The Mi25 is competing directly against Nvidia's Tesla P100 GPU,
That Tesla P100 has a core clock of 1328MHZ, with a boost up to 1480MHZ. Tesla p100 is a 300W TDP part also.
Nvidia Chip does 21.2 Tflops half precision, 10.6 Tflops Single, and 5.3 Tflops Double.
http://www.anandtech.com/show/10433/nvidia-announces-pci-express-tesla-p100
The Vega 10 Mi25 we're looking at has a core clock of 1526MHZ (no idea on boost yet) Has a TDP of 225W.
AMD Chip does 25Tflops half precision, 12.5Tflops Single, and 675Gflops Double (the rate for the AMD chip for 64 bit is 1/16 compared to Nvidia's 1/2 rate, hence Nvidia easily comes out on top there.
http://wccftech.com/amd-vega-10-20-slides-double-precision-performance-1500-mhz-vega-10-x2-2017/
I don't see AMD doing nearly as bad as you think they are.
If I go ahead and include the 300W Vega, things get uglier depending on what you look at, but we have an estimated clock of 1100MHZ for that right now.
The P100 is based on a bigger Pascal X, but even then Vega compares very favorably. (The Vega Chip arguably does even better compared to Pascal Titan X, but the GP102 isn't aimed at the same market as P100, so not a big deal)
All signs right now are that AMD has potentially finally caught up to Nvidia in the core clock speed department.
If you look at the core clock for the Pascal Titan X, it's 1417MHZ. All indications are that AMD has definitely caught up, even if they are using Glofo's 14nm FinFET while Nviida is using TSMC's 16nm FinFET.