Sub-mix levels question: which option is best?

I have a piece made up of various soft synths.  Each soft synth output is routed to a sub-mix bus, which then goes to the Master.  The reason for this is so that I can move the volume of the entire piece up or down with one fader without changing the Master fader, which stays always at 0.
 
I found that the overall level of the piece was quite low, peaking at -12db.    I'd like to have a stereo pre-master peaking at -3db.
To achive this I have various options, and I do not know which is the best choice, to give the best quality - if anyone can advise/suggest which of the following is the best choice (or something else) :
 
1.  Take the level of the sub-mix up over 0 ----( it can go actually only as far as +6db.   Question:  does going over 0 on this sub-mix cause any lowering of quality of the signal?)
 
2.  Go to each individual soft synth track and increase the levels so the overall level of the entire piece to master bus is higher.   This can be quite time-consuming with many tracks.
 
3.  Use a limiter on the sub-mix to increase the levels before it reaches the master bus (or a limiter on the master bus itself).
 
4.  Make the stereo pre-master at -12db, this is an OK level and can be increased with a limiter when mastering.

In my opinion, -3dB is far too loud for a Pre mastered song.
 
Your -12dB is spot on the money and gives the Mastering guy plenty of room to work his magic.
 
In my opinion.

I have always understood you do not want to lower the gain early in the gain stage, only to have to raise it later.  In other words, fix things early in the gain staging.
 
Bob Bone
 

yeah, i found that grouping tracks works well, cuz when you move the fader, they all move together.
Is there a way to do a sub mix, like on Logic Pro X?

Anonymungus!
Hello - here's a related Question:  Does clip gain add any distortion/modulation to audio, especially if you go up & down a few times?

Up and down a few times doesn't matter...a 64-bit engine has a lot of resolution. 
 
Turning it up won't cause distortion in the track itself, but it could lead to distortion further downstream, or change characteristics of subsequent effects like compression or distortion.

If by sub-mix, you mean using additional buses, that is sort of what the OP was doing, except he kind of took it too far for being practical.
 
A common thing to do is to route common instruments to a single sub-bus for each instrument group, which is then routed to the master, though additional layering is potentially beneficial, depending on the nature of the bus configuration you end up with.
 
This is done for a couple of reasons: 1st, grouping like instruments (like 1st violin, 2nd violin, cello, viola), by routing them to a bus (Strings), allows you the flexibility to control individual instruments as needed while also allowing the whole section to be brought up or down.  2nd, 'like' instruments usually have similar kinds of reverb and such applied to them, so you could share a single instance of whatever reverb you wanted to use for the string section.  (This saves CPU, and makes changing the effect levels and/or other parameters easier, since doing it once benefits the whole bus).
 
Folks often route each drum kit piece to its own audio track, then group those tracks by sending things like all the toms to a Toms bus, and all the cymbals to a 'Cymbals' bus, and these buses are then routed to a main 'Drums' bus, which itself is routed to the Master bus.
 
It's all about grouping 'like' instruments for making it easier to add effects to, to save CPU by reducing the number of effects instances are used, and about controlling volume of grouped tracks with single faders.
 
If you watch the first part of one of the Sonar videos, on Larger Than Life Drums, they explain this and go through the setup - and the 'why'.  Here is a link to that video:
 
https://www.youtube.com/watch?v=VH9W1MD80kk
 
You can skip ahead to 7:00 into the video, and they will shortly begin adding the extra buses, and explaining how and why they do so.
 
Bob Bone
 

Latency isn't really a problem during mixdown, it's only a problem when you want to play a live instrument.
That said, a limiter is indeed not such a good idea because that's exactly the kind of stuff the mastering engineer would want to apply with his expertise.

We have been down this road before. You're wrong, and I'm not getting into it again, you got lots of explanations last time.

In my opinion, Option 4 is the best.
 
If you need to mix for competitive loudness, remove artifacts and carefully eq your sounds early on, put compressors with very low compression ratios on each track and on sub-busses as required to gently reduce overall dynamic range. The rest can be done in mastering.

ULTRABRA
@ston: re the quick grouing, thanks for the tip.  That actually would work if I only had soft synth in a piece.   I do mainly, but sometimes there are some audio tracks in there too -- if I understood quick grouping right, it only adjusts faders of same type (meaning pressing ctrl+fader of a soft synth track would move all soft synth tracks simaltaneously, but not any audio tracks).

I *think* you can quick-group any set of faders you like, but I'd need to double check as I've been out of the DAW loop for some time now (about 7 1/2 months!)