8

I noticed my IBM XT 286 has no heatsink on the CPU nor a fan of any kind (only the PSU has a fan). I realize this was fine when it was built, but for longevity purposes, should I put a heatsink on it? Or is that a silly thing to do?

Are there other chips that could benefit from a heatsink?

Improve this question

1 Answer 1

Reset to default
15

The 80286 has a maximum power dissipation of 3.3W and an operating range of up to 70°C. Unless you want to operate near that point, no heat sink is needed. The ICs surface is alreaady way too large.

Beyond that it falls into the same class as iluminated cables and coloured ribbons. Not needed but doesn't do harm either (if done properly).

From a historic standpoint, it would be a shame to add such gimmicks to a (by now) classic machine. Especially as the XT286 is a bit less common than an early AT.

Improve this answer
10
  • Only 3.3W? That seems preposterously low considering it was the mainstream performance desktop processor of choice when it came out. Why did the TDP of CPUs increase drastically into the mid-2000s and then plateau? Could the 286 have been designed with (today's) 90W TDPs for better performance, or does it mean that today's processors could be designed-down to ~4W TDP and yet still yield acceptable performance?
    – Dai
    Dec 27, 2017 at 0:36
  • 2
    @Dai - note that many of today's processors do have TDP that low, or even lower. ARM processors with multiple cores running at multiple gigahertz can easily run without heatsinks, for example, and you'll likely find one in most modern phones. The Intel Atom processor is also similarly efficient. They sacrifice a little in terms of performance for this, but not as much you'd expect.
    – Jules
    Dec 27, 2017 at 0:49
  • 1
    @Dai What you mean 'only' ? Thats more than 13 times what a 6502 was made for (0.25W), desprite using just three times the die size (~47 vs ~16 mm²). And no, if you use back then technology a modern CPU would have the size of a letter sheet and be impossible to be produced. Remember power consumption in CMOS is linear with clock frequency, or more exact number of loads switched. Rule of thumpfor power is P= αC(V^2)f where f is the frequency and V is the voltage. Thus, halving your speed is halving your power need. Is that what you're looking for?
    – Raffzahn
    Dec 27, 2017 at 0:59
  • 2
    @Dai while neither of the things you asked for are true, a 286 could be made on a modern ULV process to have the same performance at only a fraction of a watt. As for why TDPs have plateaued, it's because it gets increasingly complicated to remove more than 150W or so from a CPU-sized area. That, combined with energy-consciousness at the data-center level (you pay for every watt twice, once as electricity and once as cooling), makes higher-power chips rather hard to sell.
    – hobbs
    Dec 27, 2017 at 5:20
  • 1
    Oh, and: Average and peak power consumption might be close together with the original 80286, given it is an NMOS (static operation wastes significant power) and not a CMOS part (where almost all power usage is either for dynamic operation or due to leakage, which is not a big problem with micrometer processes :) ) ...
    – rackandboneman
    Dec 27, 2017 at 17:07

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .