Hi Paul,
1. Am I right that I can only play 24 96 native files (via engine#3) because my off ramp / pace car will only take 24 96
No.
The Off-Ramp (which really determines it when it is in the chain) accepts 24/44.1 just the same, but only 24 (and not 16). This is why the "DAC Needs" is to be set to 24.
Don't get confused by the DAC Needs = 32 setting, because any device that accepts 16 bits (but can do 24) can also accept 32 bits (and throws out the least significant last 8 ).Thus, set DAC Needs to 24, and DAC Is to 24/96 and you can play 16/44.1 just the same. XXHighEnd makes 24/44.1 of it, at that settings.
2. Am I right that with other formats (say 16 44.1) the x2 & x4 upsampling will not work with off ramp pace car because it will not output 24 96 ?
No again.
But, x4 upsampling can't work, because it can do 96KHz max. And I'm not sure what you want to say with "will not work with off ramp pace car because it will not output 24 96 ?" because it just does. It outputs 24/44.1, 24/88.2 or 24/96. If you mean to say dat a device behind it can do 16/44.1 only (or 16/88.2 and 16/96), yes, *then* you are in trouble, because the OffRamp won't output 16 bits. Only 24.
3. Does 0.9y4 use more resource (memory) than 0.9x7 ? and is this likely the reason that with 24 96 files & engine#3 playback stutters for first 30 secs of a file ?
Hard to answer, because that stuttering shouldn't happen anyway, and can only indicate that your PC is not "on par", while at the same time it depends on quite some things. Thus, if you set "Start playback during conversion" while rpe-conversions are in order (like from FLAC), yes that would be a good reason for the sound to stutter (with a not on-par machine). If you, together with that, have a single core processor, you just shouldn't use that setting, because the conversion will take away the cpu cycles from the playback engine. But, this would also happen with multi core processors and a wrong Core Appointment setting (I use 3) or priorities for Player and Thread which are too close to eachother.
Things have changed indeed in 0.9y, which mainly is about all is preprocessed, and *nothing* is processed in real time anymore, while 24/96 sure was processed in real time before. It is hard to predict how this can differ, but it really can be solved with a multi core processor. With single core I guess not, unless the processor is powerful enough (and any notebook processor is just not). Don't look at cpu useage for this, because it is not about that. It is about the speed of task switching, and any power saving processor is just lousy on that (like a Celeron is lousy on it).
Well, I hope this helps a bit !
Peter