An update -
Okay, finally got this tested in both SAW 32bit and SAW 64bit.
I used all 64bit plugins, with jBridge to port the 64bit plugs into 32bit SAW.

The session consisted of five songs, recorded at 48k/24bit and separated into their own group of 11 tracks per song, across the timeline - around 20 minutes long. There's five aux channels with six various 64bit vst fx and a total of nine different 64bit vst fx inserted in various track/channels throughout the five songs, as well as a number of SAW native plugins. Plus, each channel has some amount of channel strip eq and dynamics. And the main output has a small number of vst and native fx, also.

The 64bit session has automation across some tracks and most inserted fx are bypassed until they are needed. The 32bit session has all fx active all the time and there is no automation. Neither session has significant editing to the original regions.

At 4 x 512 buffers, the 32bit session played normally, with a single group of 11 tracks active at around 27-30% MT load. Each additional group of 11 tracks added just 10% load, with a total load of 70% for all 55 tracks.

At 4 x 512 buffers, the 64bit session played normally, with a single group of 11 tracks active at around 21-22% MT load. Each additional group of 11 tracks added just 2-3% load, with a total load of 32-33% for all 55 tracks.

At 4 x 2048 buffers, the 32bit session played normally, with a single group of 11 tracks active at around 25% MT load. Each additional group of 11 tracks added just 8-9% load, with a total load of 61% for all 55 tracks.

At 4 x 2048 buffers, the 64bit session played normally, with a single group of 11 tracks active at around 20% MT load. Each additional group of 11 tracks added just 1% load, for the first four groups. However, adding in the fifth group jumped the MT load to 99%. I was initially able to get the load to jump back down by disengaging up to eight tracks.

Further checking revealed that I could actually load up all but one or two tracks and keep the MT at 24%, depending on where on the timeline I started playback. But, as soon as I enable one more track, or even engage a couple of channel strip processes on an active channel, the MT load jumps to 99% (or rapidly climbs). The choice of tracks and timeline position is totally arbitrary, though and doesn't seem to be dependent on any particular fx, fx chain, or timeline event. It just appears that when some amount of "load" is reached (around the 24-30% mark), the MT jumps to 99% on playback.

This doesn't happen in SAW 32bit.

Sorry for the really long-winded post, just to reach the same conclusion I started with.

UpTilDawn wrote:I'm speaking about editing/mixing in SAW, without SAC as the front end again, as I typically have done, when not tracking.

This is the configuration i always use. Many mixes have regions here and there with long gaps in between.

Whenever I'm doing any sidechaining, I patch a vst on the keyed track to let the compressor or gate to return to their 'normal' levels to avoid surprises. That has the opposite intent - to avoid auto fx bypass. I guess I have to be careful what plugin I use, as that could cause the problem being discussed. (Also if the key has any gaps between regions, I just mix a 'helper track' that is contiguous audio).

Anyway, I'm curious what you find that is causing the loading.

jmh wrote:You could try disabling groups of tracks (or maybe number select one half, then the other) to attempt to see if you can identify one particular (or some) plugin that is related to the high loading.

I had been doing this sort of thing in this particular five-song session.
That helps with the freezes.

My current curiosity has shifted to the difference between running the RME/SAW buffer settings at 2048, or 1024 (and 512 it seems) and the dramatic MT readings.

I'm speaking about editing/mixing in SAW, without SAC as the front end again, as I typically have done, when not tracking.


I think I will create a duplicate session in SAW 32bit tomorrow, using the same 64 bit plugins, through jBridge (just to see how much impact the plugins are actually having). I will run both 64 and 32 bit sessions without SAC (64 and 32, respectively) at 4 x 2048 buffers and 4 x 1024 buffers and see what there is to see.

Bob L wrote:If using SAC as a front end, then you have no control over SAWStudio buffers... SAC takes over and balances them out... SAC should be used for live low latency with buffers set at 64 or 128 max... and as for FX that cause the floating point error, you can always just keep some open room mic active on the track or bus so the levels never go all the way to zero.

Bob L

Understood, Bob. I will keep that in mind.

You could try disabling groups of tracks (or maybe number select one half, then the other) to attempt to see if you can identify one particular (or some) plugin that is related to the high loading.

If using SAC as a front end, then you have no control over SAWStudio buffers... SAC takes over and balances them out... SAC should be used for live low latency with buffers set at 64 or 128 max... and as for FX that cause the floating point error, you can always just keep some open room mic active on the track or bus so the levels never go all the way to zero.

Bob L

Okay - I don't get this...

I finally got a chance to test out the idea Bob suggested of setting the Enable Track FX Auto Bypass menu item.
But before I tried it on any channels at all, I happened to have opened the original multi-track project I'd raised the issue over and the MT load never went above 30% !!

Curious, that none of the three current sessions on the hard drive raised the MT load above 37%, including the special test session.

At first I wondered if running SAC as a front end for a few days had made anything change, but it didn't seem likely.
Then I noticed that the buffers were set at 4 x 1024, instead of 2048, as I'd been doing to compensate for the high MT loads and more frequent freezes/crashes.

So, I tried testing the current, loaded project, using 1024 and 2048 buffers - closing SAW between each change.
The MT load is actually low (at 27-30%) with the buffers at 4 x 1024 and jumps almost from the start to 99%, with buffers at 4 x 2048.

Why is this?
I don't have a clue.

Tried Bob's suggestion just now - I assume this would need to be set for every channel in the session that has fx inserted?
There's no way to do this for output channels though. So out of luck if fx on those channels should be causing an issue... right?

At any rate, it is not affecting the current session's MT load at either buffer setting.
This session has 5 songs, each on its own set of tracks in order to manipulate the multiple vst plugs inserted on certain channels that would otherwise require automating settings from song to song.
But the 99% MT load issue is the same, regardless of what session I'm running.
In fact, it appears now to be related to the buffer settings.

There is a little talked about issue you may be running into... floating point math may cause a major issue in performance when trying to process values that approach zero... as the values get very small, the processing time can begin to slow down exponentially... this can occur with plugins using floating point math when tracks haver blank areas... between regions... in some cases this can actually bog the tracks down togimonstrous delays causing the performance percent value to instantly spike to 99%... there is a special opyion on the popup track menu called Enable Track FX Auto Bypass... this attempts to bypass plugin processing when the buffers values approach zero... try turning this ON and see if that solves the mysterious 99% jumps... that would suggest that your plugin on that track is susceptible to this issue.

Bob L

Well, I can report that using SAC as a front end to SAW64 v2 while editing/mixing (and like I usually do when tracking),
has brought the MT load back down to reasonably-looking percentages - but still with buffers higher than 4 x 512.

I am able to run the previous session that was troubling me at 30% MT load, with SAC up front.
That will do nicely.

But I still don't know why SAW64 has been jumping from 23% to 99% in the same session, but without SAC and with buffers at 4 x 2048.

jmh wrote:Just throwing out possibilities between your installations as I think of them. It is easy to overlook some of the basic stuff. Happened to me, once - or is it many times? Also is 'live mode' engaged (I actually don't know what (if any) difference this might make, as I've never used it)? Single cpu?

Live mode is not engaged.

Dell Precision T1700 Desktop - I gather it's ancient (10 years old)
Intel i7-4790 CPU @ 3.60GHz (4 core)
32GB ram
nvidia rtx3060
Win10 pro
RME Hammerfall pci (pcie?)
Hyperthreading is off
BIOS running in legacy mode....

Just throwing out possibilities between your installations as I think of them. It is easy to overlook some of the basic stuff. Happened to me, once - or is it many times? Also is 'live mode' engaged (I actually don't know what (if any) difference this might make, as I've never used it)? Single cpu?

jmh wrote:...also did you you do runas administrator?

I'm pretty sure I do - whatever was recommended and seemed to work best at the time. I'll check this stuff tomorrow, just to be sure/refresh the memory.

jmh wrote:Oh yeah, what power saving mode do you have set. If it is a 99%, and you can give it more load it probably is dynamically throttling down the cycles.

Power saving is always set to maximize performance - manually set, when I last installed a new ssd and Win. 10 pro (to remove Win 10 LTSC, which caused me all kinds of grief last year, or late 2023).

jmh wrote:Do you have hyperthreading turned on?

You might try locking saw to run in a particular (performance) core. I haven't played with any of this since my last dive.

https://www.sawstudiouser.com/forums/sh ... rthreading

I believe hyperthreading is and has always been turned off.

I haven't made changes to this pc since well before I started using SAW64 quite awhile ago (I've forgotten when - maybe in Jan., this year?).

...also did you you do runas administrator?

Oh yeah, what power saving mode do you have set. If it is a 99%, and you can give it more load it probably is dynamically throttling down the cycles.

Do you have hyperthreading turned on?

You might try locking saw to run in a particular (performance) core. I haven't played with any of this since my last dive.

https://www.sawstudiouser.com/forums/sh ... rthreading

Realized I had already made a safety copy of SAW64 v1.4, so I was able to do a bit of testing on it.
I haven't yet duplicated this test on SAW64 v2.

I was able to load up 12 regions onto 12 tracks, along with the following 64bit vsts, before the MT load went to 99% and stayed there:
27 SAW native (Echo, EQ & Levelizer)
17 64bit vsts (randomly chosen - 2 comps, 1 autogain, 1 noise reduction, 1 eq).

At this point, when engaging playback, the MT load will either jump to and stay at 99%, start at 99% and slowly creep downward, or start at 23% and slowly creep up to 99%.

Deleting any single region will drop the MT load to hover around 23-25%. While deleting any single vst plug from this list will drop the MT load to 22% and stay there.

Deleting a single native plugin will cause the MT load to hover in a wide range between about 85-99%, creeping back and forth as playback continues. While deleting 2-5 native plugins will cause the MT load to slowly creep down from 99%, or creep up from 23% (I haven't the patience to see if either goes to the other end of the scale).

Deleting 6 of the native plugins finally steadies the MT load at 23%.

I haven't found a combination of regions and fx that cause the MT load to sit steadily at any percentage between 24% and 99%.
This is not true in SAW 32bit, where I routinely saw/see steady loads at any percentage.

What is this telling me (us)?
I don't know.
But I do know that I've had far more freezes and crashes with SAW 64bit (v1, or v2) than I ever have in SAW 32bit. Maybe it's the plugins I'm using?..... They aren't causing crashes in SAW 32bit, using jBridge - at least not the ones I've bridged in the past year, or so. Maybe my pc, or Win10, or the persistent and forced invasion of Windows Update??? I can't tell. I'm just curious as to why the MT load in SAW64 acts the way it does.

cgrafx wrote:Just curious, are you using SAW independently or in conjunction with SAC?

What buffer settings are you using?

Good points -
Just SAW by itself at the moment. I find it easier to work on editing, when I'm only concerned with SAW... just my way, I suppose.
But it's a good point - that the MT load might be lessened in SAW, if I used SAC as a front end all the time... maybe need to try this.

I'm running buffers quite high lately, while editing and mixing - because of this change in the MT load - currently 4x2048 in both SAWs.
That's much higher than I'm used to. Thinking back, I started upping buffers to 1024 for editing/mixing in SAW64 v1.4. I believe because I started using plugins that were demanding more of the pc than before.
This current observation of the 99% MT load in SAW64 v2 is likely an extension of that. But I'm curious as to why I'm only now seeing it.

I haven't taken the time yet to install SAW64 v1.4 again, for the sake of testing.
It'll mean having to reestablish a bunch of plugins to the new install, which will take some time.
But I plan to do so soon.

MMP wrote:I haven*t found the meters to be accurate in either V1 or V2 and have it pinned on most mixes without any noticeable issues.

I'm experiencing steady (if not accurate?) MT load readings in SAW 32bit.
Nothing like the sudden jumps from 23% to 99% in SAW64, so far.
I did witness a steady, relentless climb from 32% (IIRC) to 99% after adding a single extra instance of a plugin to one channel, while testing in SAW64 once - which I haven't flushed out much yet.