3 ms·

How many years did it take us to catch up to that Pentium 4's 3.6 GHz clock speed? Edit: For 3.6+GHz Intel chips - there was one single 4th Gen i7 in 2013 & tw

How many years did it take us to catch up to that Pentium 4's 3.6 GHz clock speed? Edit: For 3.6+GHz Intel chips - there was one single 4th Gen i7 in 2013 & two in 2014, one 6th gen (each) 2015 & 2016. 3.6GHz didn't return to more standard production until the 7th-9th gen, 2017-2019.

nsonha 5y ago
your comment makes me realize I have not thought about GHz for a long time. I think we just don't have the demand for high clock speed anymore so they didn't push for it? I'm not a hardware guy but I remember reading that added cores are better suited for mobile computing (thermal and energy wise) than clock speed.
- joe_guy 5y ago
  I believe the goal od netburst architrture was to brute force beat its inefficencies with clock. > With this microarchitecture, Intel looked to attain clock speeds of 10 GHz,[7] https://en.m.wikipedia.org/wiki/NetBurst_(microarchitecture) https://en.m.wikipedia.org/wiki/NetBurst_(microarchitecture)
- zamadatix 5y ago
  CPU wattage is (approximately) scalingConstant * voltage^2 * Frequency + staticPower. The increase in frequency not only increases the power draw linearly but requires a higher voltage which increases the power draw quadratically. Doubling the frequency may quadruple the power requirements so if you can find ways to increase performance per clock instead it'll be preferred as long as those take less than 4x the power. If that means multiple low powered cores great, if that means many specialized instructions that do complex things efficiently (aes, simd, whatever) great. If that means just making existing instructions faster that's fine too. Whatever doesn't break the power budget as there is only so much power you can dissipate without exotic cooling solutions.
zamadatix 5y ago
AMD was doing 3.9 with 4.5 turbo in 2012 and just like the P4 it didn't really do much per clock.

nsonha 5y ago

your comment makes me realize I have not thought about GHz for a long time. I think we just don't have the demand for high clock speed anymore so they didn't push for it? I'm not a hardware guy but I remember reading that added cores are better suited for mobile computing (thermal and energy wise) than clock speed.

joe_guy 5y ago
I believe the goal od netburst architrture was to brute force beat its inefficencies with clock. > With this microarchitecture, Intel looked to attain clock speeds of 10 GHz,[7] https://en.m.wikipedia.org/wiki/NetBurst_(microarchitecture) https://en.m.wikipedia.org/wiki/NetBurst_(microarchitecture)
zamadatix 5y ago
CPU wattage is (approximately) scalingConstant * voltage^2 * Frequency + staticPower. The increase in frequency not only increases the power draw linearly but requires a higher voltage which increases the power draw quadratically. Doubling the frequency may quadruple the power requirements so if you can find ways to increase performance per clock instead it'll be preferred as long as those take less than 4x the power. If that means multiple low powered cores great, if that means many specialized instructions that do complex things efficiently (aes, simd, whatever) great. If that means just making existing instructions faster that's fine too. Whatever doesn't break the power budget as there is only so much power you can dissipate without exotic cooling solutions.

zamadatix 5y ago

AMD was doing 3.9 with 4.5 turbo in 2012 and just like the P4 it didn't really do much per clock.