You should have res to pump order after that it doesnt matter.In 99% of watercooling systems there is no "optimal" component ordering, EXCEPT that which minimizes tight bends and overall tubing length. This is not always true as there can be a significant difference (read: measurable) when dealing with extremely high heat loads, like those found in systems in which one or more components are actively cooled by a TEC. Because of water's high heat capacity, there is very little change in water temperature due to the heat dump from components that may be considered "parasitic" to the loop's most efficient operating conditions. What do I mean by this? In your case, this only really applies to the pump.
Specific heat of water (c) = 4.186 J/gram-°C
By definition 1 Joule (unit of energy) = 1 Watt-second (unit of power)
therefore, c = 4.186 W-sec/gram-°C
Meaning, if you pump "dumped" a modest 15W of heat into your loop:
Q = cm(ΔT) where Q is the heat added (15W), c is the specific energy of the material "receiving" the heat energy (water), m is the mass of the material, and delta T is the rise in temperature of the material.
Since this would ultimately depend on flow rate (to be given as g/sec, L/hr, gal/min or whatever other units you desire) we are going to simplify our example by assuming your flow rate is 2L/min:
(2L/min)(1 min/60 sec)(1000 g/1L) = ~42 g/sec (1L water = 1000 g by definition)
Q = cm(ΔT) => 15W = (42 g/sec)(4.186 W-sec/g-°C)(ΔT) -> ΔT = 0.08 °C
As you can see, certainly not a lot of temperature rise due to the pump alone. So assuming the water temperature entering the pump was 25°C exactly, this would mean that the outlet temperature from the pump would be 25.08°C. If the CPU (or any other block) was located such that the outlet of the pump directly feed it then the supply temperatures would be 25.08°C! (assuming that your tubing was a perfect insulator)
Try repeating this calculation with higher flow rates (3L/min, 4L/min, etc.) and you will see that in any case there is not much to be gained by sending the water from the pump through the radiator before any cooling block/device. In fact, due to the extremely small water temperature increase from to the pump I'm no so sure you would see ANY difference in radiator outlet temperature (a passive heatsink's ability to reject heat is directly proportional to the difference in temperature between the medium being cooled and the medium providing the cooling, in this case air at ambient).
Finally, since the GPU and CPU block locations are the only other real question, I can only provide you with the same advice. Using the equation above you can calculate the approximate rise in water temperature across the block (I say approximate because Q in the given equation can only be approximated from true processor power consumption since calculations, as provided, would have to assume that 100% of the heat from the processor is transmitted to the block and then to the water, which is untrue...). Then it's just a matter of determining which component you think needs that extra degree or two of cooling to perform best. At this point it's really a matter of preference and/or individual testing (which would require that you test the loop with each block individually installed, to remove the other as a competing variable, and then plotting the overclock results as a function of supply water temperature).
I hope I've at least partially answer your question. If not, here's the bottom line: I believe you will find that there is little to no performance difference to be achieve by ordering components in your loop. Instead, work to minimize flow resistance (smooth tubing bends and as little tubing used as possible). Any optimizations here will allow for maximum flow
p/s: I'm not a pro and just share my 2 cents. I hope I helped you a little bit to get stuff clear. Otherwise, these links might be helpful for you:
Goodluck for your watercooling setup!