If one is planning just to get the highest CPU frequency (MHz) or raw power (MIPS), for certain reason, then the lower Vcore rated Tbred B 1700+ DLT3C would have an edge to get the highest MHz number.
On the other hand, CPU MHz or raw power is not the only means for achieving highest
system performance which including CPU raw power, cache and memory performance, memory bandwidth. The Barton 2500+ would consume less power than the Tbred B 1700+ DLT3C for delivering the same overall system performance. The main argument is that Barton 2500+ can achieve the same level of system performance as the 1700+ by virtue of the larger L2 cache, even the Barton 2500+ is running 5% lower in CPU MHz and raw power compared to a 1700+. The larger L2 cache (256KB more) improves program performance ranging from 0 - 10+ %, depending on the particular applications. As a result, Barton 2500+ consumes lower CPU power (~9% lower) and die temperature (for the same cooling).
From the following overclocking characteristics of a 2500+ and 1700+, at 1.8-1.9 V, by reading from the graph, a 2500+ running about 7% lower in MHz, would consume about 12.5% lower power and about 3 C lower in die temperature.
Overclocking characteristic (voltage, frequency, power, current, temperature) of Tred B 1700+ DLT3C and a Barton 2500+
(to get to the graph, may have to scoll the page up and down for the right post).
Further, for a given cooling setup for the same system performance, lower power is less likely running into stability problem due to temperature than the case by simply pushing voltage and CPU MHz.
