Out Go The Lights In Taiwan (NOT)

“TSMC had mulled setting up its 3nm wafer fab in the US due mainly to the availability of stable power supply there. For any 12-inch wafer plant, more advanced process requires higher power consumption, with electricity consumed by 3nm process likely to double that by 5nm process. Accordingly, after a massive power outage occurred on August 15, 2017 around Taiwan, doubts had deepened over whether Taiwan’s power supply could secure normal operation of a 3nm wafer fab in the country.”
 
See Commentary: Triple-win decision for TSMC to build 3nm wafer fab in Taiwan
For a guy currently running some 45nm kit, it’s exciting to see Taiwan head for 3nm. That’s just a few atoms in size. That’s competitive with the world. That allows all kinds of choices for chips. Imagine how many ARMed cores one could build with that kind of resolution. It’s mind-boggling. I can see solar-powered smartphones real soon now. I can see all kinds of computer-equipment getting some power from ambient lighting, RF or even body-heat.

I’m zeroing in on some 10nm equipment. I doubt I can wait for 3nm but it gives me room for future growth of IT in my property. It’s all good to have choices.

About Robert Pogson

I am a retired teacher in Canada. I taught in the subject areas where I have worked for almost forty years: maths, physics, chemistry and computers. I love hunting, fishing, picking berries and mushrooms, too.
This entry was posted in technology and tagged , , , , . Bookmark the permalink.

13 Responses to Out Go The Lights In Taiwan (NOT)

  1. oiaohm says:

    Deaf Spy 5% of a limited number of CPU cores you are seeing from the OS might not be 5% of the total processing power of the machine.

    Arm idle is not 5% https://www.slideshare.net/linaroorg/idling-ar-msinabusyworld

    System Management Mode on x86 is why you hear x86 people talk about 5% CPU usage being idling what is in fact a miss usage of the term. Arm idle is basically dead stop by book its CPU at dead stop with no load..

    https://en.wikipedia.org/wiki/Idle_(CPU)
    Deaf Spy basically the correct usage of the term Idling on CPUs is if the CPU is truly doing nothing.

    Then how would you call a device that works only within 5% of its capacity?
    Sorry using the word device meaning idle has a different meaning. Yes terminology is a evil at times. Please note device idle has to include all chips and that can be 5% is idle but to be correct the general CPU has to be at 0% or its not really idle.

    Deaf Spy basically this is another case of using incorrect terminology and lack of understand how to apply the terminology on your part. It does not help that people coding on x86 miss use the term because they don’t want to talk about the elephant in the room being System Management Mode.

  2. Deaf Spy says:

    5% is not idling

    Then how would you call a device that works only within 5% of its capacity?

    And, don’t forget, why exactly do you personally need the so many ARM cores?

  3. Deaf Spy, contradicting his own argument, wrote, ” It idles. The CPU attributes to maximum 5% of the time spent on these works, the rest is all IO operations.”

    5% is not idling. Someone is converting path/file-name to track/sector/block information. It’s not the drive. Someone is buffering. It’s not the drive alone. Data is being integrated into the information system of the computer. It’s not the drive.

  4. oiaohm says:

    And do you know what the CPU does during these activities, Robert? It idles. The CPU attributes to maximum 5% of the time spent on these works, the rest is all IO operations.
    Again what CPU. Because the CPU on a hard-drive does not idle that much. The CPU in many IO controllers does not idle that much.
    https://github.com/brandonlw/Psychson
    This here is hacking a USB storage device controller cpu. So when you are doing IO form a USB storage device is it running at a max of 5% no it running like at 90 Percent.

    You are reading from particular brands of harddrive the cpu inside is a standard arm core cpu and yes again it running at more than 90% at full IO. I guess you are thinking of the x86 cpu on the motherboard that is basically off loading work to a stack of different CPUs some are in fact arm. What gets interesting on some arm boards in the SOC chips is that you have access to all cores include the ones doing like Sata talk and other operations you cannot access simply on a normal x86 motherboard design. So you might need a arm soc chip with 16+ cores arm to have 8 like cores for applications. I guess you would not have been thinking that something like a sata port could be sitting on a arm cpu gpio pins. Basically think winmoden on a lot more scary scale.

  5. Deaf Spy says:

    One needs to translate file-system or network requests into drive operations, buffering, error-handling etc.

    And do you know what the CPU does during these activities, Robert? It idles. The CPU attributes to maximum 5% of the time spent on these works, the rest is all IO operations.

    So, what do you need these fabled many ARM cores for?

  6. Deaf Spy wrote, “What exactly do you intend to do with these fabled many ARM cores, Robert? Serve files from a disk? (Hint: you don’t need a CPU for that 🙂 ).”

    Of course you need computing power to serve files. One needs to translate file-system or network requests into drive operations, buffering, error-handling etc. That can be done with hardware but that’s just another CPU these days. USB drives can have CPUs for pity’s sake. Hard drives have a controller that does a lot but to serve files you need a higher level of protocols.

  7. oiaohm says:

    Well, well… I am yet to see a smartphone that can go with one charge for three days. That should come first.
    https://www.engadget.com/2017/05/08/oukitel-android-phone-10000mah-battery/
    Deaf Spy has existed for a few years smart-phones than can do 14 days per charge in standby and over a day in full usage. Ok one huge battery. 10000mah is almost triple what is in the galaxy 8+ what even in benchmarking usage run for 8 hours. Yes the 10000mah smartphones can do about 3 days in general usage.

    These, er, “cellphones” will be without screen then? Is that how you imagine this brave new telephone world to work, eh, Robert?
    There is interesting technology
    https://www.extremetech.com/computing/114609-smartphone-screens-with-built-in-solar-cells

    To make it better we need screen and processor to be more power effective. Yes 11 percent generation using the light we cannot see effectively like UV. So screen can be solar panel just not the most effective solar panel.

    What exactly do you intend to do with these fabled many ARM cores, Robert? Serve files from a disk? (Hint: you don’t need a CPU for that 🙂 ).
    Define of CPU. Controller boards on hard drives and ssd drivers have what is a CPU for them that can be arm. So serve files from disk without a cpu pull the other one it plays jiggle bells.

    Lot of people want to ignore how many cpus are embedded inside a PC/smartphone with software they cannot control.
    https://github.com/brandonlw/Psychson
    And at times people work out how to hack different ones of those embedded cpu and reprogram them.

  8. An Out Of Phase Transistor says:

    These, er, “cellphones” will be without screen then? Is that how you imagine this brave new telephone world to work, eh, Robert?

  9. Deaf Spy says:

    What exactly do you intend to do with these fabled many ARM cores, Robert? Serve files from a disk? (Hint: you don’t need a CPU for that 🙂 ).

    I can see solar-powered smartphones real soon now

    Well, well… I am yet to see a smartphone that can go with one charge for three days. That should come first.

  10. oiaohm says:

    https://en.wikipedia.org/wiki/Single-atom_transistor

    ram currently we can produce 1 single atom transistor in a room temperature device. Current systems of lithographically that can be used for 5 and 3 nm don’t work.

    There is lithographic methods for 1nm.
    http://www.sciencedirect.com/science/article/pii/S0030399214000267
    Suspected to be tune-able for .5 nm. These methods don’t work at all for single atom either. So is a new factory for 1nm and .5nm again.

    Also there is alternative to single atom transistor being return to valve based methods. Either way once we get to single atom transistor or it competitors getting smaller appears off the cards.

    3nm, 1nm, 0.5nm then more step thats end. 0.5nm is most likely the end that we can use atoms as any from of insulation.

    Single atom and sub nm valve both depend on empty space being used as insulator. That does create a few structure problems and a lot more sensitivity to contamination and miss placement.

    If it remains 4 years to step down levels some time in 2030 we hit wall this most likely will not be a wall that will be able to be researched around. 1nm and 0.5 nm is still able use silicon as substrate but the circuits themselves are not in silicon because silicon atoms are on the large side at 0.2 nm. So its a little bit of a debate if 1nm and 0.5nm are silicon tech.

  11. ram wrote, “How is this even possible? References please.”

    If you move charged nuclear or atomic particles around in cyclic particle accelerators they emit electromagnetic radiation which can be used to etch silicon masks preferentially to draw the circuitry of a chip. Accelerator technology is now at the point where extreme ultraviolet radiation or even X-rays are possible. The shorter the wavelength of the light used, the finer the details possible. Accelerators capable of beam energy of gigawatts are possible. I don’t know what fraction of that energy is available as synchrotron radiation but it’s enough for practical applications like making 5nm chips. My former boss at the University of Manitoba Cyclotron Laboratory returned to Korea to work on the development of a synchrotron for such purposes back in the 1980s.

  12. ram says:

    How is this even possible? References please.

  13. oiaohm says:

    That time line is interesting. If Intel sticks to their time line when Intel gets to 5nm TSMC will be a year into 3nm production.

    We still have 1nm and 0.5nm technology that has been tested in the labs to come into production. After 0.5nm its fairly much how to create single atom transistors at room temp if possible and then that is the wall to transistor technology.

Leave a Reply

Your email address will not be published. Required fields are marked *