A memristor is a passive two-terminal electronic component for which the resistance (dV/dI) depends in some way on the amount of charge that has flowed through the circuit. When current flows in one direction through the device, the resistance increases; and when current flows in the opposite direction, the resistance decreases. When the current is stopped, the component retains the last resistance that it had, and when the flow of charge starts again, the resistance of the circuit will be what it was when it was last active.
Memristor theory was formulated and named by Leon Chua in a 1971 paper. Leon Chua is to circuit theory as Einstien is to relativity. Thirty-seven years after he predicted the memristor, a working solid-state memristor was created by a team led by R. Stanley Williams at Hewlett Packard. Here's an informative video of R. Stanley Williams (http://www.youtube.com/watch?v=bKGhvKyjgLY) giving a keynote presentation on memristor technology at the UC San Diego Center for Networked System's Winter Research Review 2010.
In 2009, Massimiliano Di Ventra, Yuriy Pershin and Leon Chua co-wrote an article extending the notion of memristive systems to capacitive and inductive elements in the form of memcapacitors and meminductors whose properties depend on the state and history of the system.
Gravock
Homemade Memresistor with test circuit (http://sparkbangbuzz.com/memristor/memristor.htm)
Gravock
Quote from: gravityblock on March 02, 2016, 01:26:07 PM
Homemade Memresistor with test circuit (http://sparkbangbuzz.com/memristor/memristor.htm)
Gravock
A Tip On Making Nyle Steiner's Homemade Memristor (http://www.angelfire.com/oh3/ebjoew/Memristor.html)
Gravock
Memristor simulation is possible with the Falstad Circuit Simulator (http://www.falstad.com/circuit).
Gravock
HP abandoned memristors and took it off their roadmap for their super computer "The Machine". Everyone else has been focusing on ReRAM but I think the jury is still out on if this is just another memristor variant (HP wants to say it is so the millions they flushed down the drain on Memristors didn't go to waste). Intel is now focusing on 3D Xpoint although very little is known about it and based on what has been released it looks like another CBRAM variant that is apparently 1000x NAND and higher density then DRAM.
Quote from: Nink on March 02, 2016, 02:44:30 PM
HP abandoned memristors and took it off their roadmap for their super computer "The Machine". Everyone else has been focusing on ReRAM but I think the jury is still out on if this is just another memristor variant (HP wants to say it is so the millions they flushed down the drain on Memristors didn't go to waste). Intel is now focusing on 3D Xpoint although very little is known about it and based on what has been released it looks like another CBRAM variant that is apparently 1000x NAND and higher density then DRAM.
Two letters for you at this time: AI
Gravock
Not sure memristors have the synaptic capabilities everyone was hoping for and I don't know any research teams focused on neuromorphic (cognitive) computing who are seriously considering memristors in the AI field. But you can prove me wrong in 10 years.
Quote from: Nink on March 02, 2016, 08:22:15 PM
Not sure memristors have the synaptic capabilities everyone was hoping for and I don't know any research teams focused on neuromorphic (cognitive) computing who are seriously considering memristors in the AI field. But you can prove me wrong in 10 years.
No, I'll prove you wrong right now, and not in 10 years. You are worst than wrong as always, lol.
Researchers have developed a new kind of neural circuit that uses memristor technology to replicate the complex human brain. The "intelligent" circuit was able to perform some human tasks that computers usually struggle at, such as image classification (see image and reference link below).
Reference: Memristor circuit recreates the brain and carries out human tasks (http://www.digitaljournal.com/technology/cutting-edge-circuit-emulates-the-brain-to-make-human-decisions/article/433171)
Gravock
Here's a video demonstration (http://www.youtube.com/watch?v=3ZRIPdr1lug) in how a coherer (http://en.wikipedia.org/wiki/Coherer) can be configured as a memristor.
The basis for the operation of the coherer is that the metal particles (beads in this case) cohere or cling together and conduct electricity much better after being subjected to radio frequencies. The radio signal from the antenna is applied directly across the coherer's electrodes. When the radio signal from a "dot" or "dash" came in, the coherer would become conductive. The coherer's electrodes were also attached to a DC circuit powered by a battery that created a "click" sound in earphones or a telegraph sounder, or a mark on a paper tape, to record the signal. Unfortunately, the reduction in the coherer's electrical resistance persisted after the radio signal was removed. This was a problem because the coherer had to be ready immediately to receive the next "dot" or "dash". Therefore a decoherer mechanism was added, to tap the coherer, mechanically disturbing the particles to reset it to the high resistance state.
Coherence of particles by radio waves is an obscure phenomenon that is not well understood even today. Recent experiments with particle coherers seem to have confirmed the hypothesis that the particles cohere by a micro-weld phenomenon caused by radio frequencies flowing across the small contact area between particles. The underlying principle of so-called "imperfect contact" coherers is also not well understood, but may involve a kind of tunneling of charge carriers across an imperfect junction between conductors.
Gravock
Thank you Gravock for the interesting posts. I am surprised you don't have a few more people commenting about the information you have shared. I am not sure how I would make use of the memristor but I am old enough to know that any knowledge gained may come in handy at a later time. It is also interesting that this simple device can be made at home pretty easily. Do you have any simple applications for it since I don't think I will be building any neural computers any time soon.
Carroll
Quote from: gravityblock on March 02, 2016, 11:37:00 PM
No, I'll prove you wrong right now, and not in 10 years. You are worst than wrong as always, lol.
Researchers have developed a new kind of neural circuit that uses memristor technology to replicate the complex human brain. The "intelligent" circuit was able to perform some human tasks that computers usually struggle at, such as image classification (see image and reference link below).
Reference: Memristor circuit recreates the brain and carries out human tasks (http://www.digitaljournal.com/technology/cutting-edge-circuit-emulates-the-brain-to-make-human-decisions/article/433171)
Gravock
You are kidding me, a 100 Synapse memristor created by university students is your proof. As I said NO SERIOUS RESEARCHERS are considering memristor as solution for cognitive computing.
I am suspecting this is just an out of work HP researcher who probably spent the last 10 years of his life on memristors joined UC and convinced his students to work on his abandoned project.
IBM Synapse chips are capable of 46 BILLION SYNATPIC operations per second but as I said you come back to me in 10 years and show me where memristor technology is.
Quote from: Nink on March 03, 2016, 08:43:20 PM
You are kidding me, a 100 Synapse memristor created by university students is your proof. As I said NO SERIOUS RESEARCHERS are considering memristor as solution for cognitive computing.
I am suspecting this is just an out of work HP researcher who probably spent the last 10 years of his life on memristors joined UC and convinced his students to work on his abandoned project.
IBM Synapse chips are capable of 46 BILLION SYNATPIC operations per second but as I said you come back to me in 10 years and show me where memristor technology is.
Most of the research in memristors in regards to AI is jointly funded by the SyNAPSE program (DARPA) and the National Science Foundation (NSF), thus DARPA is yielding the benefits of this research. The project is primarily contracted to IBM and HRL who in turn subcontract parts of the research to various US universities. So, it should come as no surprise that a 100 Syanapse memristor was created by university students, since this is where the funding is being funneled into.
Reference: DARPA's SyNAPSE Program (http://www.artificialbrains.com/darpa-synapse-program)
Gravock
Quote from: gravityblock on March 03, 2016, 11:27:43 PM
Most of the research in memristors in regards to AI is jointly funded by the SyNAPSE program (DARPA) and the National Science Foundation (NSF), thus DARPA is yielding the benefits of this research. The project is primarily contracted to IBM and HRL who in turn subcontract parts of the research to various US universities. So, it should come as no surprise that a 100 Syanapse memristor was created by university students, since this is where the funding is being funneled into.
Reference: DARPA's SyNAPSE Program (http://www.artificialbrains.com/darpa-synapse-program)
Gravock
HP was also originally part of the team that received funding from DARPA but they never even made it passed phase zero feasibility study. I recall HRL received something like 10 or 11 Million for phase 1 so HP convinced them to continue down their memristor path. ~40 Million dollars later HRL pumped into memristor tech funded by DARPA and it never really amounted to anything so after 8 years of R&D all you have to show for it is your 100 Synapse memristor chip.
If Moores law kicks in and DARPA and others invest 100's of Millions in memristor tech your 100 Synapse memristor chip will have scaled to a whopping 3200 Synapse over the next 10 years.
Quote from: Nink on March 04, 2016, 12:18:28 AM
If Moores law kicks in and DARPA and others invest 100's of Millions in memristor tech your 100 Synapse memristor chip will have scaled to a whopping 3200 Synapse over the next 10 years.
It appears DARPA doesn't agree with you, and I don't either.
Gravock
Quote from: gravityblock on March 04, 2016, 12:46:38 AM
It appears DARPA doesn't agree with you, and I don't either.
Gravock
You are right DARPA does not believe that they will even reach 3200 Snyapse in 10 years and I don't either.
They are not stupid enough to invest any more money in memristors. Game over
Quote from: Nink on March 04, 2016, 01:21:42 AM
You are right DARPA does not believe that they will even reach 3200 Snyapse in 10 years and I don't either.
They are not stupid enough to invest any more money in memristors. Game over
Please provide a reference to your false assertion above.
Graock
DARPA had a budget of 100M they gave IBM ~56 Million and HP / HRL ~ 42M. The project wound down last year and the rest of the money went to various awards to universities etc. Most of this info should be public knowledge. The problem was memristors is they don't scale as you need a 30+ nm air gap for each memristor where IBM can use FinFET technology that can scale the entire synapse chip down to 14nm and => 5nm around 2020
Please provide a reference to show that DARPA is investing in Memristor after the SyNAPSE project was completed.
I personally don't think there is a very good future for these hybrid circuit devices. Quantum
Computing is just about as hard circuit wise and its capabilities can't be touched by any other
computational mechanism especially for those used in AI. Similar to free energy, Quantum
Computing is so computationally disruptive that it may not be allowed to grow freely but that
doesn't mean its shadow won't be felt competitively. If AI can be based in quantum computers
and natural intelligence doesn't already use it, humans may wish to stand back. These are just
my opinions.
Quote from: mscoffman on March 04, 2016, 03:27:58 PM
I personally don't think there is a very good future for these hybrid circuit devices. Quantum
Computing is just about as hard circuit wise and its capabilities can't be touched by any other
computational mechanism especially for those used in AI. Similar to free energy, Quantum
Computing is so computationally disruptive that it may not be allowed to grow freely but that
doesn't mean its shadow won't be felt competitively. If AI can be based in quantum computers
and natural intelligence doesn't already use it, humans may wish to stand back. These are just
my opinions.
I agree. Quantum is probably the future of AI but it is really hard to separate fact from fiction on the current state. D Wave is boasting a 1000 qubit but I personally think it is all smoke and mirrors and all they provide is quantum simulation. If they were really true quantum computers with 1000 qubits all existing simultaneously in multiple states, at 1000 qubits they could break every encryption system on the planet. IARPA (DARPA baby brother) commissioned IBM to work on solving this problem last year but I think we need to clearly define just what the definition of a multi state quantum computer actually is before we can agree that anyone has actually developed even a true single qubit quantum computer.
Quote from: Nink on March 04, 2016, 08:00:27 AM
DARPA had a budget of 100M they gave IBM ~56 Million and HP / HRL ~ 42M. The project wound down last year and the rest of the money went to various awards to universities etc. Most of this info should be public knowledge. The problem was memristors is they don't scale as you need a 30+ nm air gap for each memristor where IBM can use FinFET technology that can scale the entire synapse chip down to 14nm and => 5nm around 2020
Please provide a reference to show that DARPA is investing in Memristor after the SyNAPSE project was completed.
I asked for a reference for your assertion that DARPA doesn't believe they'll even reach 3200 Snyapses in 10 years, and your reply above is on DARPA's budget, scale-ability, etc. I never made the claim that DARPA is investing in memristors after the SyNAPSE project, so you asking for a reference is a red herring.
Knowm's new memristor improves efficient AI processors (https://thestack.com/world/2016/01/21/knowms-new-memristor-improves-efficient-ai-processors/)
Gravock
Knowm - Natures transistor (http://knowm.org/)
Gravack
Quote from: gravityblock on March 05, 2016, 01:23:05 AM
I asked for a reference for your assertion that DARPA doesn't believe they'll even reach 3200 Snyapses in 10 years, and your reply above is on DARPA's budget, scale-ability, etc. I never made the claim that DARPA is investing in memristors after the SyNAPSE project, so you asking for a reference is a red herring.
Yes you did but when you made that claim you didn't actually realize the funding had ended. Now you are trying to back away from that claim. DARPA has funded Memristors in the past but they do not currently fund memristors
Quote from: gravityblock on March 03, 2016, 11:27:43 PM
Most of the research in memristors in regards to AI is jointly funded by the SyNAPSE program (DARPA)
Moores law is agreed upon by all IT companies and still holds true today after 50 years. DARPA understands Moores Law hence the reason their projects span 5 to 10 years. The dependency for Moores law to be maintained is both time and money. DARPA is no longer investing in Memristors (and as a publicly funded organization if they were there would be a public record of that investment) and no one else is. except unemployed startups who waisted 10 years of their life on memristors are now begging for money to keep this obsolete, failed technology alive. Even HP who invented memristor abandon the project after DARPA ended the funding in 2015 and HP are now off working with SanDisk on a ReRAM variant. HP are playing up their investment in Memristors as being a contributor but their is actually no Memristor technology involved in the new SCM SanDisk is developing. All HP is bringing to the table is $.
So best case for memristors is Moores Law and thats 3200 Synapse per IC in 10 years but the probable case is the startups will get zero funding and the technology will go the way of bubble memory.
If you want a link to Moores law I suggest you try google. If no one is investing in Memristors (including DARPA) then not even Moores law will be maintained.
Quote from: Nink on March 05, 2016, 08:36:59 AM
Yes you did but when you made that claim you didn't actually realize the funding had ended. Now you are trying to back away from that claim. DARPA has funded Memristors in the past but they do not currently fund memristors
Moores law is agreed upon by all IT companies and still holds true today after 50 years. DARPA understands Moores Law hence the reason their projects span 5 to 10 years. The dependency for Moores law to be maintained is both time and money. DARPA is no longer investing in Memristors (and as a publicly funded organization if they were there would be a public record of that investment) and no one else is. except unemployed startups who waisted 10 years of their life on memristors are now begging for money to keep this obsolete, failed technology alive. Even HP who invented memristor abandon the project after DARPA ended the funding in 2015 and HP are now off working with SanDisk on a ReRAM variant. HP are playing up their investment in Memristors as being a contributor but their is actually no Memristor technology involved in the new SCM SanDisk is developing. All HP is bringing to the table is $.
So best case for memristors is Moores Law and thats 3200 Synapse per IC in 10 years but the probable case is the startups will get zero funding and the technology will go the way of bubble memory.
If you want a link to Moores law I suggest you try google. If no one is investing in Memristors (including DARPA) then not even Moores law will be maintained.
Memristors are necessary for a continuation of moores law (http://www.scientificamerican.com/article/moores-law-computing-after-moores-law/). As Pavlus explains, memristors make it possible to combine storage and random-access memory. "The common metaphor of the CPU as a computer's 'brain' would become more accurate with memristors instead of transistors because the former actually work more like neurons—they transmit and encode information as well as store it," he writes. Memristors can transmit, encode, and store it by working more like neurons. Brains don't separate processor from memory, as we do with a CPU and RAM.
Knowm came right out of DARPA, and the co-creator of DARPA SyNAPSE founded Knowm. The co-creator of DARPA, Alex Nugent (Knowm), has been awarded government SBIR and STTR contracts to furter develop the techonology (reference article (http://knowm.org/technology/)).
Knowm Inc. exists to fill a niche in the rapidly evolving technological landscape and lead the computing industry toward neuromemristive processors. The roots of Knowm Inc. were planted in 2002, when lead inventor
Alex Nugent began patenting his ideas around adaptive computing architectures and founded an intellectual property holding company called KnowmTech. Initial seed funding for the endeavor was made possible by business woman and entrepreneur, Hillary Riggs. The portfolio now includes over 40 patents spanning memristive components and circuits all the way to large scale neuromorphic architectures.
Alex Nugent co-created and advised the DARPA SyNAPSE program and more recently has been awarded government SBIR and STTR contracts to further develop the technology. In 2012, physicist, electrical engineer and software developer, Tim Molter, joined the effort to lead software development and further design chip architectures with Alex. This collaborative effort lead to the publication of the formal introduction to AHaH computing in early 2014: AHaH Computing–From Metastable Switches to Attractors to Machine Learning. More recently, Knowm has partnered with key experts in the field to further ramp up efforts on a path to commercialization. Collaborators include memristor fabrication pioneer Kris Campbell, Ph.D. from Boise State University and memristor circuit designer pioneer Dhireesha Kudithipudi, Ph.D. from Rochester Institute of Technology. Most recently, investor and consultant Sam Barakat has joined the team to help launch Knowm Inc.
Gravock
Quote from: Nink on March 05, 2016, 08:36:59 AM
If no one is investing in Memristors (including DARPA) then not even Moores law will be maintained.
As you can see in my previous post, memristors are currently being developed and commercialized by the co-creator of the DARPA SyNAPSE project through Knowm, thus the continuation of Moores law.
Gravcok
Exactly as I expected some out of work folks who lost their jobs after the DARPA project finished up and wasted 10 years of their life on memristors begging for money to continue the project on obsolete tech. Maybe you will prove me wrong in 10 years and they will find some suckers and burn through another $40M to reach 3200 Synapse by 2025 in accordance with Moores law but I highly doubt it.
By the way Alex Nugent actually received that money between in 2011 and 2013 and he burnt through 1.5M and has nothing to show for it.
Quote from: Nink on March 05, 2016, 02:43:30 PM
Exactly as I expected some out of work folks who lost their jobs after the DARPA project finished up and wasted 10 years of their life on memristors begging for money to continue the project on obsolete tech. Maybe you will prove me wrong in 10 years and they will find some suckers and burn through another $40M to reach 3200 Synapse by 2025 in accordance with Moores law but I highly doubt it.
The co-creator of DARPA SyNAPSE is the founder of Knowm, and he didn't lose his job after the DARPA SyNAPSE program finished. He was awarded government contracts to further develop the technology after the DARPA project finished, which he co-created.
Alex Nugent ---> Knowm ---> DARPA SyNAPSE ---> Alex Nugent / Knowm ---> Government contracts awarded to and commercialization of memristors through Knowm.
Memristors are necessary for the continuation of Moores law in computing. This continuation of Moores law in computing is contrary to your unreferenced 3200 Synapses by 2025 with memristors.
Gravock
Quote from: gravityblock on March 05, 2016, 03:16:38 PM
The co-creator of DARPA SyNAPSE is the founder of Knowm, and he didn't lose his job after the DARPA SyNAPSE program finished. He was awarded government contracts to further develop the technology after the DARPA project finished, which he co-created.
Alex Nugent ---> Knowm ---> DARPA SyNAPSE ---> Alex Nugent / Knowm ---> Government contracts awarded to and commercialization of memristors through Knowm.
Memristors are necessary for the continuation of Moores law in computing. This continuation of Moores law in computing is contrary to your unreferenced 3200 Synapses by 2025 with memristors.
Gravock
Sorry I thought you knew how to count. Here is moores law doubling every 2 years.
2015 100 Synapse
2017 200 Synapse
2019 400 Synapse
2021 800 Synapse
2023 1600 Synapse
2025 3200 Synapse
Quote from: Nink on March 05, 2016, 03:21:11 PM
Sorry I thought you new how to count. Here is moores law doubling every 2 years.
2015 100 Synapse
2017 200 Synapse
2019 400 Synapse
2021 800 Synapse
2023 1600 Synapse
2025 3200 Synapse
Wrong, memristors is necessary for the continuation of Moores law in computing, and 3200 Synapses by 2025 with memristors isn't a continuation of Moores law in computing. For one, your 3200 synapses by 2025 is based on the 100 synapses created by the university students in 2015. The 100 synapses created by the students was a proof of concept and in no way represented Moores law for memristors in 2015. The students could have easily created 1000 synapses in their lab. 100 synapses was all that was required for their proof of concept.
Gravock
Then why didn't they create 1000 Synapses. I will tell you why because it doesn't scale, the fact they managed to squeeze 100 Synapse on single die is quite impressive and the only reason they achieved it was because they had an out of work ex HP memristor research employee who had to take on a teaching role and used this for further development on a project that he no longer had funding for.
BTW I see you did not address the fact that no one is actually funding memristors. There are just a bunch of people looking for funding but no one actually funding development so as I keep saying even 3200 is optimistic at best. I am looking for funding to build a cold fusion reactor, it does not mean it will actually work it just means I am just looking for funding.
Quote from: Nink on March 05, 2016, 04:29:19 PM
Then why didn't they create 1000 Synapses. I will tell you why because it doesn't scale, the fact they managed to squeeze 100 Synapse on single die is quite impressive and the only reason they achieved it was because they had an out of work ex HP memristor research employee who had to take on a teaching role and used this for further development on a project that he no longer had funding for.
BTW I see you did not address the fact that no one is actually funding memristors. There are just a bunch of people looking for funding but no one actually funding development so as I keep saying even 3200 is optimistic at best. I am looking for funding to build a cold fusion reactor, it does not mean it will actually work it just means I am just looking for funding.
They didn't create 1000 snyapses because only 100 was needed for their proof of concept. You don't build two houses if you're only going to utilize 1 house.
Funding for memristors is now being funneled into Known in the form of government contracts and commercialization (taxation) and profits for further research and development. In other-words, the taxpayer is funding memristors through government contracts and profits from consumer purchases.
Gravock
Quote from: gravityblock on March 05, 2016, 04:48:38 PM
They didn't create 1000 snyapses because only 100 was needed for their proof of concept. You don't build two houses if you're only going to utilize 1 house.
Funding for memristors is now being funneled into Known in the form of government contracts and commercialization (taxation) for further development. In other-words, the taxpayer is funding memristors through government contracts with Knowm.
Gravock
The funding Nugent had ended in 2013 I already told you that and you ignored it. Commercially all knowm have come up with was 8 memristors on a single IC
Quote from: Nink on March 05, 2016, 04:56:29 PM
The funding Nugent had ended in 2013 I already told you that and you ignored it. Commercially all knowm have come up with was 8 memristors on a single IC
Look at the timeline below. The eight, 8, discrete memristors in a 16 Pin Ceramic DIP IC is a proof of concept and in no way represents Moores law in memristors. Lets have a look at the raw research die (http://knowm.org/downloads/Knowm_Memristors.pdf) (second image below). The research die was created to allow study of device operation over a wide range of device sizes. The die are 7860μm by 5760μm and consists of 9 columns of devices, each corresponding to a different device size. The size is listed at the bottom of each column. Each column contains 20 rows of each device size per column, for a total of 180 devices. They're available (http://overunity.com/knowm.org/product/bs-af-w-memristors) for research and development. As you can see, they didn't try to cram as many memristors onto a single IC that is currently possible. 180 device sizes with 180 memristors is sufficient to allow for the study of device operations over a wide range of device sizes. The raw research die itself is much smaller than the package of the IC. Once again, this doesn't represent moores law in memristors.
Gravock
So you agree they have no funding and the best they have made is an 8 bit Memristor. If we use them as a baseline the best they can hope for is 256 bits no parity. I was generous and took UC lab results and stated 3200 Synapse but as everyone knows what is achievable in a lab, in no way reflects anywhere close to what is commercially achievable.
Edit
I guess your silence is agreement. Get back to me in 10 years.
Quote from: Nink on March 05, 2016, 07:14:45 PM
So you agree they have no funding and the best they have made is an 8 bit Memristor. If we use them as a baseline the best we can hope for is 256 bits or 8 bytes no parity. I was generous and took UC lab results that no in way reflects anywhere close to what is commercially achievable.
No, I don't agree with you. There are 180 devices on the raw research die of varying sizes ranging from 1 to 30 micrometers with 180 memristors on it (see images below). Each device has 1 memristor. By taking the smallest device consisting of 1 memristor, which is 1 square micrometer (see image below), then it is theoretically possible to place at least 45,270,000 devices with a memristor on a 45.28 mm
2 die. This isn't taking into account half of the surface area of the die that isn't being utilized, so we'll say this unused surface area will make up the space between the 45,270,000 memristor devices (5,658,750 byte memristor). The current die is a proof of concept for research and develepment due to popular demand and the number of memristors on it doesn't reflect how many memristors can be placed on the die.
Gravock
Quote from: Nink on March 05, 2016, 07:14:45 PM
Edit
I guess your silence is agreement. Get back to me in 10 years.
My 3 hour of silence was because I had to drive out to the lake house (2 hour drive round trip) to bring a few things (stove, small refrigerator, misc. items) back to the house.
Gravock
Quote from: gravityblock on March 05, 2016, 10:14:48 PM
No, I don't agree with you. There are 180 devices on the raw research die of varying sizes ranging from 1 to 30 micrometers with 180 memristors on it (see images below). Each device has 1 memristor. By taking the smallest device consisting of 1 memristor, which is 1 square micrometer (see image below), then it is theoretically possible to place at least 45,270,000 devices with a memristor on a 45.28 mm2 die. This isn't taking into account half of the surface area of the die that isn't being utilized, so we'll say this unused surface area will make up the space between the 45,270,000 memristor devices (5,658,750 byte memristor). The current die is a proof of concept for research and develepment due to popular demand and the number of memristors on it doesn't reflect how many memristors can be placed on the die.
Gravock
It is now obvious that you have absolutely no understanding of what a wafer of raw dies is and how the semiconducter device packaging process actually works.
All I can do now is wish you luck with your memristor recommendations and future investments in memristor technology.
Quote from: Nink on March 05, 2016, 11:43:49 PM
It is now obvious that you have absolutely no understanding of what a wafer of raw dies is and how the semiconducter device packaging process actually works.
All I can do now is wish you luck with your memristor recommendations and future investments in memristor technology.
ROFLMAO!!! The above isn't a scientific or mathematical rebuttal. The 16 pin IC has 8 devices with 8 memristors (1 memristor per device) and the raw research die has 180 devices with 180 memristors (1 memristor per device). The IC and the raw research die don't correlate with each other in regards to the number of devices and memristors. The number of devices and memristors, along with the theoretical limits of such, has nothing to do with how the semiconductor device packaging process actually works. Another red herring from you.
Gravock
Reference: (CNN) So long, transistor: How the 'memristor' could revolutionize electronics (http://www.cnn.com/2015/02/26/tech/mci-eth-memristor/)
In a transistor, once the flow of electrons is interrupted by, say, cutting the power, all information is lost. But a memristor can remember the amount of charge that was flowing through it, and much like a memory stick it will retain the data even when the power is turned off.
This can pave the way for computers that will instantly turn on and off like a light bulb and never lose data: the RAM, or memory, will no longer be erased when the machine is turned off, without the need to save anything to hard drives as with current technology. But memristors have another fundamental difference compared with transistors: they can escape the boundaries of binary code.
There is a physical limit to the number of transistors that we can pack on a chip, and we are already approaching the miniaturization threshold of this technology. It is inevitable that, one day, we will need to move away from silicon based computing.
The memristor technology is a candidate for this crucial step: "It could mean the end of the silicon era, giving us lower power consumption, the ability to compute more information, increased data storage and completely new logic patterns for our computers," says Rupp.
Memristors don't require a silicon layer and different materials can be used as a substrate. This could create a new class of microchips, that could eventually be integrated in everyday items such as windows, clothes or even coffee cups. Once again, this has nothing to do with how the semiconductor device packaging process actually works.
Gravock
Meet The Memcomputer: The Brain-Like Alternative to Quantum Computing (http://www.popularmechanics.com/technology/a16308/memcomputer-quantum-computing-alternative/)
Di Ventra explains that this very sort of problem, which could take our most advanced computer decades, could take a memcomputer only seconds.
Gravock
How a single memristor can perform all logic operations via spike operations and how a memristor with two detectable hysteresis curves can be used to store more than two values (https://hackaday.io/project/7344-memristor-computing).
Gravock
Quote from: gravityblock on March 06, 2016, 03:02:54 PM
How a single memristor can perform all logic operations via spike operations and how a memristor with two detectable hysteresis curves can be used to store more than two values (https://hackaday.io/project/7344-memristor-computing).
Gravock
IMPLY based logic functions with a memristor, awesome! All of these functions so far rely on the logical IMPLY function, which can be achieved by applying a read/condition voltage to one memristor and a write/set voltage to another. As the IMPLY function is a universal logic functions, all logic functions can be derived from a series of condition and set voltages. A single memristor to perform all logic operations! How many transistors would that take, LOL? Transistor count for generic logic functions based on static CMOS implementation (https://en.wikipedia.org/wiki/Transistor_count#Logic_functions)
Gravock
Quote from: Nink on March 02, 2016, 08:22:15 PM
Not sure memristors have the synaptic capabilities everyone was hoping for and I don't know any research teams focused on neuromorphic (cognitive) computing who are seriously considering memristors in the AI field. But you can prove me wrong in 10 years.
First programmable memristor computer (https://news.umich.edu/first-programmable-memristor-computer-aims-to-bring-ai-processing-down-from-the-cloud/) aims to bring AI processing down from the cloud! The University of Michigan proved you wrong in only 3 years!
Gravock
Quote from: Nink on March 04, 2016, 12:18:28 AM
HP was also originally part of the team that received funding from DARPA but they never even made it passed phase zero feasibility study. I recall HRL received something like 10 or 11 Million for phase 1 so HP convinced them to continue down their memristor path. ~40 Million dollars later HRL pumped into memristor tech funded by DARPA and it never really amounted to anything so after 8 years of R&D all you have to show for it is your 100 Synapse memristor chip.
If Moores law kicks in and DARPA and others invest 100's of Millions in memristor tech your 100 Synapse memristor chip will have scaled to a whopping 3200 Synapse over the next 10 years.
Wrong again! 5,800 Synapses in only 3 years. Please note: In the experimental-scale memristor computer, there were more than 5,800 memristors. A commercial design could include millions of them.
Gravock