Taiwan has implemented new export restrictions that prevent Russia and Belarus from receiving decent CPUs. Among the processors that are banned include chips with frequencies that exceed 25 MHz, offer performance greater than 5 gigaFLOPS or higher and an arithmetic logic unit with an access width of 32 bits or more, and those with “more than one data or instruction bus or one serial communication port that provide direct external interconnection between parallel microcircuits at a transfer rate of 2.5 MB/s.” ICs with “more than 144 pins or basic gate propagation delay time of less than 0.4 nanosecond” are also banned from exportation to Russia and Belarus. One of the first Intel processors that could hit 25 MHz was the i80386, first released in 1985.

Image: WikiPancu

Alignment and exposure equipment for wafer production using photo-optical or X-ray methods, such as lithography equipment which includes image projection and transfer, step-and-repeat operation (direct step on wafers) or step-and-scan operation (scanners) processing, as well as scanning electron microscopes designed for automatic inspection of patterns of semiconductor devices are under the ban as well.

Semiconductor equipment has become a main category of Taiwan’s machinery exports in terms of export value, accounting for 32.0% of total export value in 2021, with China, Hong Kong, Singapore and the US being major markets.

Source: DigiTimes

Go to thread

Don’t Miss Out on More FPS Review Content!

Our weekly newsletter includes a recap of our reviews and a run down of the most popular tech news that we published.

14 comments

  1. Does this include systems that were designed in Russia but are manufactured by TSMC?

    Examples from wiki:

    Baikal BE-M1000 ... The CPUs are made by TSMC in a 28nm process
    Elbrus Elbrus-8SV ... Tech. node 28 nm, TSMC process
  2. Does this include systems that were designed in Russia but are manufactured by TSMC?
    I imagine it does - although only from the fabs located in Taiwan. TSMC also has a fab located in mainland China, and it wouldn't exactly be hard to have the chips ordered and papered to Mainland, and then... just keep going North to Russia.
  3. I imagine it does - although only from the fabs located in Taiwan. TSMC also has a fab located in mainland China, and it wouldn't exactly be hard to have the chips ordered and papered to Mainland, and then... just keep going North to Russia.
    I wonder if that were the case if Taiwan would pressure TSMC to not fab for Russia from any of its plants regardless of where it is located.
  4. I have some 386's laying around if Rus is interested. I don't accept Rubles or potatos though
    They could pay you in bit coin I'm sure... just give the friendly hacker your wallet ID...
  5. I have some 386's laying around if Rus is interested. I don't accept Rubles or potatos though
    They could pay you in bit coin I'm sure... just give the friendly hacker your wallet ID...

    Or maybe they could pay you in oil :p
  6. I wonder if that were the case if Taiwan would pressure TSMC to not fab for Russia from any of its plants regardless of where it is located.
    Pressure? Sure.

    Succeed?

    It's not in Taiwan's best interest to antagonize China, and you don't get a plant in China of any sort as a foreign entity without government oversight. I doubt their mainland staff will have a choice, push comes to shove.

    I have some 386's laying around if Rus is interested. I don't accept Rubles or potatos though
    We can (and should!) laugh, but for perspective, the F-22 baseline used 486s. Which were cutting edge when it was designed. Also, they work.

    I shudder to imagine just how much of our modern consumer-facing computing is just... overhead. I'm betting that unless one is gaming or working some other compute-intensive content creation or simulation workload, that the overhead between actual compute work and everything else is like 99.99%.

    We were running RTS games on 486s with multiplayer over dial-up. Once we hit a real performance wall, the next round of user-facing performance increases will come from better utilization of the hardware we have.
  7. We can (and should!) laugh, but for perspective, the F-22 baseline used 486s.
    Yeah it's amazing how much hardware you ~only~ need if you can optimize for it and target just exactly what it is you are wanting to do.

    When I was in the service, we ran some very critical safety equipment. It was contained in 12 cabinets, each about 3'x3'x6' inside of a space about the size of a single car garage. The entire operation would easily have ran on a 486. Everything was entirely analog / 1960 relay based, except for one tank level controller - which used an 8080 processor just to maintain water level in what was essentially a very fancy float valve (albeit a very important one)
  8. Yeah it's amazing how much hardware you ~only~ need if you can optimize for it and target just exactly what it is you are wanting to do.

    When I was in the service, we ran some very critical safety equipment. It was contained in 12 cabinets, each about 3'x3'x6' inside of a space about the size of a single car garage. The entire operation would easily have ran on a 486. Everything was entirely analog / 1960 relay based, except for one tank level controller - which used an 8080 processor just to maintain water level in what was essentially a very fancy float valve (albeit a very important one)
    If it ain't broke don't fix it. The only reason we replace hardware when we do is we can't take the risk of failure and these devices have a finite service life.
  9. The only reason we replace hardware when we do is we can't take the risk of failure and these devices have a finite service life.
    Well I can also see the efficiency angle.

    In my example, you have 12 full size cabinets -- that eats a lot of real estate. That eats a lot of man-hours in service and calibration. That eats a lot of spare parts inventory. That eats a lot of power consumption. If you could consolodate all of that down to ... one piece of equipment that fits in a 2U rack mount... and can run diagnostics and calibrations with just the push of a single button - that's a significant savings.

    Or in the IT world - if you have computations that run X% faster, or with Y% energy savings - that translates directly into cost savings right there, and you can calculate a Return on the investment.

    You have a very good point about "it ain't broke" - and that's exactly why that ancient equipment is still there -- it's been proven to work beyond a shadow of a doubt, and nearly any failure mode you can think of has been witnessed, analyzed, and mitigated or planned around. And when you have an organization with effectively no limit to their budget, or man power, such as the Military -- they see nothing but risk. At least until their budget gets slashed.
  10. Well I can also see the efficiency angle.

    In my example, you have 12 full size cabinets -- that eats a lot of real estate. That eats a lot of man-hours in service and calibration. That eats a lot of spare parts inventory. That eats a lot of power consumption. If you could consolodate all of that down to ... one piece of equipment that fits in a 2U rack mount... and can run diagnostics and calibrations with just the push of a single button - that's a significant savings.

    Or in the IT world - if you have computations that run X% faster, or with Y% energy savings - that translates directly into cost savings right there, and you can calculate a Return on the investment.

    You have a very good point about "it ain't broke" - and that's exactly why that ancient equipment is still there -- it's been proven to work beyond a shadow of a doubt, and nearly any failure mode you can think of has been witnessed, analyzed, and mitigated or planned around. And when you have an organization with effectively no limit to their budget, or man power, such as the Military -- they see nothing but risk. At least until their budget gets slashed.
    Depending on your area in IT you also run into diminishing returns on survivability of hardware. Sure the compute power available today means you can run what was in a 5 node cluster before in a singe 2u Chassis today... that doesn't mean it's the best route to go.

    And for some software having a billion threads to run on just increases your licensing costs to astronomical levels and provides no real tangible benefit to performance day to day.

    But yes IT has generational improvements to IPC, and Cores, and memory and storage.. and it adds up to increased performance per watt. Something we as desktop consumers don't care about. And being able to reduce that power footprint is a big win... BUT.

    It's also a big risk... when you cut your footprint down too much you loose out on redundancy. It's always a balancing act and depending on the criticality of your hardware being able to deliver day to day is more important than the footprint... within reason.

    Just some other things to think about... things you learn (or have learned.) in a long and storied career within the world of IT.
  11. It's also a big risk... when you cut your footprint down too much you loose out on redundancy. It's always a balancing act and depending on the criticality of your hardware being able to deliver day to day is more important than the footprint... within reason.
    Well, you're still not supposed to go full dagron smash on it all :ROFLMAO:

    Gotta map out the logic and ensure that you maintain redundancy while reducing overall footprint.

    And when you have an organization with effectively no limit to their budget, or man power, such as the Military -- they see nothing but risk. At least until their budget gets slashed.
    Budgets tend to ebb and flow with the political affiliation of the party in control of the Executive, but in general, they're not unlimited. Congress has the unlimited budget, the DOD is still held to account (if poorly). And it really is about managing risk all around.

Leave a comment

Please log in to your forum account to comment