Generally the Centrino will get you longer battery life. If all you are worried about is battery life, the Centrino will be the way to go. The processor speed will open up all kinds of arguments if mentioned "generally" and possibly even specifically.
hope to use any OS release in the next 3 years you will want 64bit. Which will be faster at 2.0ghz each will be debatable, both will seem faster than 2.0ghz by design (compared to P4 which has been discussed, was a poor design). They should be comparitable, probably some win on one, some on the other. Any application that can take advantage of SSE3 the AMD will use and beat the Centrino hands down -- and more applications are starting to use the SSE3 extension set. Centrino will be noticably more efficient in wireless communication. Since you mentioned a 3 year life span, and the possibility of doing bio data analysis, I would say Turion for the 64bit future support and the SSE3 extension set to speed up some applications
The difference in CPU speed was 5% between the Turion and PM, yet memory processing will be 30%. This was why you see some tests show the Turion as faster, some the PM. Anything strictly core processing will show the PM as slightly ahead, anything that processes large datasets through memory to the CPU will show the Turion as ahead. But if you look at that comparison, they run pretty close, I would look at other features, often the PM is feature rich unless you get a good sale on AMD.
most games will not use the SSE3, they've barely moved to SSE2, one of the reasons is you don't compile your binaries for the highest hardware. SSE2 is mainstream now, cheap and easy and old news, so binary wise all games will be SSE2. The specifics of the game will test graphics card, memory and CPU in combination which is why one game will run better on on piece of hardware and another will run better on another piece of hardware. Assuming the same graphics, I would put the two at a tie, no edge either way as one will get an edge there, the other the edge elsewhere.
Turion actually beat Centrino in battery life when running simple applications like word, excel, etc..., but not when running other more cpu intensive applications. Why is it that people are saying that the Centrino is faster than the Turion? Both systems were virtually the same thing, and the Turion was able to allow a faster graphics board.
A lot of it comes down to what you run on the system. There is a lot of confusion because of how the Centrino runs. Intel chips as you know are CISC design, they keep getting larger instruction sets to handle more capability. There are also extension sets that are industry standard SSE/SSE2/SSE3, this adds more complexity to the mix. AMD tosses in their extension of 3Dnow just to make things more complicated. You cannot really compare CPUs on the Ghz level anymore with any chip. It all depends on the microcode and how fast it can process the given function, and how often that function is called from the application. How often can the processing cycle loop of function within an application fit within the CPU cache, etc. Then you get into processor prefetch of data, and more.
So it all comes down to user evaluation for their needs. In a perfect world where data is designed such that it is bad for the Turion, but better for the Intelligent prefetch on the Centrino, Centrino gets a boost in effective speed. When the instructions are such that they can take advantage of the SSE3 extension, Turion gets a boost in speed. If one has a larger cache, then it may get a boost in speed. When you jump to a new location in execution (call a function) the cache has to be emptied and reloaded, now the larger cache can slow you down, too much random activity defeats the purpose of the processor cache and prefetches (and then you are on raw horsepower).
QUOTE from 1src Forums.
AMD cpus don't gain much, if any at all, from SSE code because they don't provide any speedups using packed SSE data. So all that stuff about SSE3 code being faster on turion can be thrown out the window. AMD mostly includes support for compatibility reasons.
And all AMD and intel x86 cpus today are CISC on the outside and RISC on the inside. In fact, most current CPUs are like that.