Not to be a smart ass but can you hear the difference between 44.1 and 96k? Really? (I do not want to get into a you know what match about sample rates) Are you tracking and seeking low latency by shrinking buffer size? If you insist on high sample rates (which is fine by me) and increased bit depth you may want to "render"/freeze some of your tracks to ease the load on your pc. Create a backup of your project then open new one with stems or submixes. you can then tweak away and render those to re-import into original. Old school bounce in digital age. I am a bit confused by our statement about 32 bit vs 24 bits. My PC DAW uses a 32 bit program (this is b/c some of my plug-ins are only 32 not 64). Now when one speaks of 24 vs 32 and sample rate in the same sentence Bit depth = noise floor, dynamic range. 32 bit processing is a different creature all together. Hope this helps, not trying to be difficult. Be well all.
I said I record at 92/32, not because I can hear a difference, but to avoid rounding errors. Digital processing is simply math, and results become more accurate when they are significant to more decimal places. It seems to me (please note the use of that phrase) that, given the amount of digital processing I apply to my tracks, I lessen the potential of digital artifacts, thereby preserving harmonics and intentional detuned beating, by using the higher sample rate. As for bit depth, as you note, it lowers the noise floor and, given that I don't have a proper studio, that is important when I do noise reduction. It also allows for greater dynamic range, which is a factor in my music, and allows me to be more "quick and dirty" when I record. Also, 32-bit processing in my DAW is floating-point, as opposed to fixed point, which, again it seems to me, reduces rounding errors.
As for load on my computer, that's a non-issue. A larger ASIO buffer is needed to handle the additional data but as long as the computer can keep up with the data, it does not mean latency is increased. As I noted in my earlier post, I have used as large as a 4k ASIO buffer and, on my system, it resulted in no perceivable latency.
The only downside in working in 92/32 is more data is produced so more data has to be stored. My home system is LAN based and has a total of more than 15 terabytes of storage spread among the various things that I do with computers (there are, at any given time, 7 to 9 computers on my LAN). My music is backed up to three separate RAID systems and I'm not remotely close to running out space, and won't for the foreseeable future. I'm used to editing video (also as an amateur), and the storage and processing requirements for that make audio recording and mixing a walk in the park for my systems.
So, what do these rounding errors sound like? Can you post a sample?
No. I am an amateur and I don't record for any reason other than my own pleasure and edification, and that of my writing partner, as well as some friends whose opinions about the
composition, not the recording, I solicit. My projects are not intended to be distributed to anyone and certainly not heard by everyone but, rather, to be produced and performed for an audience at some later date once they are finished. I don't record for anyone else, don't want to record for anyone else, and will never produce recordings that are intended for commercial (or, for that matter, public) distribution. What little free time I have is spent composing and recording and I have no interest in convincing anyone else that I am right. My system works for me, and that's all that matters. As I mentioned in another thread, as useful and helpful as HR is, many of its participants appear to believe that there is only one kind of music, to be created in only one way, and for only a single end result. That is simply not true.
If I am wrong, I am wrong, but so what? I don't tell anyone else that they should record and mix at 96/32 and I don't know why anyone would have a vested interest in convincing me that I shouldn't. My points in my original post were that too-small ASIO buffers typically result in crackles and pops and a large ASIO buffer does not necessarily result in perceivable latency, and I am correct about that. Someone specifically asked me what sample rate I record at and I answered.
I like this guy's video very much (and liked it and said so the first time it was posted). My concerns have nothing to do with stair-stepping, but with sonic artifacts and, specifically, those that effect harmonics and beating, introduced as a result of lower precision mathematics being used for digital processing. I don't believe the video addresses that.
Everything I do prior to CD creation is 92k/32 bit, high, I suppose, but at least it's not 192k. 
