That difference in clocks vs IPC is most probably showing Sunny Cove's best case scenario, where it performs 40% better than Skylake / Kaby Lake / Coffee Lake / Whiskey Lake.
The problem is that on average, per Intel's own official statements, the average IPC gain is 18%.
The other problem is that per Intel's own official statements, 10nm+, which is the node being used to make mobile Ice Lake, can't clock as high as 14nm++. And we can see that, as they announced a max of 4.1 Ghz Boost, whereas just this year they launched a Whiskey Lake based 15w mobile CPU with a 4.8 Ghz boost, the i7-8665U.
And from what I recall, the fastest SKU that they've announced so far only turbos to 3.9 Ghz, not even 4.1 Ghz.
And finally, the other problem is that not only is it mobile up to 28w, it only goes to 4C/8T. None of this allows them to have competitive desktop SKUs and when it comes to mobile, it will be a draw in some cases, and a slight win in others, but nothing outstanding.
They have been talking a lot about decoupling the process from the architecture, so who knows maybe if 10nm++ or 7nm isn't doable in time, maybe they can put Ice Lake on 14nm++ for desktop parts to buy some time. The TDP will be higher most probably, though.