I dont really believe you could make a faster computer with less transistors. Every new generation of CPU's has more transistors than the previous, and thus increased performance is directly related to transistor count.
You are totally wrong here on a number of counts. You have been well and truley fed the Intel diatribe that bastardises (rides roughshod over )Moores Law.
Which simply says that the number of transistors on an integrated circuit doubles every 18/24 months. Nothing about its relationship to performance.
The problem with your statement is that it is too simplistic and fails to recognise that more doesn't necessarily mean faster (better performance). The more transistors on the an IC means more heat and more heat means more reistance and more resistance equals.....poorer performance. Other things have happened to remedy this that have nothing to do with the number of transistors.
The architecture used within a Desktop PC is fixed by legacy, and thus we all use variations on X86/64.
Absolute rubbish - you don't know what you are talking about. The architecture of a desktop is constantly evolving the fact you refer to the x86/64 is a clear example of that. If it was fixed we would be still be using 4004 processors with 4004 transistors. However, the central processor is only part of the overall performance of a desktop. In the same period betweem the 4004 and the current crop significant developments in the GPU, the ALU, RAM, transistor design and other integral parts of the desktops architecture have had far more impact on the performance than the transistor count alone. This is before considerations are given to improvement in algorithms and software design.
More efficient architectures exist, and are used in things like your smartphone or PS3, where those legacy constraints dont exist.
Even more rubbish - once something is built is becomes out of date - how many ps2 games work on the ps3?
Having said that, while some architectures of CPU have less transistors for a given performance level than other architectures, doesnt alter the fact that you still need to increase the transistor count within that architecture as performance increases.
That statement is an oxymoron.
Using a computer or its cpu is a bad anology to the topic of debate.
The only relevance I can make to its use in this debate, and it has the opposite of legacy issues, is that is that better performance in electronics has come about by focusing on the needs of the customer and delivering what the customer wants. That is why ARM has been so successful using low power but powerful processing that now appears in nearly every form of electrical equipment across the globe,
Like desktop computing, lighting has evolved over time as new ideas replace old whether by design (CFL, LED etc) or legislation (banning certain incandescent lights).
The critical point here is that neither BAS or Radweld (used to stick that stuff in my Austin 1500 radiator) is actually wrong.
The cut and paste stuff that BAS produces for downlights is correct if that is what the customer wants and Radwell is right to highlight his views to the OP that needs to be aware that these lights might be considered over the top for his needs.
The issue of whether the lights are for commercial or domestic situations is one for the customer ultimately to decide.