OT - Aether Reverb ver. 1.5

[John Ludlow]

I just upgraded Aether from 1.0.1 to 1.5 last night. I thought I would report back here to the group about it. It IS noticeably smoother now. It was a remarkable sounding reverb before and it has improved. There are some additional controls now as well. I haven't played with all of them yet. The biggest control addition, though, is that there are several settings now that allow one to trade processor time for additional quality. Also, the list of (very nice) presets is now really huge and they're divided, usefully, into types.

The documentation still isn't finished (if it was ever finished for ver 1.0.1 - I didn't get it). But, what there is of it is very comprehensive. There were a lot of controls before to shape early and late reflections - now there are more. I'm looking forward to the day the docs are finished.

The only serious downside to me is that Aether 1.5 is a serious processor hog whereas Aether 1.0.1 was not. Before, if I assigned it to a track it bumped up my SAW usage percentage by a few points - maybe to 8 percent. Now, for the first time, I can easily kill SAW with just 1 playback track plus Aether running. Even if I leave it at the lowest quality settings - although I can run everything - it uses much more processor than before. When I choose the higher quality settings, not all of the capabilities of the software are even available to me because my i7 920 isn't fast enough to keep up! A week ago I doubted I would ever say that. Also, whereas my previous impression (just from visually looking at the Task Manager cpu usage monitor) was that it used all 4 of my cores pretty much evenly - now it really only uses one - and that one is pegged. The others just tag along.

So, ironically, although it produces a better output, it is seemingly less usable for me than it used to be unless I limit myself to writing to a single track with reverb at a time. I guess that's a presumption on my part - I haven't actually attempted to record - but it seems reasonable considering playback performance. I haven't decided yet whether I'll do something prop-headed with my processor to increase my overall processing capability or re-insall the old version of Aether, which seemed incredible until I heard the new version, and which the (free) upgrade overwrote. One option might be to invest in water cooling and overclock. That might cut down on noise too, actually. Or - maybe I should switch to a fast i5. I think I could get one that uses the same socket, so I could keep my motherboard.

The main point is that it brought what I considered to be an astonishing amount of processing power to it's knees. If you are using an older processor - consider that carefully before purchase or upgrade.

[sebastiandybing]

That is strange, on my 3.16 ghz core2duo it use 18%.
Sebastian

[John Ludlow]

That is strange, on my 3.16 ghz core2duo it use 18%.
Sebastian

Wow. That is worth bringing up with Andrew. What do you have set within the quality selection area? Does 1.5 use both cores for you?

[sebastiandybing]

Its set to default.

Realtime:
oversampling=1x-off
modulation=1-balance
interpolation=1-excellence

ER set to L-R stereo
LR set to L-R stereo

Hall1

SAW shows 18%

In win XP´s the performance meter only shows activity on cpu2

Yesteday I updated my home comp with a new mb and cpu I5 661,
when I get home I will make a install and test on that one also.

Sebastian

[mOjO Fet]

What sample rates are you both using?

Michael

[sebastiandybing]

24/44.1

[John Ludlow]

24/44.1

I was using 24/96. Another difference is the preset I chose. When I use:

Hall1 at 24/44.1
oversampling=1x-off
modulation=1-balance
interpolation=1-excellence

...my usage hovers at 6%. Also, if I do Hall1 at 24/96 I hover around 30% usage (whereas, in 1.0.1 it was closer to 4%). I usually mix and apply reverb at 96. On the other hand, if I try that with 'psytrance delusion' - playback is not possible at all, whereas if I do it at 44.1 - it's much lower. So, the main thing seems to be the sample rate. I guess that makes sense.

I find that at 96k I can't use 4x oversampling at all, but I can use 2x oversampling and set modulation at 2 - 'ultra-smooth' and still keep rolling with most, but not all of the presets. For instance, using 'Tundra' in that configuration playing back a single track puts me over the 90th percentile in SAW usage.

Sebastian, I'd be interested in knowing how you do at 24/96 using the lowest quality levels in Aether and also, at some higher quality settings. Can you use the heavier reverb presets, like Tundra, at all at 24/96? You do have the advantage of a higher clock rate and since only one of our cores is being used anyway my having 4 to your 2 would not seem to be an advantage - maybe you'll do better?

One thing I didn't mention in the original post is that Aether now has 'offline' settings which are contrasted against 'realtime' settings. This would seem to indicate some opportunity to process an audio file using quality settings that aren't possible at all, for heavy processor reasons, in realtime. But, I haven't stumbled over how to accomplish that yet. I can't imagine how it would be possible within SAW at all anyway. Maybe using mix-down like the old SAW Pro Normalization?

[John Ludlow]

One thing I didn't mention in the original post is that Aether now has 'offline' settings which are contrasted against 'realtime' settings. This would seem to indicate some opportunity to process an audio file using quality settings that aren't possible at all, for heavy processor reasons, in realtime. But, I haven't stumbled over how to accomplish that yet. I can't imagine how it would be possible within SAW at all anyway. Maybe using mix-down like the old SAW Pro Normalization?

Yep - that turns out to be it. At the highest quality setting for both SAW and Aether, a roughly 30 second 44.1/24 track, using Hall1, takes over 5 minutes to write to a new file. But, the results are stunning. Core 1 on my machine is pegged for the entire time. Presumably that time would over double at 96k.

Also, I think that at least some of my earlier difficulty had to do with having moved the slider from 'auto' to 'force offline' and then back. I presumed that this would reset the quality settings to what I had been while in 'auto' previously. That seems not to be the case. I believe that from that point on, Aether is using the higher (right side of screen) settings. It seems to disregard one's preference to move 'back' to auto. I presume this is a small bug. I found that I could leave the slider set to 'force offline' and simply modify the settings within it to achieve 'realtime' results.

[sebastiandybing]

it all make sense now,
I did do a little change in the hall1 reverb and that is
to change the late reflection to L-R stereo instead of the default
L - R mono. this will use some extra cpu.

On my I5 661(3,3ghz) hall1(LR=L-R stereo) is using 10% and 18 % on a E8300(3,16ghz),

I will do the 24/96 and tundra test later,

Anyway the best thing about the v1,5 is the 0 latency, I am now using it in
SAC.
And the sound, even in the low settings it sounds better than the TC4000 or
anything I have heard.

Sebastian

[John Ludlow]

It is remarkable, isn't it? Try writing a new file from a single track using the reverb while it's in 'obsession' mode and 4 times oversampling. It takes a long time to render, but that's the best reverb I think I've ever heard.

By the way I checked the length of my file and it's really closer to a minute than 30 seconds - so, that's closer to a 5 to 1 render-to-realtime ratio than the 10 to 1 ratio I reported earlier with everything set to the best quality.