 I am A. N. Sandalkar from IIT Bombay, Department of Electrical Engineering. My interest area is VLSI design. We have a around 40 hour course in this area of advanced VLSI design. I will be assisted or rather I will be coordinating this kind of course with three other faculty. The details about the course content and the faculty and what they will teach I will speak about in my next talk. Today I will start to try to talk with you something about historical perspective and future trends in CMOS, VLSI circuits and system design. Let us start with the very early 20th century. There are many inventions in 20th century like airplane, nuclear power generation, computers, space aircraft and things of that kind. However, everything has to be controlled by electronics as is now known. So many people ask what exactly is electronics? So electronics I must say it is the most important invention of 20th century and it essentially means the flow of electrons in a circuit and therefore these circuits were called electronic circuits. The modified version or the integrated version of same circuit is called integrated circuits and this course is essentially talking about very large scale integrated circuit design. Not in an IC the present things like mobile phone cannot be made. Just to give a just to take you even before 50 years before what we were say for example, world of 1958 we are the first artificial satellite Sputnik which was injected by Russian ejected Russian in 1957. We started with radio and today we are at TV. You can see from here initially we started with vacuum tubes even televisions were vacuum tube as late as 2000 year. Now only we have LCD LED displays earlier we have all vacuum tube based TVs. The first solid state transistor actually appeared in 1955 and the credit goes to Sony. So as I said electronics is the most important invention in 20th century. These are integrated circuits what how they actually have progressed in last 100 years. We started with vacuum tube and today we are at very large scale integrated circuits. Just to tell you 6 years ago it was the 100 year anniversary of a vacuum tube. The first vacuum tube were made by LiddyForest and it looked like something like this. And today when we are talking of a VLSI this is a SRAM which we are showing you this is a 64 K SRAM from Intel and you can see the change in the structure change in the operation of a single vacuum tube compared to what today we call memory semiconductor memories. The first computer which appeared in the scenario of 2000 which was even before 2000 was credited to Babbage. This is called difference engine which was proposed by Babbage in 1832 and it has 25,000 mechanical parts and in those days it cost something like 17,470 pounds. In today's money it may be millions and millions of dollars equivalently or may be crores of rupees in Indian money. So this kind of a mechanical system was first thought which later on became registrics or some kind of register system. And please remember the first register system was used or rather popularized by the most famous company in computer the IBM actually they also have their first machine which was mostly mechanical system. The first electronic computer came in 1946 was named as INIAC essentially it was electronic calculator some kind of equivalent but it was made out of vacuum tubes. You can see this has been put into a garage of a huge building or the basement of a huge building and it even it was doing only four small operations but it requires such a large area large systems large in size it consumes shell of a power and it has a very short time short life filaments available on the tubes. So you have to keep changing the tubes. Let us take a comparison if for this INIAC equivalent if I now make a Pentium 4 in INIAC equivalent circuit or equivalent system it will be at least two heights of a Empire State building in New York and will require for to cool it and most two Niagara Falls every minute to throw on it. So the kind of system from where we started in 46 and the kind of system we are now talking in 2012 is a hell of a difference large improvements large something which can do many more functions than what we thought in 1946. Please remember in life whether we talk about circuits we talk about anything at the end of the day anything can be done if there is money after all economics matters or money matters. So for you those who are money minded and maybe one should be these days let me show you some kind of a graph which says we are looking for customers who breathe, eat, live only electronic systems. So as I was talking to you about economics in case of semiconductor industry this economics is related to what we say. If you see the lower most triangle inverted pyramid the lower part the semiconductor equipment material requirement is around 100 billion dollars the semiconductor itself is around 400 billion dollar market electronic equipment is around 1,050 billion dollars and overall impact on political microeconomic environments which uses these kind of electronic components and systems have around 50,000 US billion dollars. So this is the kind of money in which electronics is now involved and therefore one should not take things very lightly why we are progressing so fast because somehow we want to see that we do much more profit than what we have been doing today. Many students have asked me over the years that how do you really know this when before you start a particular system on chip that is you want to fabricate a chip what will be the cost of production. So I have actually taken it as old data 2005 data not that it is the current value system but the idea of evaluation of any product money is shown here. Say for example, 1000 wafers per month if that is what your company can manufacture of course they are large number of parts 600 dollars are actually required for 6 inch quarter micron wafer thickness wafer. So therefore it is roughly for the lot you are talking it is 3.5 dollars per centimeter square is the charge you require for the silicon. Add for packaging around quarter dollar pin for you know quad kind of flat package this is the most costliest part per pin because it may be 100 pin, 200 pin package so you will lot of cost actually goes per pin on the packaging. Then you add around 200 dollars an hour for 256 pin mixed signal tester about 1 second to move sides. 6 inch wafer and 180 centimeter square 8 inch which is 1 310 centimeter square is the area typical I am talking and from this numbers then one can probably evaluate please remember I have not added the design cost which at times may be larger than this but generally it is now found that the cost of chip is actually goes in the testing which is the highest amount of money one spends just to say that my chip is working or working well to the specification. Coming back to what I was talking I just thought these two slides to show you that why semiconductor industry is doing well or why so much effort is put on the semiconductor industry why Intel IBM HPs are known or Texas instrument is known world across the simple reason as I say is the amount of money which they are generating or spending on people as well as on systems. So let us go back to say history how semiconductor devices started way back in 1947 first point contact bipolar transistor in germanium was made and the credit goes to Bardin in Britain and they won Nobel Prize in 48 first junction actually it was not 48 it 48 it started in 1956 the first bipolar junction transistor and not the point contact transistor was actually first suggested and his group then manufactured or fabricated was due to William Shockley. There is an interesting story on Shockley as I proceed ahead I will talk about that he also one same time 1956 Nobel Prize were awarded to Shockley Bardin in Britain for invention of transistor. In 1958 the first integrated circuit was suggested and was actually made by Jack Kilby then he was at the Texas instruments he is late Jack Kilby now and he won in 2000 the Nobel Prize. In 1959 the first planar IC came and that credit goes to Robert Noyce and he is the most famous person in integrated circuit I will come to it when it comes next slide. And the major invention of today's this came through the efforts of Kang and Atala Bell Labs and they make first MOS transistor. Here are some interesting photographs the first point contact transistor shown on your left is a point contact done by Bardin and Britain you will be surprised what they were actually trying on a simple germanium piece they were putting two probes and trying to feel whether they it can amplify signal so it was quite trivial when they started but when they started bonding it they suddenly realize just they got some kind of amplification and that made you know sudden change in thinking of most people that oh simple material with two contacts can actually do amplification which was vacuum tube required a huge area huge power 300 volt supply contrast to this it required hardly 5 volt supply and very small current low power and was still doing amplification for signals. So this was an invention of 1947 which made the today's world whatever we see in most of the integrated circuit area is essentially due to the first such invention by Bardin and Britain on your right you can see there are two three people sitting one is Bardin on the left the right is Bratton and the lower one is the famous person William Shockley he was the head of the group which was supposed to government of your united state that time asked them to actually make some kind of replacement for vacuum tube and he was heading that group and Shockley went in 1948 4748 he was in Caltech as a visiting faculty and during this time Bardin and Bratton actually invented so when the patent was filed by Bellas about this point contact transistor it did not put Shockley's name and Shockley was furious when he came back he said this is unfair because most of this discussion which went through before 47 was with me by both Bardin and Bratton and he had a tough fight with Bardin Bardin then actually you will not believe but in 1951 Bardin left the group and actually started working on some other area which is called superconductivity and he won his second Nobel Prize in 1959. So Bardin was the very most furious person that Shockley wants all credit but Bellas did not file a patent along with Shockley. So Shockley started working on germanium junction transistor instead of point contact and in 1954 he first time actually showed how a junction transistor can work and for this invention of his along with Bardin and Bratton all three were awarded Nobel Prize. There are interesting histories about Shockley please go through some Wikipedia kind of things to know how Shockley really behaved in his life. All said and done this is slightly magnified version of the same transistor point contact germanium transistor which was made in 1954 display this is this photograph is actually taken from Bellas Museum where this transistor is still actually shown. In 1958 you know the another youngs engineer then Jack Kilby joined Texas instrument and the way he was he was actually asked to help people in making some CRT tubes kind of base circuits and he had much more time to while away. So he was in sitting alone he used to find that why you know only transistors have been put separately then you put on a board something a resistor this why not I can make all components in silicon or in a one same material and then if I join it will be a universal kind of circuit in which everything is made out of a semiconductor. This idea not only he thought but he also actually introduced that itself on a board and under your left the first integrated circuit as made by Jack Kilby is shown to you though it looks far away from today's integrated circuit but the main important point that this was the first IC which has one resistor two resistor one capacitor one transistor all together on a circuit and that has worked the importance is it was working circuit. Please remember Jack Kilby was not assigned this project even after this invention by Texas instrument. So that is the irony of life that you are asked to do something for which you think you know more. However Jack Kilby's invention was never treated very high till so late as 2000 in which he was then awarded Nobel Prize. Simultaneously when Jack Kilby was doing something on the kind of things he showed Robert Noyes then at Fairchild and then who then invented then he started company along with others Intel. They actually were thinking of making all components in silicon and they called PN junction based silicon device isolation technique which allowed components to be separated in the silicon. There was another scientist or another engineer Kurt Lehueck who was that time at Sprague Electric and then joined USC as very distinguished professor at University of Southern California. They also worked on PN junction isolation theory and that allowed us to actually separate components inside a silicon block. Unfortunately both died before Kilby got the Nobel Prize probably he would have to share the money as well as the credit if had they been allowed in 2000. So here is the first planarite integrated circuit you can see from here there are three parts annular form one this one downwards and one center and these are contacts. So PNP transistor was first made by Robert Noyes inside the silicon itself all diffusions inside all contacts on the top and this was the major invention I must say which allowed integrated circuit to manufacture in large number of densities and very cheap methods. This so called planar technique still stands and this is the way almost every semiconductor chip is made. Here is the photograph of Robert Noyes who actually also he was also working with there is called eight dirties among them is Gordon Moore Intel's noise then there used to be Chang. All these were working with a new company in Menlo Park in California under the headship of Shockley but as usual Shockley never wanted to be credited to any one of them and therefore they all ate dirties left and started a company Fairchild and this Fairchild was a camera company they did some work and then they realized that in a camera company their choice they were only making C series. So they left Fairchild and started their own company and then called Intel Corporation. What today you all see is because of probably Shockley because if Shockley would have been a good man probably Intel would never have started. To sum up on this kind of device I have a interesting slide you know one of my very famous distinguished colleague at 2Q Institute of Technology is Professor Hiroshi and this slide actually was made by him so I thought I always show to everyone. He said how the device has devices technology have progressed over the years so you are left you can see Professor Iway was a boy of 12 years in 64. However in 2012 this is how it looks he is 60 years old and in 2062 another 50 years if you add he will be 112 years and I do not know how will you look at too long a time for anyone for him as well as for me. So what has happened in 64? Transistor just started to be used for radio. Yeah most people thought amplifiers is all that you need some small logic was made and the history shows that this part was though always novel but was not very difficult. It was rather easier to actually achieve successes faster. Slowly when you say every system became based on integrated circuits particularly silicon integrated circuits over the years last you can see 2012 onwards till 2062 any progress in silicon IC industry or in silicon technology will be very very difficult. So for us it was a nice time we could do small things and get credit for it probably in future when you start doing work in this area you will find very difficult to progress as the pace with which we all progressed or the technology progressed. So in Nutshell in 1900 diode came the 1906 triode came Wilson's invention into 1926 misfit and MOSFET were first announced never made but announced first bipolar transistor came in 47 to 50 both Barden-Bretton and Chocolet's inventions first IC came in 58 Kilby noise and Lehovic first MOSFET appeared in 1960 by Kang and Attala and first large scale integrated circuit appeared in 1970. The names were given because we started looking for the number of components on chip. So first we started vacuum tube first 20 years it took us to reach to transistor concept next 30 years we could make IC's and last 10 years up to 70 we were spending on large scale integration and I do not know next last 30 40 years we are in the era of VLSI or ultra large scales. What was the difference from 1900 to 2000 otherwise in this performance wise the first electronic circuit was huge power and very slow huge power consuming and comparatively low speeds lower speeds we wanted to look for lower power dissipation. So we went from vacuum tubes to solid state now silicon technology allowed high integration and when we went to CMOS we have a low power high speed very high integration devices put on a single chip and we are still continuing for high integration very high speed and low power they remain the same parameters in which integrated circuits are being advancing now. Please remember that in 1925 to 33 twice there was a research done on MOSFET including Shockley was actually working on MOSFET and Shockley has a file legal file with many people saying that MOSFET was his concept unfortunately Bell Lab did not file a patent on the name of Shockley otherwise probably this would not have happened. Shockley of course always had a brighter idea than most people of his time so nothing that he could not think because of very bad interface property then between semiconductor and the gate insulator MOSFET could not be realized till 1960 and the word I always say even Shockley could not make in spite of all the intelligence he thinks he had. The first MOSFET as a concept was given by 19 by a Air Force officer of Royal Air Force of England Britain he actually filed his first patent the name is Lillienfield. He has a US patent on a MOS surfaces and MOS devices the MOS transistor was suggested by him he had a aluminum oxide as insulator aluminum as a back contact so is aluminum the top contact and he did actually suggest that this may amplify as far as history goes but actually this was not his original idea this idea was not so original because by the same time a German scientist Oscar Hill also was working on MOS and has actually has patented independent of Lillienfield in 1934 he had another patent on MOSFET but that was a European patent so no one knew that Lillienfield has a US patent so there was a fight between Hill this but also the fight couldn't last because no device could be made. What I meant by interface properties can be seen from here between see germanium and germanium oxide if this is your insulator the because they are two different materials they are all silicon to germanium to germanium oxide bonds are not satisfied and therefore there are some charges left which are called dangling bonds charges which are called interfacial charges and the net effect was they were shielding the effect of voltage on the gate material and not every voltage could be then put into induced into the semiconductor and then there was another problem people could immediately think that if the carriers have to move they should move fast but there are lot of scattering what we call carrier carrier scattering and also field scattering which allowed very small amount of current to flow so it was found when the first MOSFET was made drain current which was measured by then was several orders smaller than what was theory was expecting and as I keep saying this was even Shockley could not explain then though later time when suggested that the surface states have become interface states the states were actually explained by Shockley and therefore they were also called Shockley states. By 1960 the first MOSFET actually was made at Ballard's and the two people Kang and Attala this lady is not Attala by the way he is Mrs Kang they actually made the first MOS transistor and they made it out of silicon instead of germanium and they have SiO2 as the interface as the insulator and the interface between Si and SiO2 was far superior than germanium-germanium oxide there was very little shielding between insulator and gate and because of that the first MOS transistor worked a figure could be shown here this was the source area this was the drain area and in between this was the gate and that was covered by aluminium so this was a first MOSFET which appeared in 1960. Then we changed from metal gate to polysilicon gate to reduce the capacitances and we still kept on having SiO2 and silicon substrate for almost till 2000 or even today many Intel circuits or IBM circuits still have SiO2 as the insulator they have changed modified SiO2 with something else but silicon dioxide is still going strong but in 2010 probably or 2005 onwards we have shifted out of SiO2 and new high-K dialectics are coming. So typically using a gate field effect one can see from here the gate could then control the carriers below by the induction Gauss's law and then connect between source and drain and large current could flow this was the first MOS transistor way it was made successfully. Two types of MOSFETs were made one due to N type MOSFET which was in which the electron motion was possible because the if you apply a positive voltage on the gate the negative charges are induced and between N silicon source and N silicon drain electron channel could be created we started with the P substrate therefore it is called inversion layer electrons could move under the electric field laterally and constituted current. So we say N type because the carriers were electrons where in the case of P channel we started with N substrate we have a negative voltage on the gate to actually create inversion layer of holes between P silicon source to P silicon drain holes can move and therefore hole motion was also possible and we therefore declared this is P type P type MOSFETs. You will be wondering that I am talking about advanced VLSI course and why this because the first course which I gave was a web course in which all this was not actually provided. So I thought if you are reading VLSI design one and VLSI design two in continuation let us at least come back and show what things happened over the years. Yeah this is something same since we are looking for a MOSFET as a switch we found out it is much easier to actually say how it off and on. So we said if you apply a 0 volt on the gate let us say your N channel MOSFET going so you require positive Vg to create an inversion. So if I less than that particular voltage we call threshold voltage right now we put 0 voltage which is less than positive Vt then there is no channel between source and drain so therefore there is no current between source and drain so we say off state no drain current of course there is a leakage current but that is very small and we still declared as a off state. On the contrary if we apply gate voltage which is larger than threshold on the gate an inversion channel can be created of electrons, source drain if I apply voltage across this is like ohms law this is a resistor you are applying two contacts the drift current can flow and we call that as on state or one for the transistor. So on and off 0 and 1 could be created out of MOS switching and therefore it became the most important candidate for switching circuits or logic circuit. Over the years I again now summarize quickly the years through which we went through in technology first MOSFET transistor Linenfeld and Heil in 3525 CMOS 1960 but plagued with manufacturing problems please remember the first CMOS was attempted as early as 1960 it was not only PMOS it was actually they started both NMOS and PMOS together NMOS could never be turned on because it was always on. So we just could not create a CMOS circuit out of it was depletion mode was always created PMOS was then we continue to work with PMOS devices till late 70s most calculators available then was PMOS based circuits 1960 to 1970 then NMOS technology was improved and then we shifted from PMOS to NMOS because N channel has larger mobility because of electrons so more current and lower voltages and therefore NMOS replaced PMOS individual case for improvement in speed first Intel processor 4004 and 8080 were made in NMOS early 75. In 1980 when both PMOS and NMOS technology were well controlled the first combined device was made complementary MOS which has both P channel and channel together and this has much more advantage because it actually shows much lower power consumption or much lower power dissipation compared to either NMOS which is the more power consuming device than NMOS PMOS but both were power consuming device compared to when the CMOS is used. Then then we went from CMOS to other technologies like by CMOS we always thought that bipolar circuits were very fast compared to CMOS for many years the obvious reason was there was huge current in a bipolar transistor and a capacitor can be charged much faster therefore speed of a bipolar circuit is always larger at least till some years but it was found that the power is also very large to create that high speeds and our ultimate aim for all MOSFET improvements were larger speed at low power so people thought if we merge by bipolar technology with CMOS we may get advantages of both unfortunately did not happen as well it actually got disadvantages of both in a larger number than the advantages and except for very specific circuit some input circuit where TTL inputs are to be given by CMOS did not actually capture the imagination of most of the designers or most of the system designers we also shifted to high mobility materials like gallium arsenide we thought that silicon has limited electron mobility compared to any other material then we say okay look for higher mobility materials we went through gallium arsenide silicon germanium and even host of silicon nitride is being tried silicon carbide has been tried unfortunately none of them have as good an interface with insulator as a silicon and therefore silicon continued to remain as the benchmark material for all integrated circuit technology though I will show you at the end maybe if time permitting beyond move is silicon going to stay yeah possibly we also worked on SOI silicon on insulators we are also working on interconnect different interconnects like copper locate these are the new inventions in last 10 years aluminum was replaced by copper it's a very interesting technology thinking that if copper is known good better conductor both electricity wise as well as by thermally why at all we started with aluminum why not copper unfortunately copper for many years even now is called poison to silicon device because it actually creates the levels in the band gap of silicon which actually reduces the lifetime so copper was now never allowed probably if you come even to a lab like ours in 1985 which we made in IIT Bombay we had no copper lining copper tubing anywhere inside the lab because we were told that any moss or bipolar circuit will not work if you have a copper tubing of gas also coming so no question of putting directly on silicon however I there was a dam seen process came from the efforts of Texas instrument IBM I think everyone claimed I don't know who should be given credit they actually replaced copper and to protect copper from silicon they had some kind of a cladding around which was titanium nitride and using this they could then make an interconnect of copper which has a higher conductivity and therefore lower resistance and therefore higher speeds we also wanted to put more than one layer of metal interconnects so between the two metal interconnect lines we put low capacity low dielectric constant materials insulators like many of them HFC and many others were tried glass of course is the only one available earlier we try to see whether lower k materials can be used air of course is the best but you can't put two metal lines separated by air so one has to find some material which can give physical support 1970-71 the first generation of LSI has appeared this is Intel's 1103 DRAM you can see what kind of structure it has what kind of how it looks the first microprocessor in as I said earlier is Intel 404 404 and just for our Indian students I may say India's only manufacturing company which was semiconductor complex at Chandigarh somehow purchase the know-how to manufacture 4004 God knows why I don't want to say more than that so repeat performance what I said so far 1960 integrated circuit came so people started thinking how do we name improvement in number of transistor or technology so they say okay first one when we have appeared it has only 10 transistor maximum so we call integrated circuits I see then we say oh if you have more than 1000 kind of transistors we say it is a large scale integration in between there with this small scale medium scale integration then we say 1980 when we say we are larger than 10,000 transistors we started talking of very large scale integrated circuits by 90s this became some kind of a 10 1 million transistors then we say ultra large scale but in if it is one than more than a billion now as they are coming up probably we may call it I don't know what a giant large scale or whatever it is giant scale but so happened the designers did not like these technology names so they kept on calling anything beyond VLSI as VLSI and therefore it's stuck so even if now you have a billion transistor on chip it will still call a VLSI chip whereas technology people would not like it to call VLSI because for them VLSI means around 10,000 transistors alone so what is integration I just come back again you say okay there are multiple devices on one substrate and this question as I say is always asked how large is very large okay so we say okay small scale integration chips I would give numbers TTL has 74,000 series or 7400 series which typically has 10 to 100 transistors then you have 74,000 series from TTL which has around 100 transistors plus and we say okay this is a medium scale and when we actually moved away from Bipolar and we went for large scale 1000 to 10000 transistors we actually started on large scale and above a 10000 as I said we all started calling VLSI so even if now you have million transistors or billion transistors everything we say is very large scale if I do a course in VLSI design and if I do not utter a word or utter a name Gordon Moore probably I will be fooling myself this is our so-called demigod for VLS integrated circuits VLSI integrated circuits Gordon Moore was one of those who joined who left Shockley's company in this and then he joined along with Noyes and others at Fairchild and then started Intel with them he was a very I must say he must he was a very visionary person when the technology was being grown at Intel and earlier at Fairchild he figured out the way technology is improving on the same dyke because of the improvement I can see that components will doubling so first he thought it may double every 18 14 14 to 18 months but later on he said it may double every every year he say it will double the company a large scale you can say it will start doubling or essentially you can equivalently say exponentially with time the number of components will double on chip it was certainly I must say it was amazing visionary pronouncements and you can see 1980 itself according to what Moore thought you should have million transistor on the chip and yeah we did cross that barrier in 1980 so as if there is a joke going on in both technology group and design group that all of them are working to see that Moore is correct as if so we work strive very hard both technologically as well as design ways so that the Moore's law still is agreed to by almost everyone even if we do not reach what Moore thought then we say okay deviation from Moore's law but we kept on talking on Moore's law all through our careers and maybe you will also keep talking in your whole career till you work in the area of integrated circuits just to give some numbers of transistors the first Intel 4004 was 2300 transistors which was working at 1 megahertz clock ultra spark from sun which has 16 million transistors 2 gigahertz Intel p4 which appeared in 2001 has 42 million transistors and in 2003 HP's first PSS850 which has 140 million transistors currently many of the processors quad ones or ethylion or others from AMD all have more than 400 billion transistors on chip so here is the Moore's law okay so what essentially Moore's law has as I already said on your right right of the scale these are the number of transistors this is a billion number as I show this is sorry this is billion this is 10 billion so you can see the face of Gordon Moore I am happy to show you that yeah I have met him in one of the conference in US of course met many flittingly I do not think he knows me or I know him but it is a great thing to know who is this so called Gordon Moore so from 1970 till 2010 all the processors if you see are actually as per the number of transistor count as modern Moore suggested and they kept on following the years of course it is not 100 percent straight line linear everywhere a little bit slope change here but again it has started rising in a similar old fashion and therefore one can say Moore's law is back in full force for example Deer Cole Intel ethanium II processor which was announced in 2008 okay it has more than a billion transistors so Moore's law essentially was telling that the component density or transistor density will double every year was followed till very very late this is something two components shown on the same base one is memories the other is microprocessor why why people chose this because an integrated circuit or VLSI if someone asked you the marker what decides the technology node or what decides the progress the two devices we always discuss or two such different parameters we look for one is the improvement in speed and other performance of a microprocessor the second is memory how many how many bits memory you can create in a smaller number and what is the excess time so these are two though they are more made out of MOS transistors but their operation is different from each other one is purely based on logic the other is based mostly on the charge or discharge of a capacitor and because of that the progress of a MOS technology was always gauged based on the memory as well as microprocessor so the Moore's law if you apply on these two components uh microprocessor in this you can see not exactly one-to-one correspondence but yes by 2010 4G 4GB DRAM is available in the market which means the Moore's law is still getting followed okay whereas we have already reached ethaneum Pentium 4 quads so we are already crossing the numbers of transistor which was projected as early as 1970s and we are still going strong with it now question is always asked how long this will last well at the end of this first lecture of mine probably you may have some idea about where it will end if at all because if at all is a word I keep using because there is a statistical theory that there is no zero probability so one cannot say nothing will happen yeah I mean there is a probability may reduce but may still happen one doesn't know Moore figured out that by 2000 that his double every year law is not being followed so he projected that it's called Moore's second law he says okay double every two years first he said 18 months then make it 24 months and now he says okay every three years okay so all that he is now changing the slope but there is still a law which is being seen as Moore's law and people are trying to meet what Moore says so this is called same this is a graph between the years as well as the logic number of per chip or logic or gates per chip and you can see it is slightly separated with the same graph Moore's law I have been shown separately this is for memories and this is for processors one and you know this line is essentially why I this graph has been shown in the integrated circuit manufacturing one of the major worry right now is what we call how to print the small dimension if you reduce the dimension you have to print something on a silicon buffer of that dimension nanometers say 30 nanometers 20 nanometers 10 nanometers so when you print something you need process called lithography transferring image from one to the other now this is the limiting point right now we are still using photolithography which is called 193 nanometer process and maybe in 2010 and onwards or 12 onwards we may go for what we call extended UV process which is still not been available to any manufacturer Intel is working IBM is working Ti is working but that is not on the manufacturing though it has successfully been tried so the limitation now people are saying is not because of the DRAM thinking or this it may come essentially because of the lithography process may not allow you to go much smaller but all sudden done if there is a problem there is a solution and therefore I do not see why it will not occur this is same figure again so we what I am trying to show you on this our journey over the years have become micro to nano we started with dimensions which was 100 microns in 1970 I worked on a chip which was a transistor which has a 100 micron channel length in 1976 of course by a few years we have been to 10 microns but we start our first mass transistor in TIFR was just 100 micron length and we made it and it worked the importance it worked then we of course went to 10 microns in the next mask so in the 60s we started with 100 some microns and by 90s end of 90s or even early 2000 we are talking of a million transistor on chip so this is what we call really milestones we went from IC to VLSI to nano now so what essentially nutshell Moore said when you say double he says essentially says that when the size of a feature smallest feature on the chip reduces so he says every technology improvement will be improving it by 0.7 times it will reduce the number by last 0.7 times and which he now says every three years the new modern law he says 0.7x so for example if you are working on 90 nanometer technology earlier so 0.7 of that is 63 so now then we say next technology will be 65 nanometers if I multiply it by 0.765 it will be around 40 odd number so it is called 45 nanometer node if I multiply by 0.745 nanometers it will become 32 nanometer node if I multiply by 0.7 to 32 it will become 22 nanometers and that is how nodes were actually described and you can see the behind all this was Moore multiplied by 0.7. Now this as you increase the chip size now 16 percent every year and you reduce the size of the transistor 0.7 into 0.7 that is half of it obviously the number of transistor will increase because you are increasing the size and you are reducing the size of the transistor so obviously number of transistor will keep on increasing every year when the new chip will be appearing. Now the question came till 70s or 80s early 80s the designers used to say I want to put such a large system it will require a 10 million transistor or 8 million transistor but your technology cannot give so many transistor on chip by 2000 or even 1995 the reverse has happened there are number of transistors available can be as high as 800 millions or a billion transistors but there are no system which can actually implement all of them in one chip. So now the designers have more problem than the technology because designers do not know what to actually I should put they started putting 4 processor on chip so that oh I have now acquired it but actually one processor is anyway what you are using on the same chip you put 4 so 4 times but even then there are number of transistors available is much larger than most designers can even think. This is a good micrograph of Intel 4044 if you see how complex it looks these are if you see these are the pads you can see from here there are around 14 pads on a 4004 computer micro processor by the way it works I have worked on 4004. This is the current Pentium then it started in 80s ends of 80s Pentium II which is now it looks much more component density than what micro this earlier 4004 had of course please remember this everyone ask us in IIT at least I know why we are still teaching 8085 in my opinion 8085 micro processor is one of the basic micro processor architecture and any new architecture unless you change from what that architecture has it will follow same 8085 architecture in modified form and keep using it unless you go from sys to sys to is risk I do not know how changes can be made otherwise there is a standard procedure of actually executing data and as long as that happens 8085 remains our workhorse however we to improve the speed to improve the functionality of the micro processor improvements were made and the one of the major improvement is now coming is availability of large amount of cash on chip and we will show you this later. So, this is Pentium II this is what all of you are working right now on your desktops in Pentium IV micro processor 2 gigahertz and now of course 3.4 gigahertz and soon it will be 4.8 gigahertz. So, the kind of you can see from here any structure which is looks blackish because they are identical these are actually cash and now one believes that there will be a huge cash area on the micro processor rather than the controller part or the chip register part or ALU part the major decision of doing things will rest on how much cash we have and how fast we can take data out of cash and put to processor and back to cash. So, this kind of newer way of doing faster analysis will come and that is the only improvement one expects in coming years. Last slide for day these are typical commercial memories we have a 70 MB Intel SRAM which is my on the left this half of here and half down then this is Samsung's 2 GB DRAM. Of course, I have a photograph of 4 GB, but I think this was better. So, I put I have 2 GB photograph you can see this is this is 2 GB DRAM. So, if I can put 2 GB, 2 GB, 2 GB, 2 GB you can see there is a 918 GB DRAM and then if I even I put 2 then I can actually improve the speed because I can share the work and then we will call DDRAMs. It is essentially same it is a fast DRAMs as the word went, but essentially it is a dual DDRAM and because of that the speed improves. So, the idea is now to parallel so many DRAMs and may create sooner or later 64 GB DRAM itself. The another memory which is very very dominantly used by most logic systems or electronic system uses what we called as a NAND ROMs. ROMs does not require power ROMs for retaining a data whereas DRAMs SRAM do require power. So, this was a another area where much of the research went through and 8 GB NAND ROM is what is now marketed and people believe that sooner of course their speed is not as close to SRAM, but closer to DRAMs. So, maybe they will first replace DRAM may replace SRAM and DRAMs will be replaced by NAND ROMs sooner SRAM may not remain or SRAM will never be called SRAM, but will be called RAMs alone. So, I coming back to this slide again. So, at the end of the day I must tell you that whatever can be manufactured is only can be done. The manufacturers only look for profits because after all they are invested money and any system to be manufactured they first should actually find out that this system is going to be in what larger product, what is the window of that product in which it is going to be marketed and if what is the performance that system requires from your IC chip and if you cannot produce within that window time probably your whole product will not be of any consequence. And therefore, whatever people keep saying in design it is a very good design. I always say there is nothing called very good design. Any design which can go into a system and gives money to the manufacturer is the best design. Thank you very much for the day. We will come back on the next time and continue with this remainder part and also give you the more details about the course and my other colleagues. Thank you for the day.