of cores as well. So far it just about has kept the price down or the same.

But getting past around 4 GHz or so, for consumer priced cpus, seems to be an apparent bottle neck these days. Taking IPC (Instructions Per Clock) into account as well, there has only been a moderate improvement (e.g. 5%), per generation from Intels last few releases.

tl;dr
Moore's Law, Maximum performance for a single core, has hit the buffers (massively slowed down or stopped), these days.

But Moore's Law as regards ever cheaper cores, more cores per die/chip, and lower cost overall SoC/cpus/Gpus-low-end (at lower ends of the performance market), are to an extent, still carrying on with some head room, before the limits of Moore's Law hit them as well.

Graphics cards already can have thousands (and maybe tens of thousands, now or later) of "small" graphic cores. They are not full cpus in the PC sense. But can usually do floating point stuff, and handle a huge number of threads. They are usually at a lower frequency (e.g. 1 GHz), but there are so many of them, they become extremely powerful when used in parallel.

The techniques for making extremely small feature sized integrated circuits, these days, has become increasingly expensive. This is damaging Moore's Law, unfortunately. It use to be that ever smaller transistors were cheaper and cheaper to make. But the price has stabilized (very approximately), nowadays. It might even go up in the future, for even smaller feature sizes, as very expensive techniques have to be used, in order to make the chips in the first place.

But lower performance chips, can still become cheaper and cheaper, but it might take a longer period of time to achieve in the future. As cheaper methods of making less tiny featured chips, are invented.

tl;dr2
Moore's Law is mixed up these days. It depends on specifically what you are talking about. Some things are still partially following it (e.g. Flash memory, maybe graphics, core count and cheap cpus), but others have slowed or almost stopped (e.g. Cpu max GHz frequency).

[MK14]

Maybe a better way of looking at it, is to look at the "Moore's Law" of op-amps, over the years.

Originally, approx 1960's, IC op-amps were extremely expensive. Something like $50 each, so considerably more in today's money. The price went down, and down over the years, and since the 741 (especially), they have been perfectly usable and useful in electronics.
But since the late 1970's (I'm not sure of the exact date), they basically stopped getting any cheaper.

Also they got better and better, in various respects. But again, by and large in the 1980's, you could basically get ones, which are perfectly usable, even today, specs wise. But there probably has been the odd improvement here and there.

tl;dr
Moore's Law seemed to apply to op-amps, price and capabilities, from the 1960's till about 1980 (again, I'm not 100% sure of the exact dates). Then it mostly stopped. With some exceptions in some areas, if you want to nit pick.

Digital logic gates, had a sort of Moore's Law, initially being very expensive and not as fast, as they were to end up being.
Approx 1960's first available, dropping in price and increasing in speed and/or lower power consumption.

These days, although they are cheap, they are partially obsolete, because they have been mainly replaced by MCU's, PALs/CPLDs and of course FPGAs. Some sources suggest that if you need more than one logic IC and/or might want to make later firmware changes, then CPLDs/FPGAs are the way to go. Certainly logic boards with fifty or a hundred logic gate series ICs, are mostly a thing of the past. Except some vintage builders etc.

Maybe one day instead of cpus as such, we will have some kind of massive FPGA and/or million or a billion or more cored calculation monsters.

[shadow_2609]

Let's keep this example about the diode, ohm's law doesn't define that a resistance is constant for types of devices, it gives just relationship between current, voltage and resistance, in fact we could calculate the resistance of that diode at 0,2 volts if we know the current, for ohm's law R=V/I, let's say 200nA, R=0.5/0.0000002= 2.5Mohm, so it does respect the law.
And, like we said, it will take a lot of times to reach to an end, it's just slowed down
Thanks for the reply!

We seem to have gone off topic. But to answer, anyway.

Ohms Law falls to pieces, because it can no longer predict the answer.

If I give you a specific FET or bipolar transistor (a diode was used above, because it is a semiconductor), and there is precisely 10 volts between the collector and emitter (Bjt), but I refuse to tell you what the base/gate's voltage/current etc are. Tell me what current will flow (between C and E), using Ohms law ?

You probably CAN'T because the Fet/Bjt is not an ohmic device. (With base/gate voltage/current unknown, you don't have enough information to work it out).

Moore's Law has similar problems with it. It basically falls to pieces, once the fundamental laws of Physics, limits have been reached, and/or it is no longer affordable to develop new and better devices.

Toilets kind of stopped developing about 100 years or more ago. I think it is a similar concept.

The other problem/fault with Moore's Law. Is that for many people. We have reached the "Good Enough" as regards performance/speed for a modern day PC. Since many people are only browsing the internet, watching Youtube etc etc. Creating Emails.
Modern days PCs are "Good Enough", so why go and buy a better/faster PC, as it already does everything they want.
Hence lack of funding/market for future ever faster cpus.

tl;dr
We may need killer applications, which could be future VR (virtual Reality) and/or gaming systems (but much realistic), or maybe even very smart AI.
But currently, modern PCs are good enough for most people, most of the time. If it ain't broke why fix it ?
Which sadly (if you like super fast cpus), damages Moore's Law, at least as regards cpus/PCs.

Sorry my fault for this off-topic hehe.
By the way now I understand a lil better what you are saying, we misunderstood each other... but this thing about Ohm's law make me tingle for some reason hahaha.
And yes, the market now point to an average between performance/price, of course with some exceptions (you know, pc player tend to go berserk on powerful pc hardware, like for example alienware, or even 3d designer which have to a powerful gpu/cpu for working).
For the killer app I hope it will not be another IOT thing (or IOS like our dave says, which has a really good point).
Thanks and sorry for the off topic 

[MK14]

Sorry my fault for this off-topic hehe.
By the way now I understand a lil better what you are saying, we misunderstood each other... but this thing about Ohm's law make me tingle for some reason hahaha.
And yes, the market now point to an average between performance/price, of course with some exceptions (you know, pc player tend to go berserk on powerful pc hardware, like for example alienware, or even 3d designer which have to a powerful gpu/cpu for working).
For the killer app I hope it will not be another IOT thing (or IOS like our dave says, which has a really good point).
Thanks and sorry for the off topic 

No problem, it is very easy to go off-topic, I seem to do it all the time.

Because Arm have been a lot below the performance of Intel cpu chips, there has been quite a climb in Arm's performance, capabilities and number of cores. So Arm has not necessarily hit the end of Moore's Law, but I think they are getting ever closer.
MCUs we use in electronics, such as PIC/Arm etc, are a long way from the Moore's Law partial-end point.
Since embedded cpus are easily/cheaply up to about 200 MHz (I know you can get much faster ones, such as >500 MHz), so a long way from 4 GHz (although limited power dissipation also limits the frequency).

Also most are only single core (I know there are some with more than one core in the MCU), so there is a long way to go, as Arms (in mobile phones) are already at about 8 cores.

tl;dr
Most PICs/STarms (mainstream electronics use ones, NOT mobile phone high end ones) and such like, are using relatively ancient technology, so are miles and miles from reaching the ends of Moore's Law.
By the time they reach the end, the price of making the Silicon will hopefully be really cheap by that time. For typical MCUs.
They are already considerably better than they were, about twenty years ago.
MCUs have probably easily got 10 .. 20 or more years, before they hit the solid Moore's Law wall. By then, better technologies might have been invented, allowing further advancements, anyway.

[Assafl]

Moore's law came from Intel ...

Moore's editorial was published on April 19, 1965 when Moore worked at Fairchild.
Intel did not exist at the time. It was founded on July 18, 1968

I am humbled and stand corrected...

That is what happens when one assumes anything - even things one is sure of for 30 years... Thank you from making me less ignorant!

On a few side notes after reading the original Electronics article (for the first time!):
1. It actually came from digital systems IC fabrication - so not even Microprocessors....
2. He specifically states that Analog circuits will not gain as much as digital.
3. He defines the IC as an User Irreducible Unit (hence one that cannot be split - unlike a circuit).
4. Not limited to ICs. Also thin film technologies...
5. These were the days of ceramic and gold packages - so he states packaging is so high cost nobody cares about yield (hilarious)
6. Presently the optimum was 50 parts per IC - so this is squarely RTL TTL and OpAmp days..


https://drive.google.com/file/d/0By83v5TWkGjvQkpBcXJKT1I1TTA/view