rbtbob: Too bad what's left of Ovonyx has been already exiled to a university basement and the "company" will file for Chapter 11 by the end of 2012. Those robots will be quite funny since they will be slow memorizing, extremely unreliable, and will always run hot!
You are still commenting on PCM. Since you are so sure that PCM will never be commercialized in volume why do you continue to post so much about it. Who do you feel you need to refute?
If you are indeed correct the current ENER stock price is proof enough.
What is your motive here Robert?
Tyler Lowrey (Ovonyx) was issued patent 8,053,753 on 8 November for a complete thin film device with PCM and logic based on chalcogenide threshold switches. I believe these thin film computers will greatly advance the creation of robotic personal assistants. It will be a fairly straight-forward tweak to build in the cognitive functions recently documented here on EE Times. I hope Stan Ovshinsky lives long enough to see these robots demonstrate the culmination of his extraordinary work.
Mr Non Volatile Memory:-I think given the right product and demand, manufacturing costs can in most cases be reduced as part of the normal learning curve.
Unless, that is, those manufacturing costs are related to yield and linked to problems of a fundamental nature. If the VLSI2010 paper on a high performance PRAM cell (17 x 7.5nm)from Samsung is the precursor for the structure used in the 8G-bit, what I consider are unsafe extrapolations and conclusions in that work gives rise to my lack of optimism as far as the 8G-bit device is concerned. In that cell the calculated reset current density was 6.2 x10E+7Amps/sq-cm, it will be very interesting to see how Samsung have dealt with that as far as both matrix isolating devices and electro-migration as well as element separation in the PCM material from whatever its causes (J or E linked through the electrical conductivity of molten chalcogenide).
R G.Neale: Significant performance improvement is not sufficient. For volume manufacturing, costs matter as well. PCM has consistently lagged not only on the performance curve but also on the cost curve. PCM will never be commercialized in volume, so you are right to be less than optimistic. After all, more than 40 years have passed since you wrote the seminal paper on PCM showing that unlucky 256 bit (or was it 512 bit) sample!
To date (including the one to be discussed at IEDM 2012 in a week or so) there will have been four 1 G-bit PCM memory chips from different manufacturers. I think in order to judge the future prospects for PCM, rather than requesting an answer to the question, where can I buy a 1G-bit PCM, it is more important to know the real and true reasons why you cannot buy one.
My view is there are two essential needs for the success of the introduction of any new memory technology; at any point in time it must offer an order of magnitude performance improvement over what is already available. Of even greater necessity, and to build confidence is the need for a continuity of products of increasing bit density and performance, accompanied by design wins. In the product continuity steps for monolithic PCM from 512M-bit to 8M-bits, the missing steps of devices with bit capacity of 1G-bit and 4G-bit is unusual to say the least. It maybe the powers that be at Samsung recognized they have slipped too far behind and decided that it has to be 8G-bit now or nothing. In the past ISSCC was always made more exciting waiting for the product announcement a week or so after the scientific paper was presented. Therefore we might have two bits of PCM news to look forward a 1G-bit product announced by Hynix, followed by the 8G-bit from Samsung in April. Personally on that count, based on the history of PCM developments todate I am less than optimistic.
Since they are going to display the PCM with current manufacturing technology, Samsung might be spending money and efforts behind the development of the Random Access PCM and Nonvolatile PCM, to know which one becomes preferable we will have wait along with time.