The 90 nm Prescott Pentium 4 was much more power hungry than the previous 130 nm Northwood Pentium 4.
Even worse than the TDP was the fact that the 90 nm Pentium 4 had huge leakage current, so its idle power consumption was about half of the maximum power consumption, e.g. in the range 50 to 60 W for the CPU alone.
Moreover, at that time (2004) the cooler makers were not prepared for such a jump in the idle power consumption and maximum power consumption, so the only coolers available for Pentium 4 were extremely noisy when used with 90 nm Pentium 4 CPUs.
I remember when at the company where I worked, where we had a great number of older Pentium 4 CPUs, which were acceptable, we got a few upgrades with new Prescott Pentium 4. The noise, even when the computers were completely idle, was tremendous. We could not stand it, so we have returned the computers to the vendor.
A current AMD CCD is ~70mm² and can drop around 120 W or so on that area. E.g. the 9700X has one CCD and up to a 142W PPT, 20 W goes to the IOD, ~120 into the CCD.
edit: (1) this account/IP-range is limited to a handful of comments per day so I cannot reply directly, having exhausted my allotment of HN comments for today (2) I do not understand what you take offense at, because I did not "change [my] original argument" - you claimed, a P4 die is much smaller, I gave a counter example, and made the example more specific in response to your comment (by adding the "E.g. ..." bit with an example of a SKU and how the power would approximately split up).
Since Milan the IOD consumes up to 40W during extended PPT loads (The right term for the numbers you are talking about which is more keen to Turbo of the older P4s ie. 130W tdp on Prescott). It's also important that PPT refers to power delivered to the socket, not directly to the CPU, and shouldn't be confused with TDP. Editing comments to change your original argument is cowardly behavior, so I'm ending this discussion.
I mean, if what you want is P4-class performance, the modern semiconductor industry is excellent at delivering that with low TDP. An Apple A18 Pro [1] gives you over 7x the single thread performance of a Pentium 4 Extreme Edition [2] at 8 W TDP, compared to 115 W for the latter.
The 90 nm Prescott Pentium 4 was much more power hungry than the previous 130 nm Northwood Pentium 4.
Even worse than the TDP was the fact that the 90 nm Pentium 4 had huge leakage current, so its idle power consumption was about half of the maximum power consumption, e.g. in the range 50 to 60 W for the CPU alone.
Moreover, at that time (2004) the cooler makers were not prepared for such a jump in the idle power consumption and maximum power consumption, so the only coolers available for Pentium 4 were extremely noisy when used with 90 nm Pentium 4 CPUs.
I remember when at the company where I worked, where we had a great number of older Pentium 4 CPUs, which were acceptable, we got a few upgrades with new Prescott Pentium 4. The noise, even when the computers were completely idle, was tremendous. We could not stand it, so we have returned the computers to the vendor.
The die was much smaller…
Die size: 135mm²
A current AMD CCD is ~70mm² and can drop around 120 W or so on that area. E.g. the 9700X has one CCD and up to a 142W PPT, 20 W goes to the IOD, ~120 into the CCD.
edit: (1) this account/IP-range is limited to a handful of comments per day so I cannot reply directly, having exhausted my allotment of HN comments for today (2) I do not understand what you take offense at, because I did not "change [my] original argument" - you claimed, a P4 die is much smaller, I gave a counter example, and made the example more specific in response to your comment (by adding the "E.g. ..." bit with an example of a SKU and how the power would approximately split up).
The tdp is for the whole cpu with multiple ccds and iod…
Since Milan the IOD consumes up to 40W during extended PPT loads (The right term for the numbers you are talking about which is more keen to Turbo of the older P4s ie. 130W tdp on Prescott). It's also important that PPT refers to power delivered to the socket, not directly to the CPU, and shouldn't be confused with TDP. Editing comments to change your original argument is cowardly behavior, so I'm ending this discussion.
You added wrong numbers and shifted the metric from tdp to ppt. There seems to be a reason for your restrictions. Goodbye.
Which was huge in the era when CPUs didn't underclock themselves at idle to save power and coolers looked like this: https://www.newegg.com/cooler-master-air-cooler-series-a73/p...
Some coolers today still look like that but they're on chips drawing 35W or so while idling at <2W.
I mean, if what you want is P4-class performance, the modern semiconductor industry is excellent at delivering that with low TDP. An Apple A18 Pro [1] gives you over 7x the single thread performance of a Pentium 4 Extreme Edition [2] at 8 W TDP, compared to 115 W for the latter.
[1]: https://www.cpubenchmark.net/cpu.php?cpu=Apple+A18+Pro&id=62...
[2]: https://www.cpubenchmark.net/cpu.php?cpu=Intel+Pentium+4+3.7...