I didn't hear a guitar in either song...
The first song is running a pluck of some kind through a cabinet impulse response. You can find a bunch of free cabinet IR samples around, and there are several free Convolver plugins out there.
The second song sounds like a RhodesEP-ish sound running through distortion.

Nonetheless, a guitar is still useful to have in the production process. Buying an electric guitar and an audio interface might be your best solution. You can get a relatively cheap version of both (guitar, interface), and it's an intuitive instrument to play. That was my solution to the problem (I had never played guitar before). Ended up using it in my next song for a slow melody that took at least 100 recording attempts. There are a ton of Youtube tutorials out there on learning to play the guitar that I need to do here at some point (learning chords and techniques).

For amp simulations, there is the fantastic LePou Amp Sim Pack, and if you can find them somewhere, the entire series of free plugins from Ignite Amps (I can provide them since the site seems to be down).
And for cabinets, this is a good start.

This KVR thread has some possible solutions that you may already know about. Specifically Voxengo Deconvolver, which isn't super expensive.

This is an interesting video from Armin van Buuren about how he syncs his visuals when DJing. Can't remember where I saw it originally...

It's deceptively simple.

Start with a sin wave.
Use amp envelope with almost no attack and a short decay.
Use pitch envelope with a tiny decay and one octave range.
Add a keytracking lowpass filter that is 3 octaves above the note.
Portamanto/Glide that is less than 20ms.
Delay.
Reverb.

Also make sure to record a sine sweep or impulse using stereo mics, and then generate an IR for it. That way, you can "re-use" the space as frequently as you want with a Convolver plugin.

No problem, glad I could help!

I think I was expecting the logarithmic scale of dB to relate directly to sample values, but yeah, that's probably not the case beyond values of -1.0 to 1.0.

I have no clue how much the SSE/AVX instructions affect the math, but I was under the impression that they were there purely for acceleration purposes, i.e. fewer AVX instructions required for vector tasks. Maybe they do use double precision internally or something...

I was thinking more along the lines of values losing decimal precision as they grow in size. The extreme example being that single precision floating point can only represent even integer values between 2²⁴ and 2²⁵.