[Shadowraven]

153bp is about 120whp on an mx5 and seems very low to be getting from an M45. An M45 set up should see you reach 180whp. I'd just keep the M45 with a suitable reduction pulley if I were you. Going NA is really hard, you generally sacrifice low end and mid range for peak power and really need to know what you are doing with matching parts together or you'll likely see very minimal gains. There's a build on here where being 2mm off on the manifold primary diameters cost 20hp. You'll also need a lot of dyno time and a tuner who is very knowledgeable. To put it into perspective the aforementioned build made over 200hp NA but spent over £3000 on dyno tuning alone IIRC.
The M45 will add low down torque and be much more fun and usable on the road rather than a 5k rpm only screamer. It'll also idle much better and be easier to live with. I say just turn the M45 up. M62s are very similar, they make a touch more power but do so more efficiently which is the real benefit. Same for the rotrex, they are a lot of money and are more for those who want turbo levels of power but for what ever reason dont want a turbo. As you already have an M45 there's little reason to upgrade to either IMO, no when you've not got all you can out of it yet.

BTW, the coldside M62 kits use an MP62 which has an internal bypass, are a lot of money, and very hard to find in the UK. Before you consider one you're best off looking at the fitment and intake routing issues as they are very tight and not something I'd be happy to live with.

 

[Shadowraven]

Ah sorry.

180whp is hard to get to NA (I know you said bhp), but it's easily enough with a charger, and 180bhp won't be much higher than you are now. As I said, you really need to know what you are doing, cam pulses need to be matched with exhaust sizing and lengths, everything needs to aim for about the same power band, there's too much in it for me to bother, and even if you do you are still giving up the low end and drivability you'll use most on the street. To get an idea what is really involved have a look at some of the NA threads on here. Just throwing parts on without understanding how they need to match together will likely net little but a lighter wallet.

Cost wise an M62 can be done with some bells and whistles for under £2000, mine sits at £1500 but I also have things like a new exhaust, racingbeat style intake, engine mounts etc that you like don't need. You are better off going M45, a fresh build can be done for around £1000 as you can keep the stock clutch in most cases, and as you have most of the main stuff already by the sound of it it'd be silly not to. I'd not recommend anyone pursue a supercharger or NA build if they had a budget of £5000, supporting mods like suspension and brakes aside that budget would net you a very very fast forged turbo build. The gt2860rs build I'm starting shortly is set to cost £3000 excluding the mapping, and that's with a new mishimoto oil cooler, 600cc injectors, new garrett turbo, new mid pipe, kraken manifold and 3" DP, you could certainly cut some of the cost out of it if you wanted.

 

[Shadowraven]

As above, Ed Kellett/Garage42 kit DIY welded can be had for £150. M62s can be had or £150, DIY EFI ECU is £270, intercooler and piping is about £200, rx8 injectors £60-80, bypass £40, wideband £170, reduced pulley is expensive at £200, and a clutch is £300 but that's things mostly there at £1560 in parts. There's about £500 left for boost gauge (£30), IAT (£20), filter (£40), second TB (£15), idler pulleys (£20 tops), BOV (£90) if you want one, clamps and other little things you might need. It's perfectly do able. With an M45 the reduction pulleys are under half the price and you shouldnt need a clutch so you can save £400 ish there.

 

[Shadowraven]

They do yeh, it's like £120 for the upgrade, but you might also want a gauge and I dont think that the upgrade comes with the sensor. A mate just bought one and it's the impression he's given me. Once you factor those costs in there's little in it and I'd go external for the visibility. The AEM ones also seem to have a better response rate than some integrated controllers.