Is THIS the TRUE future of audio plugins??

[@PaulThird]

www.gpu.audio/paulthird FYI I just I noticed I accidentally edited a small section out where I talk about the gpu's needed. As far as I'm aware the tech only runs on AMD & nVidia right now as gpu audio have partnerships with them both so they are working together to develop the tech. They have said that expensive desktop GPU's are unneccisary for this tech and consumer grade gpu's (laptops etc) should be more than enough. Take that information as you will. In depth lecture about the technology kzbin.info/www/bejne/i568n6J8pJV1mJo Also.. What are your thoughts, do you think that utilising gpu is the future of audio production?

[@joeking9760]

This to me sounds like the return of the sound card, cause basically we will just slap an "audio card" into one of the PCI slots. I discovered the website gpu.audio a while ago and immediately thought about the implications and possiblities. While at its current state the GPU is being utilized I think there will be a more dedicated ways of handling audio processing in the future.

[@Bthelick]

The business model of bespoke plugins on a hardware platform is still a dangerous one. UAD leaving the hardware space is a sign of the times, and that's a LATE example. A bespoke daw should be even more dangerous. The adoption rate will be super niche until they can get existing platforms (vst) translated.

[@PaulThird]

They are very tight lipped on the daw idea. I have no idea where their heads at with it tbh. I don't even know how it would work. Maybe something like luna where it's only gpu plugins 🤷‍♂️

[@SoundcastStudios]

You have actually made me far more interested in GPU audio than I was before. I originally saw it as "ugh another plugin" but it's far more than that. It's an idea as well. Very interested to see what the future holds for audio.

[@PaulThird]

🤓🤓

[@helmanfrow]

I've been dreaming of audio on GPUs since I read a paper saying it couldn't be done 15 years ago. Been following GPU Audio (great name) for a while now. Very excited.

[@PaulThird]

🤓🤓

[@henrikpetersson3463]

UAD-1 used GPUs.

[@PaulThird]

Not the computers on board gpu though. Plus interesting why they ended up moving to sharc chips for their DSP 🤔 From what I read there were driver issues with the computers mistaking the uad-1 cards for the nvidia GPU. I'll ask my uad guy exactly why but makes sense why this would happen

[@henrikpetersson3463]

@@PaulThird Well Chromatic Research stopped making the MPACT2 chip they used for UAD1 and the SHARC processors had really good and fitting specs for the time so it was kind of a natural move. There has been more powerful dual core SHARCs released since, also with an ARM core. But according to UA they had problems implementing them.

[@makemusicordie]

Keep it up, Paul! Makes me glad to know you’re starting to get more work mixing etc! Clearly you are extremely dedicated, so it’s well deserved!

[@PaulThird]

🤜🤛

[@Bthelick]

Shame it can't run the vst standard on the GPU, only bespoke stuff. Having said that, I've been hoping for real time ray traced reverb. Let's hope someone manages it because my reverb coding isn't up to par.

[@PaulThird]

Still early days. I'm just gonna keep an eye on how it develops over the years 🤓

[@dirtyharry1881]

@@PaulThird It's not really that early. I'm hearing about this for years. I hope Im wrong, but this ain't going anywhere. It will be another example of good ideas that were not adopted and finally got surpassed by sth else, like Laser Disk etc.

[@PaulThird]

True, the expansion of it all hinges on other devs taking the tech on board and integrating it with their algorithms but surely now that we have physical plugins using the tech you would presume lots of devs would at least look to utilise it to try and get an advantage over the competition. Could be a case of simply offering gpu versions of what they already have. Gulfoss master being one. Pain in the hoop on my i9. 15% cpu haha convolution outside of reverbs I could imagine would be a nightmare to pull off but if acustica pulled it off.. Would completely change the landscape for them. It's the main thing holding them back. I'm an an idealist so I like to think it'll progress but I undetstand that it may have a ceiling. but it does make sense for devs to experiment with the gpu tech and see if it can be integrated and expanded over the years

[@dirtyharry1881]

@@PaulThird that depends on many factors. For example: compatibility. Macs are still a big thing in audio. Does it work wiith apple chip? Does it offer anything in this case? Will the x86 platform survive the decade? If it doesn't, will it still make sense to utilise the GPU? And GPUs in the PC world nowadays are pretty expensive. Does it matter which platform you're on? Does it matter what GPUs you use? Fan noise in GPUs is also a thing. Some of them sound like an airplane taking off - no way to predict, it's a lottery! Adding another expensive and unpredictable variable in the audio setup, seems to add one more layer of complexity and variability to an already convoluted chain of expensive shit... Anyway, if I were an investor, which I'm not, I would not open my wallet so easily...

[@PaulThird]

Said in the video that it has M1 compatibility. In my pinned comment I included info about the gpu's needed. Nvidia and AMD only just now but they have stated you don't need expensive gpu's. Consumer grade gpu's in laptops will be enough according to them. I think it's worth experimenting with seeing as it's of no cost to us. You can always try demos and see what the fans you have are like. I mix on headphones anyway so I can't see it being much of an issue for guys like me but I hear you. I just think it's worth looking into as another option. If it's too little too late then it is what it is. The way I see it with my i9 I'm still really tight for sessions and if there's gpu there to utilise then I could really do with any extra processing as I don't have a couple of grand spare for a new higher performing desktop. I'll just sit back and keep an eye on how it develops

[@nedim_guitar]

Interesting stuff! Congrats on being busy with clients and the podcast!

[@PaulThird]

🤜🤛

[@johnvender]

Like wow. Using the GPU to take the load off the CPU seems so obvious but there is no argument this is far from trivial when it comes to work that is inherently sequential. Big thumbs up to these people. Loved your call out to Tunca :)

[@PaulThird]

🤓🤓

[@Hassan_Omer]

More power to you Paul. Thank you for sharing this insight and congratulations on getting busy with client work.

[@PaulThird]

🤓🤓

[@tonyadams6985]

Always informative. Thanks for your effort in delivering great content!

[@PaulThird]

🤜🤛

[@0e0]

amazing video and explanation..and love to see your confidence.. in full support

[@PaulThird]

🤜🤛

[@mikelo303]

One word: FINALLY! And everybody was telling me this is not possible on music forums. I was advocating that after Sony announced GPU based Tempest Engine using ray tracing technology for audio years ago before PS5 launch. And now what, audio smart asses? Explaining IT guy that something in IT is not possible? Where are you now? Even plugin developer were arguing. Eat that! I just hope all those major companies won't cheap out and start using and licensing it. I hope small developers will force them to do that. I hope this will become industry standard. And we need new style of DAW's based on things we can automate and overcome, not on old music desks. Like, channels with automated gain staging, gain compensation. Automated plugins like Sonible. Preformatted templates for music production, with busses and mastering chains set up. Things that are always required but are still necessary to be build every time step by step. Channels that process basic tasks automatically (gain -12, clip to zero, add plugin chain like izotope, communicate between plugins eq like Sonnox Claro, remove resonances, basics automated, etc.). There are plenty of things a computer can do to speed up our work in digital format based on what we need and not on how it was always done on the hardware desk.

[@PaulThird]

I utilise as much software as I can. If it gets me there quicker without massively losing out on sound quality then I don't care what anybody thinks. Many people are scared of AI but in the right hands and used in specific ways for specific tasks.. Can only further help the architect build upon his idea

[@RealHomeRecording]

Eat it, indeed!!!!

[@bertfader]

its a great idea, but seems without support for the big hitters like Omnisphere, KONTAKT etc. it will dead in the water, or am i missing something? hope you can enlighten me to the potential for using other vst's. i have just bought 2 NVidia grid headless graphic cards for processing audio and video, but finding info about using these for audio is thin on the ground. i realise the vst's I've mentioned probably need them to create an updated version for this. I guess its early days for this.

[@PaulThird]

It's beneficial for those who can't afford expensive processors or work remotely with laptops with lower performing processors. The biggest vst benefit I can see is from convolution. If they can integrate it with convolution plugs like acustica that would be pretty big

[@bertfader]

thanks for your reply. I'm wondering if there DAW will incorporate some type of vst wrapper, the potential is huge and if your using big sample libraries it will open it right up to running huge orchestral templates from just 1 pc@@PaulThird

[@ronnielad1928]

Keep it rocking my dude,, Lovin the podcast as well,, mixes are soundin awesome 🤘🤘

[@PaulThird]

🤜🤛

[@matt_nyc_audioengineer]

Great video Paul!! Just gave it a try, the reverb sounded like crap but my CPU meter read 0 lol!! How cool, I'm looking forward to seeing where this goes. Sorry for this but I wouldn't be fan boy if I just sat idly by. You made a comment about M1 chips not being good for machine learning. I would double check that fact hahaha!!! That's absolutely not true. Here is just one quote "The M1 is a breakthrough for machine learning at the edge, with the ability to execute 11 trillion operations per second, achieving up to 15x faster machine learning performance. Using cutting-edge 5-nanometer process technology, the M1 is packed with 16 billion transistors!". Sorry wouldn't be a proper Apple fanboy if I didn't say anything!! 🤣🤘🏻 Thanks again for the video bud!

[@PaulThird]

Hmm here is what the GPU audio CEO said in an interview "The second point is that CPUs are occupied with many processes already and there is a huge variety of hardware systems for users today, for example the new MacBook M1/M2 systems. Regardless of the fact that one silicon chip is both powerful and highly efficient - it can run simple plugins, and lots of them - but it can't run things like live machine learning without encountering significant issues. GPUs enable this machine learning to happen, so that's one of the great use cases we have for desktop and laptop grade systems. We've spoken to people with the latest Macs and they are encountering CPU limits with Spatial Audio. In terms of other resource intensive technology - you can look at what's going on with 3D and 2D graphics and machine learning algorithms; most of them work on GPUs, utilising them as a hardware platform. If the models need to be refined, we can train the models quickly. It's the same for audio - if you want to process some neural network models to some audio buffers, you have to use GPUs. Adaptive tools like noise removal, tone transfers, deep AI effects are some examples of those next generation machine learning techniques we can design and execute using GPU power." So I don't know if its the 'live' machine learning aspect but I know a lot of guys with M1's still complaining about acustica and convolution maxing out. Atmos as well as guys working film and stuff Think about tone empire and baby audio machine learning.. Both fixed at 44.1 and tone empires is an absolute mess of a plugin so there is still lots of limitations to machine learning right now

[@matt_nyc_audioengineer]

@@PaulThird Strange it certainly is conflicting information. If you Google the subject you will see the results will back up what I was saying HOWEVER I would tend to put a little more trust in the people who actually use the machine to do those kinds of things. I don't have any experience with it personally obviously lol. I'm just going off what I read and have heard. Now I have been using the M chip since it came out. I got an M1 Mini right away and then an M1 Max Mac Studio as soon as they were released as well. So I can testify to the fact that they are super fast and efficient but that's about the extent of my knowledge. This is very interesting. As I said, a quick google search will show MANY results that favor my comment but I don't know if those people are just going off benchmarks, stats, or real-world experience. I wonder what the truth is!! This is going to bother me now... I even just went and asked Chat GTP and Chat GTP said M1 machines are great for Machine Learning 🤣 Also, at least in Logic, I'm not having issues with any Acoustica stuff. As a matter of a fact I can run many more Acustica plugins now than I ever have. But you need to remember I am using a DAW made by Apple on a chip made my Apple lol. It would be quite embarrassing if they didn't get that right lol! I know many colleagues with the same machine in other DAW's that are NOT having the same experience. Specifially Pro Tools.

[@PaulThird]

Tbh man it's a hard one cause obviously GPU are promoting their own tech and M1 is the biggest hurdle they have in regards to the tech being relevant today. So I take it with a pinch of salt but at the same time we do only have to look at the machine learning plugs that have been released and listen to the devs talk about the limitations caused by computational processing. They are all learning at low sample rates and even the tone empire compressor isn't exactly cpu friendy for a plugin that can't be oversampled. We both aren't doing live machine learning to make plugins so we are pretty much in the dark but I would look to what's been created as proof that cpu isn't there yet to utilise machine learning to its fullest, maybe argue its the devs own limitations but odd how 2 seperate one's I've seen have had the same limitations and issues

[@matt_nyc_audioengineer]

@@PaulThird Agreed...

[@cshortlife9179]

GPU audio looks great but I think GPU fan noise is going to be an issue particularly with laptops, but also with desktop computers.

[@akagerhard]

Interesting point.

[@PaulThird]

What about if you mix on headphones though 🤔

[@akagerhard]

@@PaulThird noise is more of an issue for recording than for monitoring. Yes, in a perfect world you have the computer in another room, but in a perfect world CPU isn't an issue anyway.

[@PaulThird]

Would you not just record further away from the laptop though? Haha surely we're not talking AC loud 😅

[@akagerhard]

@@PaulThird Of course you'd try to record further away from your laptop in a professional setting. I'm trying to get you to think about bedroom-producers that have to press record on their own and have a mic pretty close to their laptop or audio-pc, because walking 3 meters is not an option in every bedroom and costs a lot of time when you think about comping a whole song line for line. GPU is surely rather aiming at people that are lower on funds, or am I reading the room all wrong? I think GPU usually not being cooled by a noise-optimized cooling system (like my CPU is..) is an interesting point.

[@Taylor370z]

I don't know if there is a way to tell if it's getting offloaded to the GPU, but at least in FL Studio is doesn't seem to be. In the discord they did mention that Image Line has been tough to work with and something about it being a little tricky with FL for some reason. I tried the most recent release and seemed to be the same. So just FYI for FL users, they work, but, not sure if they are using GPU.

[@lib747]

Paul, excellent analysis and insight. Historically, the biggest roadblock to audio processing on GPU was data transfer to/from the GPU, and it looks like they've figured it out. I'll take it for a spin on the M1. Truly exciting!

[@PaulThird]

🤓🤓

[@Fwuzeem]

Only if it's 0 latency it is sequential, but if it's not they'll use the buffer size to process everything in that time window. The larger the buffer size means that the CPU can work on other things, as it outputs the same amount but not as often, so it's easier on the CPU, but it's still taking the same amount of time to output it.

[@tripwillie2]

Congrats on your success Paul! You've earned it!

[@PaulThird]

🤜🤛

[@rondadornocturno1824]

There was a software for use GPU in Acustica Audio Nebula called Cuda and this didn't work well, and I talk about 10 years ago

[@PaulThird]

Hopefully 10 years is enough time where the tech has moved on to begin to integrate it 🤞

[@lance134679]

This is exciting tech, for sure. I have a decent GPU for gaming as recording is my hobby, but it sits there doing (almost) nothing while I'm working in the DAW. Might as well put that expensive GPU to work. 👍😄

[@PaulThird]

🤓🤓

[@B1ggz13]

what exponential amount of bytes is going to be necessary in order to emulate analogue gear?

[@PaulThird]

To emulate properly? Enough where the anti aliasing filters don't affect the audible high end phase, enough where the saturation and clipping can be pushed without any audible aliasing and enough where the algorithm can act like convolution where it models every move the hardware makes per dial change in the hardware. To do it right you need a lot of processing

[@mageprometheus]

Thanks Paul. I saw they support teams blue and red. I wonder if they have plans to add Intel ARC GPUs, I couldn't see it on the site. I personally don't have a problem, 16 cores and 128Gig of ram runs reaper like a dream, but I love toys and if the plugins were excellent and priced right, I might pop a new graphics card in. It would help with Resolve processing too. Take care, my guy.

[@PaulThird]

🤓🤓

[@Fwuzeem]

Can't get them working with an AMD graphics card, but they work on the Mac M1 just fine

[@PaulThird]

Works on my nvidia laptop but they said they have a partnership with AMD but that's all I know

[@samueloluyinkaojomu6548]

Good day Paul. Have you laid your hands on the new Audient ID24 interface yet.

[@PaulThird]

Nah, I've got an apollo x6 and evo sp8 as my main setup and revelator io24 & topping dx1 for my portable setup. I'm cool for interfaces now 🤓

[@samueloluyinkaojomu6548]

@@PaulThird what does the Apollo has better than your Audient ID44

[@PaulThird]

Real time dsp, unison preamps & 6 standalone line in & outs (no combo Jacks) so more line options as well as cleaner ADA for analog gear as well

[@samueloluyinkaojomu6548]

@@PaulThird I don't believe the ADDA would be better than what Audient has. And I'm not a fan of dsp anyway.

[@PaulThird]

Its scientifically better that's a fact. Whether that's a big audible change, very much doubt it but there are measurable differences which for me and this channel doesn't work cause I need to remove potential variables to make my results as credible as possible. Plus only being able to bypass the preamps on 2 channels isn't great compared to apollo which has all the line inputs completely seperate from the preamps. The ID preamps aren't as transparent as the evo preamps so using them for analog gear isn't as advisable compared to an option that completely bypasses preamps. However if you are solely using preamps, limited gear, and don't care about real time dsp then yeah id44 mk2 is definitely a good shout Once I got into the unison preamps and real time dsp for tracking.. No turning back, Its became a big thing for me

[@markmallinder7618]

Very interesting. Sounds like the future to me.Bring it on! And the sooner the better!

[@PaulThird]

🤓🤓

[@SG-4u]

A few years ago I scoured the internet for a solution like this, and there were very few practical options. I wonder if it isn’t just too late at this point. Once I moved to the M1 CPU I can run pretty much everything I need without much headache.

[@PaulThird]

Take it you don't use a lot of convolution plugs? Haha

[@SG-4u]

@@PaulThird Not especially. But even there, I wonder if technology wise plugin developers are going to kind of skip over convolution now that we have AI/machine learning tech which can produce superior results without the CPU hit. But I'll definitely check out GPU Audio, I'm all for anyone pushing the tech forward. Looks like they are going to release some distortion and dynamics plugins soon, and who doesn't want more of that to play with.

[@PaulThird]

Machine learning isn't there yet. I've tested a lot of the offerings and it's incredibly buggy with inconsistent results. Machine learning in regards to analog emulations has a LONG way to go

[@Fwuzeem]

Machine learning works really well with tenser cores, which are in RX cards, but not in other ones.

[@RealHomeRecording]

Very nice video Paul. You're making me regret not setting up that interview with these guys back when I had the opportunity... The nice thing about this development is I can justify a new graphics card purchase and use it as a partial tax write off.

[@PaulThird]

They've said you should only need consumer grade gpu but if it a tax right off then...

[@seenbelow]

Badly implemented schedulers are the bane of my existence in my professional life. This is why old specific hardware that's less powerful can still outperform several of the highest rated but general usage CPUs. This is why native DSPs are still powerful to this day for the code they were developed around. I'm glad we're coming around to use GPUs to audio rendering, so many other areas in IT are already vastly invested in this for many years, and GPUs are so powerful these days...

[@PaulThird]

🤓🤓

[@surgerywithoutknives2353]

Are you on the KZbin naughty step now ? 😂This didn’t appear in my notifications and Ive been subscribed with the bell for over a year and a half ….

[@PaulThird]

A channel member said the same thing. Be cause I didn't upload for 3 weeks. Algorithm is a bitch when you aren't consistent. Happened last time I took a break 😔

[@_thevaporz]

You won't even need a discrete gpu. Ryzen zen4 rdna3 integrated graphics are approaching ~1660ti level performance .

[@RealHomeRecording]

What? Nice!! I did not know that.

[@aldotanca9430]

Very interesting!

[@PaulThird]

🤓🤓

[@thegroove2000]

Have they actually released it yet?

[@PaulThird]

Yeah their own plugins have been out for ages

[@georgeogrady449]

Why do we still use 24 bits why not 32 bit system rack

[@GingerDrums]

I don't get it either

[@henrikpetersson3463]

In which way do you mean? 32 bits is great in some situations bit pointless (pun intended) in others. Most DAWs use 32 bits for internal processing these days afaik.

[@GingerDrums]

@@henrikpetersson3463 I understood the obvious implementation of everybody just recording in 32bit... and in one fell swoop clipping would be a thing of the past.

[@henrikpetersson3463]

@@GingerDrums 32 bit recording is awesome for uncontrollable environments. It’s a game changer for field recording for example. For studio use, not really. If clipping is a frequent problem in the studio then you’re doing something wrong. 24 bit is really enough in the studio as it keeps the noise floor low already. The downside of 32 bit recording is that the conversion is more expensive. And it’s also more resource intensive with that much data. So if you’re not in a situation where clipping is a constant worry, then there’s really no need to waste the resources. The same reason why people record at lower sample rates even though their interfaces can deliver 192khz. The potential benefits is not worth the increased resource demands for most people. This might very well change in the future, but I think it will take quite some time before 32 bit recording is the norm.

[@GingerDrums]

@@henrikpetersson3463 point is, even if one in 1000 recordings gets clipped, that's an error margin that is still happening tens of thousand times a day. Once the compute becomes cheap enough there is no reason not to implement it.

[@vigilant545]

Great idea. We need the GPU working 🤓👍

[@PaulThird]

🤓🤓

[@richertz]

Sorry but I have tried the GPU plugins out on my Mac for a bit now and they do not work - sound is very distorted and only comes out the left. If this is going to take of it has to work smoothly across platform. This was tried some years ago (back on the 2000's and failed). This might suffer the same fate. Its a big undertaking to acheive multi OS platforms, DAWS but to also cope on multiple GPU architectures? I love the concept but I think its too big a challenge. Austica are a company that changed the game with plugins and it amazes me the have done so well, so I hope the same for GPU audio.

[@PaulThird]

Hmmm wonder if it's just an apple thing 🤔

[@richertz]

@@PaulThird it possibly is. Thinking about the long term prospect it has to work on both platforms ideally.

[@uselessoldman7964]

UA did something along those lines by offloading some of the processing to its audio interfaces, so its not "revolutionary" as such but it makes sense although the advantage of buying a audio computer for music production, you didnt need a GPU costing a grand !! SO I guess the question will be what GPU do you need to take advantage of this and how much do they cost, and maybe what generation of GPU would you need ie what is the schedular programmed for? Now Paul tell us all, will my VEGA64 do or will I need to buy a 4090 or specifically in my case a 6800xt which I have seriously considered as a sensible upgrade

[@PaulThird]

They told me consumer grade gpu's will get the job done and expensive gpu's arent necissary. I'm gonna take a guess that by consumer grade they mean the gpu's in most laptops 🤷‍♂️

[@lucaslutzerler5429]

a Bless

[@aviatedviewssound4798]

I think what is holding back the AI side of things is 1st most CPU not having an Neural processing unit and 2nd developer not knowing how to use it yet, instead they are relying on the CPU for AI computation as the GPU can be x10 to 20 time more efficient than the CPU, the NPU can be x50 to x100 time depending on the technology specially if it relies on the INT4 AI model library. nowadays most AI plugin or codes rely either on 128 to 512 bit Vector based instruction set from x86 chips like SSE4.1 or AVX512. The reason apple's CPU is fast is because they've rewritten the original arm architecture and remove any instruction that is either slow, unusable aka useless compared to faster that does the calculations and it makes it possible for them to add more transistors hence also increase the instruction decoders to 8 from the previous 4 without needing to increase the frequency meaning apple CPU works in a semi parallel way. P = CV²F P = Power C = Capacitance V = Voltage F = Frequency that means the more you increase the frequency the more power it pulls also voltage also increase with frequency. x86 is so complex and the cores are so big compared to apple that due to the complexity of the x86 architecture, they have to use micro-operands that takes these complex instructions and simplifies them to simpler ones by storing them to a micro-op cache to reduce latency. That's why Intel have to rebuild their libraries. Parallel is more efficient, 12 cores clocked at 2ghz is more efficient then 6 cores clocked at 4ghz.

[@PaulThird]

My 8 core i9 is pretty shit tbh. Ive slimmed my session template down to where I'm happy sound wise and I don't max out the cpu but 15% Gulfoss master is an absolute pain in the arse haha and forget about convolution. All that needs printed 😅

[@StarcraftWarcraftovich]

M-Silicons have NPUs. There are CPUs with NPUs they are not just at the consumer-end yet. The cheapest NPU device is still GPU with Tensor Cores or any other vector-based processing available on the market. And 90% of people already paid for the device, so the question is, who is going to activate it and provide a great use of this power with software

[@aviatedviewssound4798]

@@PaulThird Don't worry windows is already on arm and Qualcomm will be releasing an arm based cpu with the same type of speed as the apple m1. Lol next year you're going to see company porting their software on windows on arm but it's gonna take less time since they already have the arm build of the software, they just need to compile a windows version of it.

[@sammythebull116]

these cpu problem is just disgusting, at the beginning my cpu handle 600 serum synthesizers after 3 months working with my pc i can load maby 80 serums its game over!! format my harddisk and the power is back for 3 months,after that the problem is back, pc have no power!! spend more than 3000 euro for my pc!!

[@PaulThird]

Why is that?

[@sammythebull116]

@@PaulThird i dont know fella, i bring the pc 9 times back to the store and still have the same problem, i good friend put a light version windows on my pc so all the apps and games and these stuff we producers dont need is not on the system he overclock my pc from 3,4 ghz to 4.0 ghz nothing helps to get my power back!!

[@robertwedewer448]

Link is no longer working

[@PaulThird]

Which one

[@fxtheproducer]

Thanks for this update Paul. It's great to see it all coming together, it can only get better.⚡️ //nice description too 😉

[@PaulThird]

🤓🤓

[@thisscottishaspie5961]

🤓🤓🤓

[@PaulThird]

Remember to check out my autism channel if you want to learn more about my life 🤓🤓

[@bastianhues]

heya paul! great video as usual, the link in the description is still the wrong one - i see you've got the right one pinned below, but just might wanna fix :))

[@PaulThird]

Think that's it fixed 😅

[@georgecain123]

U need a Nvidia 1000 series or higher GPU.

[@andrew-no2ti]

Best wishes young man with your production / mixing work outside this channel. To my mind the further you travel your route - your other audience members will agree (one hopes) it's fair to predict - we will all be the happier for you: seeing you getting on, fulfilling your aspirations, growing in knowledge, skill and confidence in your judgement, fuelling your passion. If, in balancing the joys of the family and your new work does happen to permit you to drop the occasional video, then one looks forward to hearing from you. At a guess, your deeper insight then will prove even more fascinating.

[@PaulThird]

🤜🤛

[@JohnnysaidWhat]

acustica are you listening?????

[@PaulThird]

Wouldn't hold your breath as Acustica always have to do it their own way. They'd probably rather wait another 5 years until they can say they have created the tech on their own and then will use that finally sufficate nebula into obscurity

[@JohnnysaidWhat]

Sounds about right!

[@kahyui2486]

I didn't understand any of this, completey went over my head. But thanks anyway 😆

[@PaulThird]

🤓🤓

[@higltypig]

GPU audio is a tech which we needed 15 years ago. Not any longer.

[@higltypig]

@@PaulThird maybe they could optimise their plugins a bit although if your plugin is using that much cpu then it's gotta be good eh ;) I started on an amd k62 350mhz. Sure you had to bounce a few things but now there's just no excuse.

[@Bthelick]

​​​@@TeeheeTennessy What you're saying is a little like "games shouldn't need graphics help any more for 3d tasks" You're already using external hardware help! You're audio interface is performing filters at 128x (or more) audio rate, in order to capture clean audio. Plugins are still so compromised we can't run most of them past 4 or 8 times upsampling. And most plugs at 8x kill a CPU. Let me ask you this, are you a professional? I used to run 2 UAD systems + power core + liquidmix . With a modern 8 core I don't miss that help as much these days, but I still regularly peak the CPU in sessions, especially sessions where you need the latency low, and I run the most efficient drivers on the planet by RME. cpu's haven't actually moved that far forward in raw power, only in terms branch prediction and cache. Even just last week I tried to stream live, and my brand new machine could not pay back a simple session with a few plugs at low latency whilst capturing the screen and audio without severe dropouts from CPU overloads.

[@Bthelick]

​@@higltypig not true, high Resolution audio is a ridiculous amount of calculations, and not just any, they're floating point calculations. Some modern soc computers like the pi's still don't even have floating point capability. Cpu's haven't come on that far in a decade hardware wise. You know what an "optimized" floating point calculation looks like? Rounding errors everywhere. You know what rounding errors sounds like? Jitter, noise, digital distortions. Theres no real cheats of the physics when it comes to digital audio. Especially if you want to use it in real-time. Why do you think your audio interface converters upsample at 128x? When the best plugins can only manage 8x etc.

[@akagerhard]

that's true if you chose your plugins according to it. But it's easy to overload CPU, if you're actually using the plugins that eat CPU.

[@CreativeMindsAudio]

I don’t fully disagree in certain circumstances. But i regularly hit peak CPU loads and forget streaming mixing with a DAW on the same computer. Much less a laptop. I think this could also make oversampling and/or higher sample rates (88.1/96khz) more common in indie productions.

[@magneticpitch]

it sounds wonderful, it's fuckin great, massively parallel computing... but my. guess is it will work like highways. the more you build, the more traffic you get. lame developers might get even sloppier. and the good ones will exploit the new processing to its maximum as soon as they can. software catches up to hardware pretty quick. i HOPE it goes better than that, but why would you NOT take advantage of every bit of processing power? And we'll just get used to our new ultra-low latency plugins, and just pile em on til our computers start sweating again. cheers!

[@PaulThird]

🤓🤓

[@DerekPower]

If it helps, when I first knew about GPU Audio, [I also thought] that [it] was worthy of the tag "game changer". ;) A few supplemental comments: 1 - In laptops, CPU processors are less powerful than their desktop equivalents. As a result, the laptop's GPU is more powerful than its CPU. Therefore, utilising the GPU on a laptop will not only make it a comparable tool to a desktop but also will extend the useful life of the laptop. 2 - It seems that the newer generation of Intel chips has distinct cores: Performance and Efficient. This will mean that developers will have to ensure that their programs are configured to utilize the right cores, otherwise, this will lead to performance issues. In fact, if you use Cubase, you are advised by Steinberg not to use it on a computer with the latest generation of Intel chips. This makes the GPU seem like a future-proof option for ensuring continuous, reliable and top-quality performance. So all in all, it proves that if you want something different, then ... make something different =D

[@PaulThird]

It's definitely the future

[@kilkaserco]

This would be better of GPU prices weren't through the roof due to crypto mining.

[@PaulThird]

Gpu audio advise to use consumer grade gpu's. Apparently high spec gpu's are OTT and unnecessary for use with audio

[@andivax]

Useless technology until they offer run 3rd party plugins on GPU. Which will not happen. End of story

[@PaulThird]

I dont see why when CEO said in an interview they have contracts in place with 3rd party developers but can't say who and also the SDK is in place for devs to learn to use the front end

[@andivax]

@@PaulThird time will tell. GPU Audio was born in russia 4 years ago. Now they have different people in charge. Acustica Audio fails on GPU processing (it was dedicated cuda version but they drop it). As well as Liquidsonics. Will see! Great review as always Paul.

[@PaulThird]

🤓🤓

[@magneticpitch]

LOL - glad you got paid, too!! these have been in development a long time and still feel a bit too "beta" and unreliable to me. i'm interested in the reverb for sure, but the last time i looked at their site, it just seemed like they weren't quite ready for prime-time yet. as in: expect bugs and some issues.. i might wait for the Beta Beta release lol

[@PaulThird]

That's why I'm only interested in the tech as of right now

[@georgeogrady449]

Ai faster then that

[@meru_lpz]

Guess what part of the hardware AI gets the most benefit from...

[@stupidusername38]

I'll be interested to see how this performs on a typical DAW with a modest GPU. When speccing a new PC build I always go for the most under powered GPU because I don't want one with fans so as to keep noise reduced and they are a lot cheaper than a high spec GPU

[@PaulThird]

Interestingly enough gpu audio advise a consumer grade gpu, no need for expensive gamer type gpu's