|
Post by mrholmes on Feb 18, 2021 20:04:09 GMT -6
I found this blog post very inspiring. www.ryanschwabe.com/blog/96kIf aliasing is such an dangerous thing for mixing digital why do some plug ins sound great but they introduce aliasing artifacts as well? Chris of Airwindos addresses the problem @higher sample rates with his Ultrasonic plug in. Some developers get minimal aliasing done others seem not to care. For the future. Will this problem be killed one day?
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 18, 2021 22:25:21 GMT -6
|
|
|
Post by jcoutu1 on Feb 19, 2021 0:35:16 GMT -6
|
|
|
Post by Guitar on Feb 25, 2021 16:43:27 GMT -6
I'm positive that 96K sessions sound better in my my DAW, but my computer can't really handle it, so I run 48K sessions. I guess it's not worth losing sleep over, but maybe one day I'll have some insane 8 core CPU computer or something than can do busy 96K projects. If I could do it comfortably, I would do it as the regular thing.
|
|
|
Post by viciousbliss on Feb 25, 2021 21:04:05 GMT -6
I'm positive that 96K sessions sound better in my my DAW, but my computer can't really handle it, so I run 48K sessions. I guess it's not worth losing sleep over, but maybe one day I'll have some insane 8 core CPU computer or something than can do busy 96K projects. If I could do it comfortably, I would do it as the regular thing. After reading Dan's posts and going back to what Fabien from TDR had been saying about high rates, I just converted all my 88/96 stuff to 44k and used Metaplugin to upsample stuff like Hornet tape. When I originally switched to 88/96 almost 5 years ago, I was using very different stuff. I hadn't been able to buy the Cranesong stuff and a lot of what I use now didn't exist. So my 44k plus upsampling mixes and masters sound better than what I did at 88/96. Plus I barely break 50% on my cpu now. I've been running a Ryzen 1700. It's barely overclocked, set to 3.2. At 88/96 I needed it set to at least 3.7. Supposedly UAD doesn't even save cpu resources anymore. This guy Undertow over at GS was explaining how the extra stuff the computer has to do to communicate with the satellite or whatever ended up being less effective than just getting a better cpu. Kinda makes sense because when I first got my Octo, my sessions would run out of cpu even though less native plugins were enabled. Doesn't really matter much now. You could probably get by with something like a Ryzen 1600 easily. Overclocking them is simple to do and you can use the included Wraith Spire. For the RAM, I just have a 3000 speed with 8 gigs. X-Amp was off and paging file set to auto. The computer slowed to a crawl following MixxBusTV's advice on that stuff, but he was running an Intel 9900k I believe. So I'd say try it without overclocking and then if you do overclock, just change the cpu speed a little and don't mess with the RAM stuff beyond setting it to the stock speed. At least for these original Ryzen chips. My board is B350 gaming pro. I'm sure the newer stuff is a lot more advanced and perhaps even more flexible.
|
|
|
Post by Guitar on Feb 26, 2021 4:14:37 GMT -6
I'm positive that 96K sessions sound better in my my DAW, but my computer can't really handle it, so I run 48K sessions. I guess it's not worth losing sleep over, but maybe one day I'll have some insane 8 core CPU computer or something than can do busy 96K projects. If I could do it comfortably, I would do it as the regular thing. After reading Dan's posts and going back to what Fabien from TDR had been saying about high rates, I just converted all my 88/96 stuff to 44k and used Metaplugin to upsample stuff like Hornet tape. When I originally switched to 88/96 almost 5 years ago, I was using very different stuff. I hadn't been able to buy the Cranesong stuff and a lot of what I use now didn't exist. So my 44k plus upsampling mixes and masters sound better than what I did at 88/96. Plus I barely break 50% on my cpu now. I've been running a Ryzen 1700. It's barely overclocked, set to 3.2. At 88/96 I needed it set to at least 3.7. Supposedly UAD doesn't even save cpu resources anymore. This guy Undertow over at GS was explaining how the extra stuff the computer has to do to communicate with the satellite or whatever ended up being less effective than just getting a better cpu. Kinda makes sense because when I first got my Octo, my sessions would run out of cpu even though less native plugins were enabled. Doesn't really matter much now. You could probably get by with something like a Ryzen 1600 easily. Overclocking them is simple to do and you can use the included Wraith Spire. For the RAM, I just have a 3000 speed with 8 gigs. X-Amp was off and paging file set to auto. The computer slowed to a crawl following MixxBusTV's advice on that stuff, but he was running an Intel 9900k I believe. So I'd say try it without overclocking and then if you do overclock, just change the cpu speed a little and don't mess with the RAM stuff beyond setting it to the stock speed. At least for these original Ryzen chips. My board is B350 gaming pro. I'm sure the newer stuff is a lot more advanced and perhaps even more flexible. Let's say I want to implement an oversampling workflow in my projects. Would I add just a single oversampling plugin last in the insert chain for every channel? Or is there more to it than that? What's an ideal implementation? I'm interested in this idea if it sounds as good as 96K.
|
|
|
Post by kcatthedog on Feb 26, 2021 4:35:51 GMT -6
@ viciousbliss not following your comment about UA not saving cpu, as ua runs and does their oversampling on its own chips, how can it not save computer cpu?
|
|
|
Post by Guitar on Feb 26, 2021 4:42:31 GMT -6
Audio threads, whatever they're called, I/O into the DAW, requires some CPU regardless, I think was the point.
|
|
|
Post by askomiko on Feb 26, 2021 5:00:54 GMT -6
How is 96k a problem today, when computers have been powerful enough for 100 track for 20 years? Have we regressed? Nebula plugins?
|
|
|
Post by Guitar on Feb 26, 2021 5:26:34 GMT -6
Yeah, maybe the plugins got heavier. I could run 96K a little easier with "old school" plugins for sure, but I use a lot of intenstive stuff, for example virtual instruments, Elevate limiter, this stuff chews up the real time audio quickly.
Maybe my computer just sucks, I have no idea. But 96k chokes up pretty quickly with the kind of work I do. I could change my workflow, I guess, but I like my workflow.
|
|
|
Post by kcatthedog on Feb 26, 2021 5:31:27 GMT -6
I wonder if it’s the vi? I ran 96 in logic, with UA plugs and others on a 2012 mbp i7 typically without problems other than when my hard drives got full,’: 2 internal ssd snd 16g ram ?
|
|
|
Post by Guitar on Feb 26, 2021 5:49:12 GMT -6
Yes the virtual instruments are probably the main offender in my case. If I was doing simple old school recording / mixing it would most likely be a lot more straightforward with a high rate session.
I guess I could bounce all of them down, maybe that would work, freeze all the tracks and mix, might need to try that workflow.
|
|
|
Post by Guitar on Feb 26, 2021 8:12:43 GMT -6
I made some improvements to my default template. I reduced the amount of un used plugins, channels. Also turned off all drum mixing plugins until they are actually needed.
Also started freezing instruments when the performance is done. It's a trivial thing to do, that I needed to learn. It made me be able to run a session at 64 buffer rather than 4096 buffer, big difference.
Still not running at 96K, still haven't figured out how to implement system-wide oversampling although some of my plugins do oversample already.
|
|
|
Post by viciousbliss on Feb 26, 2021 8:43:41 GMT -6
After reading Dan's posts and going back to what Fabien from TDR had been saying about high rates, I just converted all my 88/96 stuff to 44k and used Metaplugin to upsample stuff like Hornet tape. When I originally switched to 88/96 almost 5 years ago, I was using very different stuff. I hadn't been able to buy the Cranesong stuff and a lot of what I use now didn't exist. So my 44k plus upsampling mixes and masters sound better than what I did at 88/96. Plus I barely break 50% on my cpu now. I've been running a Ryzen 1700. It's barely overclocked, set to 3.2. At 88/96 I needed it set to at least 3.7. Supposedly UAD doesn't even save cpu resources anymore. This guy Undertow over at GS was explaining how the extra stuff the computer has to do to communicate with the satellite or whatever ended up being less effective than just getting a better cpu. Kinda makes sense because when I first got my Octo, my sessions would run out of cpu even though less native plugins were enabled. Doesn't really matter much now. You could probably get by with something like a Ryzen 1600 easily. Overclocking them is simple to do and you can use the included Wraith Spire. For the RAM, I just have a 3000 speed with 8 gigs. X-Amp was off and paging file set to auto. The computer slowed to a crawl following MixxBusTV's advice on that stuff, but he was running an Intel 9900k I believe. So I'd say try it without overclocking and then if you do overclock, just change the cpu speed a little and don't mess with the RAM stuff beyond setting it to the stock speed. At least for these original Ryzen chips. My board is B350 gaming pro. I'm sure the newer stuff is a lot more advanced and perhaps even more flexible. Let's say I want to implement an oversampling workflow in my projects. Would I add just a single oversampling plugin last in the insert chain for every channel? Or is there more to it than that? What's an ideal implementation? I'm interested in this idea if it sounds as good as 96K. Well Metaplugin gets inserted and then you can add a chain of plugins inside each instance. The type and amount of oversampling is selectable. You can compare plugins on their own or inside Meta and see what you like better. Sometimes I prefer no OS. Check out Dan has said about IMD and higher rates here. The type of plugins you run makes all the difference. Look at the instance counts on the Waves servers. You can easily run hundreds of most of their old plugins at 96k, but only like 8 H-Reverbs. The chip in their Extreme Server is close to a 6700k in performance. I've written pretty extensively on my experiments here and on GS. Acustica plugins can easily cripple the best processors because the single core speed doesn't change a ton. I use cpu.userbenchmark.com for reference because I did my own test there. The best single core speed is around 150, I think. My Ryzen 1700 scored around 115 at its best. The multi/64-core rating was around 1200. Using 44k with a few oversampled plugins still gave me almost double the cpu performance. So it's like I've turned my 1700 into a computer that scores 215 and 2400 if I were still using 88/96. Microcenter has some pretty good processors for under $200. You might be able to get some original Ryzen cheaper used somewhere. It all really boils down to what you use. Look for instance count tests on youtube. I booked marked the GS thread where Undertow was talking about UAD not saving cpu cycles anymore. I'd have to dig it out. I'm getting the feeling that maybe the tech was most useful when computers couldn't run much of anything. There were times where I could deactivate a cpu intensive native plugin and swap in something from UAD to avoid cpu overload. Suppose you could also swap some other native choice in too. But maybe it's possible that the UAD was contributing to the cpu overload in the first place.
|
|
|
Post by jcoutu1 on Feb 26, 2021 8:47:23 GMT -6
I can't even imagine worrying about that stuff. Who has time for all that.
|
|
|
Post by svart on Feb 26, 2021 11:29:06 GMT -6
I can't even imagine worrying about that stuff. Who has time for all that. Yeah I dunno. I run dozens and dozens of plugs on dozens of tracks at 88.2k and barely see 30% utilization on cpu/ram at 3ghz. I'm not worrying about playing tricks and aliasing and whatever a long as it works and sounds good.
|
|
|
Post by Guitar on Feb 26, 2021 12:11:03 GMT -6
I tried the MetaPlugin thing today and it didn't work for me. Too much extra complication, can't even save presets, or create organized plugin folders, which is a big time waste. I know everyone has a different workflow. I'm honing mine in right now, seems like I'll stay at 48K given how I usually work. Maybe 96K for the occasional project that 'needs' it.
I work a lot at very low buffer settings, for virtual instrument monitoring, so that's probably part of why I'm reporting a lot of high CPU meters. Some of my main plugins are super heavy too, like Elevate, so that's another part of it. Regarding what I've been reporting about my sessions.
|
|
|
Post by askomiko on Feb 26, 2021 14:47:45 GMT -6
Cliff from Fractal audio wrote a big post about bad quality of coding nowadays. Programs use gigabytes of RAM to do stuff that would actually work on a 386 with proper coding. Lazy coding, why bother when it works, at least sort of.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 26, 2021 16:00:33 GMT -6
Much bigger problems are sidechain aliasing leading to ineffective, oft random processing and oversampling with minimum phase filters. The former sounds stupid most of the time and results in poor dynamic control. The latter causes well over a thousand degrees of phase shift in the treble and wrecks guitars, cymbals, and space. This is the same thing that happens in bad converters, eg Alesis Adat, Focusrites, some Behringers. This kills the effectiveness of many otherwise great plugins from Overloud and Noise Ash simply because you can’t use them in many cases without harmful phase shift.
Audible Harmonic aliasing is easy to avoid. Gainstage! Rip out bad plugins. Stop putting Decapitator in every track. Use a milder distortion plug like the Fuse ones, Klanghelm SDRR 2, Shattered Glass, UAD 2, or even the PSP Infinistrip pres gainstaged right. Even Waves hit lightly is okay.
|
|
|
Post by mcirish on Feb 26, 2021 16:09:52 GMT -6
Can any of you guys mix a 100 track project ITB @ 96K? I found it impossible without doing a ton of bouncing down. It just wasn't worth the aggravation. I use 44.1k and set any plugin that can oversample to do that during mixdown.
|
|
|
Post by mrholmes on Feb 26, 2021 16:16:02 GMT -6
Much bigger problems are sidechain aliasing leading to ineffective, oft random processing and oversampling with minimum phase filters. The former sounds stupid most of the time and results in poor dynamic control. The latter causes well over a thousand degrees of phase shift in the treble and wrecks guitars, cymbals, and space. This is the same thing that happens in bad converters, eg Alesis Adat, Focusrites, some Behringers. This kills the effectiveness of many otherwise great plugins from Overloud and Noise Ash simply because you can’t use them in many cases without harmful phase shift. Audible Harmonic aliasing is easy to avoid. Gainstage! Rip out bad plugins. Stop putting Decapitator in every track. Use a milder distortion plug like the Fuse ones, Klanghelm SDRR 2, Shattered Glass, UAD 2, or even the PSP Infinistrip pres gainstaged right. Even Waves hit lightly is okay.
Agree on the saturation part just a few hold up. Will grab Elysia Cascade with the new 29 bucks voucher by PA. My yaw dropped sounds like real tubes and I heard a lot of them----
|
|
|
Post by mrholmes on Feb 26, 2021 16:17:51 GMT -6
Cliff from Fractal audio wrote a big post about bad quality of coding nowadays. Programs use gigabytes of RAM to do stuff that would actually work on a 386 with proper coding. Lazy coding, why bother when it works, at least sort of.
link?
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 26, 2021 16:34:58 GMT -6
Cliff from Fractal audio wrote a big post about bad quality of coding nowadays. Programs use gigabytes of RAM to do stuff that would actually work on a 386 with proper coding. Lazy coding, why bother when it works, at least sort of. Exacto. Look at what Variety of Sound did ten years ago with x86 assembly and synth edit and now consider that there are less advanced plugs than his coming out today from big brands. Overloud is the Simulanalog guy and he still crams more into no CPU use plugs than anybody. The same with PSP and PSP doesn’t use bad oversampling or 32-bit rounding in their plugs. Overloud looks incompetent in their fx coding practices compared to them. Or the Softube Weiss EQs vs Sonnox Oxford. What the hell? One is legendary, the other has a heavy GUI and some rounding. Sonnox (Sony Oxford Dynamics) and Waves (Renaissance and L1) had effective digital dynamics and peak limiting in the 90s! They had high imd but the worked! The compressors couldn’t peak limit but the workflow worked and the coding was great. The spent hundreds of thousands to millions in r&d with the best audio coders around. Compare that to most of what comes out now and the current stuff is a slapped together joke. Most of the current Waves plugs (with some exceptions like the sick RS124) can’t hang with their late 90s output. Oh and the tape plugs all suck except for the two guys who build mathematical models of tape machines almost a decade ago, including programming the high frequency bias oscillators. U-he Satin and the UAD Ampex ATR 102 are still unmatched. When they came out, they are computers and DSP chips alive. Satin has no competitors for it’s flexibility. They both have to run at 352 or 384 kHz. Nobody else has even bothered since and the sound is proof. Yes you need to read the Satin manual or know how to calibrate a tape deck to use it without presets but it’s worth it. Then we have ridiculous bullshit like the Pulsar compressor plugs where the behavior on 1x oversampling sucks so bad they put in a lookahead, the distortion sounds cheap, and the cpu use is insane if you enable oversampling to make the behavior better. Compare them and Acoustica ir plugs to the Tokyo Dawn compressors and they just can’t hang.
|
|
Deleted
Deleted Member
Posts: 0
|
Post by Deleted on Feb 26, 2021 16:37:17 GMT -6
Much bigger problems are sidechain aliasing leading to ineffective, oft random processing and oversampling with minimum phase filters. The former sounds stupid most of the time and results in poor dynamic control. The latter causes well over a thousand degrees of phase shift in the treble and wrecks guitars, cymbals, and space. This is the same thing that happens in bad converters, eg Alesis Adat, Focusrites, some Behringers. This kills the effectiveness of many otherwise great plugins from Overloud and Noise Ash simply because you can’t use them in many cases without harmful phase shift. Audible Harmonic aliasing is easy to avoid. Gainstage! Rip out bad plugins. Stop putting Decapitator in every track. Use a milder distortion plug like the Fuse ones, Klanghelm SDRR 2, Shattered Glass, UAD 2, or even the PSP Infinistrip pres gainstaged right. Even Waves hit lightly is okay.
Agree on the saturation part just a few hold up. Will grab Elysia Cascade with the new 29 bucks voucher by PA. My yaw dropped sounds like real tubes and I heard a lot of them----
It’s worth 30 bucks imo. The tube upward compression and push on that is insane on some things but I warn you it can sound 2d and digitally sheened.
|
|
|
Post by mrholmes on Feb 26, 2021 17:59:26 GMT -6
Agree on the saturation part just a few hold up. Will grab Elysia Cascade with the new 29 bucks voucher by PA. My yaw dropped sounds like real tubes and I heard a lot of them----
It’s worth 30 bucks imo. The tube upward compression and push on that is insane on some things but I warn you it can sound 2d and digitally sheened. Ok which tube sim is your favorite?
|
|