My point is to drive max 5A, and avg 2A load with a P-channel FET.
Im thinking about this fet: NDP6020P (Datasheet: https://solarbotics.com/download.php?file=2205
That should be atleast enough to drive the load.
But i cant drive that with 5V can i?
So i have this in my mind: FDN359AN (Datasheet: FDN359AN.pdf
And i would use a 10k resistor on the source of NDP6020P.
Could it work like this?
And YES, i want to control positive side!
--
This is the idea basically. VCC is 12-14V (Lead acid battery)
The source needs to go directly to +12V. Connect that 10K to the source of Q2
Oh heh what a mistake. Luckily from the drawing only.
So this was my plan:
This should be totally working circuit? When the pin of arduino is HIGH, it turns the LOAD off?
And if arduino is totally without power, its pin is LOW, so LOAD is on.
If arduino pin is HIGH, the LOAD is off?
Note the first Q2 is a N-fet! Not a transistor
Beware the Vdss of the FDP6020P is only 20volts. the best is always to choose twice the maximum voltage of the power supply.
for an automotive application the max battery voltage will be 15V so the Vdss should be 30V minimum.
What is the load? resistive or inductive?
Perhpas a gate resistor should be added to limit the Dv/Dt and perhaps a free wheeling diode.
Q2 is a Nmos not a bipolar NPN so the schematic must be updated 
If the arduino is not powered, the Q2 is open, Vgs = 0 thanks to R2 so Q1 is open and the load is not powered
Bottom of R2 goes to collector of Q2
Edit:
You changed the drawing on us now
Genesis92:
Beware the Vdss of the FDP6020P is only 20volts. the best is always to choose twice the maximum voltage of the power supply.
for an automotive application the max battery voltage will be 15V so the Vdss should be 30V minimum.
What is the load? resistive or inductive?
Perhpas a gate resistor should be added to limit the Dv/Dt and perhaps a free wheeling diode.Q2 is a Nmos not a bipolar NPN so the schematic must be updated
If the arduino is not powered, the Q2 is open, Vgs = 0 thanks to R2 so Q1 is open and the load is not powered
Yeah i was kind of worried about the 20V, but i was hoping this could help me "Rugged internal source-drain diode can eliminate the need
for an external Zener diode transient suppressor."
Its not automotive, but a sailboat. It does have a small fridge, it has label that says its 3A max and 0.4A average. So maybe the diode added to that circuit would be nice.
Other thing with an motor is this rarely used autopilot, which is only attached to the steering handle.
So a diode will be added.
Other loads are resistive. Few lamps, and umm, plotter and a sonar. So nothing big there.
--
If that FET is not ok, its doable.
Other choices are:
IRF9Z34N http://datasheet.octopart.com/IRF9Z34N-International-Rectifier-datasheet-7830620.pdf
OR
IRFR5305 smd http://www.irf.com/product-info/datasheets/data/irfr5305.pdf
I might like to use smd to save space, but TO220 casing standing up takes less space.
But those are the other choices from local dealer. Any ideas?
Genesis92:
the IRFR5305 should be a good choice.
for such power current, it's best to choose the TO220 o TO263 package. not a sot-23.
it's because the turn on current could be very high during the transistor switching and a sot-23 have a very tiny thermal inertia compared to a TO263 or TO220 case.
the fridge is perhaps a "phase change" system with only a heater (resistive load).
but you can add fast switching diodes to protect the transistors againts inductive loads 
Adding a diode should look like this?
Genesis92:
the IRFR5305 should be a good choice.
for such power current, it's best to choose the TO220 o TO263 package. not a sot-23.
it's because the turn on current could be very high during the transistor switching and a sot-23 have a very tiny thermal inertia.
Well my local shop only sells the SMD version of IRFR5305 :I
a D-pak should be ok but add little copper if you can. it can be hot for short period during turning on.
the problem vith your application is this turning on current. if there's a lot of capacitive load at the input of the powered devices this current peak can go up to 100 or more amps for a short time. enough to kill the mosfet.
this is why I was talking about Dv/Dt. A mosfet is like a variable resistor with very low value when turned on.
My company build power boxes with sealed lead acid batteries (yuasa 12v 38Ah) + charger and I'm using a high power relay instead of high power mosfet for this problem. (and for safety)
so you can try and if the transistor blows up perhaps Dv/Dt should be added
Genesis92:
a D-pak should be ok but add little copper if you can. it can be hot for short period during turning on.
the problem vith your application is this turning on current. if there's a lot of capacitive load at the input of the powered devices this current peak can go up to 100 or more amps for a short time. enough to kill the mosfet.
this is why I was talking about Dv/Dt.
I was thinking a latching relay at first, but they are too big for this application.
My luck is, that there is not really anything capacitive on the load side. So it might work?
I was actually doing some calculation (if even correctly)
So please correct me if im wrong.
If its MAX 5A current, and the RDSon is 64mohm. Then its (550.065) = 1.65W. Does it mean that the casing would have to handle 1.65W at 5A load?
And Junction-To-Ambient(PCB-mount) says its 50W max, but (550.065)*50 = 81.25C. Add ambient to that its over 100C.
So should i just ditch the SMD? I dont know how much a copper pad would work as a heatsink.
--
Edit:
But what about IRF4905 then? It has RDSon 20mohm.
If i calculate it correclty; (550.020)*62 = 31C. So it would be maximum of 31C + ambient?
yes this is why the TO220 was a better choice because it's easier to attach a heatsink
with high drain current the lower Rdson is the better. but the Nmos are better compared to the Pmos and the Rdson increase with the Vdss.
You can find SMD P-channel MOSFETs with much lower Rds that will run cooler:
Select one with Low Rds, Low Input Gate Capacitance, and package you can handle.
yes but perhaps not available at the store near him
The NDP6020P has an absolute max Vgs of +/-8V so is completely out of the game, note.
MarkT, you're right...strange thing I've not check this thing o_O.
I think I have opened the FDN359AN and make a mistake with the NDP
so the NDP6020 can be used by adding a resistor in serie with R2 to get a Vgs max of 6V or connect a 5.6v zener diode in parallel with R2 Anode side to the 12v with a resistor in serie with the gate
MarkT:
The NDP6020P has an absolute max Vgs of +/-8V so is completely out of the game, note.
So if im reading this right, it can take voltages up to 20V, but gate cannot be less than 8V from source?
Genesis92:
MarkT, you're right...strange thing I've not check this thing o_O.
I think I have opened the FDN359AN and make a mistake with the NDP
so the NDP6020 can be used by adding a resistor in serie with R2 to get a Vgs max of 6V or connect a 5.6v zener diode in parallel with R2 Anode side to the 12v with a resistor in serie with the gate
I thought of this after reading MarkT:s note. But i was afraid to ask if im wrong 
But basically i could control it with a voltage divider on the gate side?
I might be going for the IRF4905. Its more expencive than others but i need this thing to work for looooong time.
yes you can use a voltage divider with the lower side connected to the control transistor
If reset is pressed while load has power, it may continue running due to charge at gate. Add another resistor, Q2 gate to gnd.