The easy way to answer this is say: upgrade to a GT2871R (
http://www.protegegarage.com/product_info.php?cPath=160_251&products_id=1213)
The why is a little more complicated, but I'll do my best to be concise about it.
This is the compressor map for the Mazdaspeed3's K04 turbo. This is a 3D graph where the inner rings represent the area where the turbo's efficiency is highest. You have to look at this as a 3-dimensional equation.
The Y-axis of the graph represents your boost level and the X-axis represents the engine's air flow (this is determined by an equation accounting for the engine's displacement, volumetric efficiency, boost air temperature/pressure, engine RPM, etc, etc, etc).
An OEM manufacturer is going to choose a turbo that offers high efficiency at WOT but maintains a usable powerband down low for driveability. Inevitably, power down low comes at the expense of the up top power.
Anyway, looking at the map, you can see how at relatively moderate air flow and 2.1 bar (which is right around stock boost) you're at the peak of that center ring. As your RPM increases your engine is moving more air, so you're moving in a positive direction down the X-axis.. and eventually you're completely outside the efficiency of the turbo.
So, thinking laterally, say you turn up the boost to 2.3 bar (~19lbs)... you're still in a decent efficiency, but you're flowing more air.. so the point where you're starting on the graph is already further down the X-axis and you're still bound by the run-out of efficiency.. essentially, you're getting more power- both faster and stronger, but you're looking at degrading horesepower returns, more engine stress, and even more inefficiency at the top of the powerband (this is a debatable comment on this car because you are making more power and the map does compensate pretty well for "overboost" by extending the air flow range, slightly, at higher pressures).
Essentially.. making a powerful turbo car hinges around affecting that airflow equation in as positive a way as possible. Colder air, better flow, etc.. will all aid the power in coming on sooner.. but you need a comiserate turbo that will be able to handle that power at the upper end- which the K04 can not.
So, enter the GT2871R.
This is not the exact map for the ATP bolt on, but it'll get the point across.
You can see how the efficency area is focused on higher boost levels than the K04- you're going to run into that high efficiency ring up to around 23psi. It is a very good choice for a "stock turbo replacement" because it best fits the profile of a lightly modified car.
To help you understand this.. consider this situation:
An MS3 running on a hot day in vaccum at 3000rpm is moving roughly 7 pounds of air a minute. Find 1 bar on the Y-axis and identify where 7 lbs/min sits on the X-axis.
Now you go WOT- and say that by 3750rpm you're at 19psi. At that point the engine is flowing roughly 20lbs/min. So, you should find 2.3 bar on the Y-axis and 20lbs/min on the X-axis and mark that point.
Draw a line between the two and that is your spool. This will occur quickly because you're very decently inside the map the entire way up. The left side of the map is called the surge limit and you're best suited to a turbo that spools and operates in that zone as little as possible.
Taking it a step further, holding your 19psi to the 6700rpm redline your engine is going to be running about 35lbs/min. So, draw a line straight across from your 20lbs/min mark to the 35lbs/min.. and there you have it.
You're looking at a run from mid-70 percentile of efficiency across the peak of map and back by redline you're still in a very high efficiency area of the turbo.
So, the 2871 gives you great down low power and response, and will continue to make power for you as you continue modding/flowing more air. It's not a huge bump in power, but it more than suffices as a ball-bearing, drop in replacement that is going to be a long term, reliable option.
