It's a common mistake to think that a 1-clocker is 12 times faster than a standard 12-clocker. Only a certain percentage of instructions is performed 12 times faster. Many other instructions are less faster.

Let's have an example. A division shall be performed as follows:

Clocks needed:   12-clocker      1-clocker (C8051F000, Silabs)
-------------------------------------------------------------
MOV A,#data          12              2
MOV B,#data          24              3
DIV A,B              48              8
-------------------------------------------------------------
Total:               84             13

Improvement factor:       84/13 = 6.46

You see, a modern 1-clocker, like the C8051F000 from SILABS e.g., is "only" 6.46 times faster than a standard 12-clocker, when performing an integer division.



The Dhrystone benchmark test runs a compiled C program that contains about 50% assignements, 30% control statements and 20% procedures and function calls, if I remember correctly. The idea is to test the microcontroller by running a represantative application. The result of this benchmark test is that their micro runs 8.1 faster than a standard 8051 micro at same frequency.
Taking into consideration what I wrote above they obviously use a 1-clocker design and compare it with a standard 12-clocker.

Additionally to that 8.1:1 speed improvement, the allow to run their micro at a speed of 210MHz, when using STRATIX-III implementation. So, compared to a standard 12-clocker running at a speed of 33MHz, e.g., they offer a speed improvement of 210MHz/33MHz=6.36, times the 8.1:1 improvement due to their 1-clock design, which yields an overall improvement of 51.5.

Kai