The launch date of Intel’s 10th Generation Core family is drawing nearer. As it does, reports are surfacing confirming that at least the upper-tier CPUs in this family will have a power consumption best described as “formidable,” or perhaps “FX-class,” if you were feeling a bit cheeky. Or, as it happens, as roughly equivalent to an RTX 2080.
— HXL (@9550pro) April 7, 2020
According to leaks, the Core i9-10900F has a PL1 of 170W and a PL2 of 224W. PL1 is the amount of power the CPU is allowed to consume over a sustained period of time. PL2 is the amount of power the CPU can consume in short bursts. “Tau” (a term that doesn’t come up in this particular story, but that you should be familiar with) is the maximum length of time the CPU can stay in PL2 mode. A CPU with a PL1 of 150W, a PL2 of 250W, and a tau of 30 seconds would remain at boost frequency / 250W for those 30 seconds before dropping its clock to sit at the PL1 point.
You might notice we haven’t mentioned TDP at all yet. When you see an Intel CPU like the Core i7-9900K listed with a 95W TDP, this only applies to the base clock. Unless you disable all Turbo functions or use a motherboard UEFI option to lock the TDP at 95W, the CPU will boost up to the PL1 power level for its sustained clock and to the PL2 level for burst turbo frequencies. The Core i9-9900K maintains a PL2 of 180W.
Part of what makes this significant is that the Core i9-10900F probably runs at different frequencies than a chip like the Core i9-10900KF. The 9th Generation Core i7-9700KF has a base frequency of 3.6GHz, a boost frequency of 4.9GHz, and a 95W TDP. The Core i7-9700F has a base frequency of 3GHz, a boost frequency of 4.7GHz, and a 65W TDP. We don’t know what the actual performance difference is between the two, but we’d expect the 9700F to be at least modestly slower than the KF variant, with slightly lower power consumption.
I’m not going to make any claims about the 10900KF on the basis of this data, except to say that we’d expect the top-end chip of the Core i9 family to have higher PL1, PL2, and tau values than the lower-end chips. The Core i9-10980HK mobile has a tau of 56 seconds, for example, while the lower-end CPUs in the same family have a tau of 28 seconds. If there’s a power consumption gap between the K and the KF CPUs, we’d expect it to favor the F-class chip.
If these PL1 and PL2 values are accurate, we’re seeing Intel adopt a page from AMD’s book, albeit with a vastly higher-performing CPU. AMD’s FX-9590 was the fastest Piledriver-class CPU the company ever launched, thanks to offering higher base/boost clockspeeds compared with the 8-core FX-8350, but the price of modestly higher CPU clocks was a massive TDP increase from 125W to 225W. Intel defines TDP in terms of base clock, so it can hold TDPs static or allow them to increase slightly, but the value to watch in Intel’s case isn’t TDP — it’s PL1/PL2, and those values seem certain to increase this generation.
What all this comes to, essentially, is that if you’re going to buy a high-end 10th Gen core, you’d best be planning to invest in a serious high-end cooler. Top air coolers can handle 250W and there are closed-loop liquid coolers that can go higher than this. Just be advised that squeezing maximum performance out of Comet Lake is almost certainly going to require top-notch cooling.
- At Least One of the Recent Intel Roadmap Rumors Is Likely Wrong
- Intel Expects to Reach Process Parity With 7nm in 2021, Lead on 5nm
- Intel Mostly Cancels Cooper Lake: Is This Good News for 10nm?