Please recommend a CPU for low-latency .vst performance!

[MIDIBoss]

I am a musician and one of the things I do is play the EWI which is a windsynth breath-sensitive MIDI controller. I used to mainly use analog synths as soundsources, but slowly I have gravitated more towards software to the point that I'm basically always in software mode now.

My computer is a 2010 Macbook Pro 2.6ghz dual core machine. It does ok at 128 samples i/o buffer with most simple .vst tasks, but I have to lower the performance level of some of the synths I use and I can't layer very many synths at once to create multi-patches without getting the infamous nasty crackles & pops as the CPU in this humble but faithful machine tries to keep up with what I'm asking it to do.

I use a lot of Native Instruments Reaktor 6, FM8, Kontakt, a lot of Dune 2 (a real CPU hog), Dexed, TAL U-NO-LX, and sometimes Omnisphere.

Reaktor 6 is especially tough on CPU (it is a single-core .vst), and Dune 2 also has to be watered down in terms of polyphony, unison voice counts, & modulation resolution in order to get it to work on my machine.

My DAW is Usine Hollyhock 3 which makes great usage of multicore processors. Reaktor 6, on the other hand, needs very fast single core speed. My computer doesn't really have either of those characteristics.

I want to build a dedicated music supercomputer with a CPU that can really slay all my .vst problems, but I'm not certain of the best one to select. I want the computer to be able to run at lower latency settings like 64, 32, or maybe even 16 samples i/o buffer speed, and I want to be able to layer more synths together to create multi-patches without watering down the settings so much.

Right now, my idea is to build around the new Intel 8700k that's coming out, delid, and overclock it as fast as possible. It will have 6 cores and 12 threads and if I can keep it cool enough, all 6 cores will clock in at over 4.0ghz according to the rumored specs.

I am also aware that there is Ryzen 1800X which has 8 cores and is darned fast, too.

What do you recommend?

This is an example of some of the kind of stuff I'm trying to do on the computer:


That chords patch is watered down. The original version sounds much fatter than that, but I had to reduce the thickness of the tone via unison voice count reduction to get it to play with my laptop.

Looking for some better informed advice than my limited knowledge can reach!

Thanks.

 

[Sycraft]

For the lowest latency VSTs you want a modern Intel chip on a good motherboard, running Windows and an RME interface. You can push it down to 48 samples in some cases if you do that. At that point, the latency of the sound moving through the air from your speakers is often as much as your interface. The Ryzen chips, while powerful, aren't as good at low latency as the Intel chips presently.

I'd recommend looking at a Studiocat DAW. Jim builds really good systems and knows how to optimize them for real-time audio performance. When you start talking the super low latency stuff, it isn't just the CPU selection that matters, some motherboards are better than others and the settings in the BIOS matter. He's done a lot of testing and can sell you something that works well.

You also can't skimp on the audio interface, it is a huge part of the equation and yes, that does mean it'll be expensive. A cheap Focusrite just isn't going to compete with a high dollar RME when it comes to low latency work. Not only are good drivers/interface hardware a must, but the D/A converters themselves matter as their internal filters have latency and some are lower than others.

 

[B00nie]

Choose Windows if you want your unsaved music project destroyed by a windows update - or Microsoft chimes ruin your flow.

 

[MIDIBoss]

Choose Windows if you want your unsaved music project destroyed by a windows update - or Microsoft chimes ruin your flow.

Tell me more. What do you mean? Are Windows updates known for corrupting audio production files? I do make a habit of backing up my system. But I am coming from Mac where a lot of basic organizational tasks are sort of automated.

Microsoft chimes? You mean like OS system alerts? Mac has these too but you can turn them off. Is it not possible to simply mute OS system alerts on Windows??

 

[MIDIBoss]

For the lowest latency VSTs you want a modern Intel chip on a good motherboard, running Windows and an RME interface. You can push it down to 48 samples in some cases if you do that. At that point, the latency of the sound moving through the air from your speakers is often as much as your interface. The Ryzen chips, while powerful, aren't as good at low latency as the Intel chips presently.

I'd recommend looking at a Studiocat DAW. Jim builds really good systems and knows how to optimize them for real-time audio performance. When you start talking the super low latency stuff, it isn't just the CPU selection that matters, some motherboards are better than others and the settings in the BIOS matter. He's done a lot of testing and can sell you something that works well.

You also can't skimp on the audio interface, it is a huge part of the equation and yes, that does mean it'll be expensive. A cheap Focusrite just isn't going to compete with a high dollar RME when it comes to low latency work. Not only are good drivers/interface hardware a must, but the D/A converters themselves matter as their internal filters have latency and some are lower than others.

Thanks so much for your time and the great info. I checked out Studiocat's website. What's really cool is the fact that he lists out the parts he uses in the builds on his site. The build I have my eye on lists some of the same exact main pieces that Studiocat is using in one of his custom builds. So it looks like I'm on the right track anyway.

My interfaces are superb already. In addition to playing music I'm mainly a mixing guy so I had a Universal Audio Apollo Gen 1 FW 800 already and my live DAC for performance is a Dangerous Source which is USB 2.0 and seems to be very fast. I've never measured the latency but at low buffer settings it's totally playable.

The main bottleneck in my system seems to be CPU. I get CPU overload messages on the norm, even when running one instance of a really complex Reaktor 6 patch at 128 samples, for example. So I think it'll be a job for the Intel Coffee Lake 8700k to slice through when it (and an accepting motherboard) become available later this year.

 

[Pandur]

Tell me more. What do you mean? Are Windows updates known for corrupting audio production files? I do make a habit of backing up my system. But I am coming from Mac where a lot of basic organizational tasks are sort of automated.

Microsoft chimes? You mean like OS action alerts? Mac has these too but you can turn them off. Is it not possible to simply mute OS action alerts on Windows??

He just wanted to let you know that he neither knows how to configure Windows updates properly or turn off the OS sounds. Nothing else to see here.

 

[MIDIBoss]

He just wanted to let you know that he neither knows how to configure Windows updates properly or turn off the OS sounds. Nothing else to see here.

That's what it seems like to me to!

 

[Sycraft]

Thanks so much for your time and the great info. I checked out Studiocat's website. What's really cool is the fact that he lists out the parts he uses in the builds on his site. The build I have my eye on lists some of the same exact main pieces that Studiocat is using in one of his custom builds. So it looks like I'm on the right track anyway.

My interfaces are superb already. In addition to playing music I'm mainly a mixing guy so I had a Universal Audio Apollo Gen 1 FW 800 already and my live DAC for performance is a Dangerous Source which is USB 2.0 and seems to be very fast. I've never measured the latency but at low buffer settings it's totally playable.

The main bottleneck in my system seems to be CPU. I get CPU overload messages on the norm, even when running one instance of a really complex Reaktor 6 patch at 128 samples, for example. So I think it'll be a job for the Intel Coffee Lake 8700k to slice through when it (and an accepting motherboard) become available later this year.

Well just remember with low latency stuff a CPU overload doesn't always mean that it is max out instruction wise, it can mean it can't switch tasks fast enough to service all the processes needed in the very short window. Not saying you don't want a good CPU, you do, just that the audio interface is hugely important since every time data is input or output it has to generate an interrupt, switch over to the audio drivers, handle the audio, and switch back. It is critical that the hardware and software involved in that are very fast if you want to go for super low buffers like 64 samples.

You also want as fast a processor, GHz wise, as possible including even possibly overclocking (though careful with that, it can cause stability issues with audio shit) because ultra low latency demands fast execution of a chain of instructions, just doing more in parallel doesn't cut it. Fast RAM is also useful for low latency. It doesn't matter at all for like gaming, but for ultra low latency, it helps.

Then like I said, have to configure the system right. Shut down all the processor throttling and low power states. Have it run at max speed all the time. That does use more power and generate more heat, but there is a delay when a core jumps between speeds and you don't want that. Disable VT-x and VT-d, the virtualization stuff is great, but again, more latency. Shut down any integrated device on the motherboard you aren't using like serial, wifi, onboard audio, etc. Tell your GPU to run at max power all the time, again switching power states takes time. Consider disabling ethernet. A pain I know but man, ethernet cards cause big DPC latency spikes, the stupid things are just not that efficient latency wise.

Just be aware the lower you wanna push the buffer, the harder it is to get stable audio and the more you'll need to optimize the system. If you are willing to run a 1024 buffer, heck you don't need to worry about anything. Use whatever you like, leave the system in a default power efficient state, it'll all work great. If you want 32 samples you'll have to have a properly built system, that has been well optimized, using a really good sound card.

 

[B00nie]

He just wanted to let you know that he neither knows how to configure Windows updates properly or turn off the OS sounds. Nothing else to see here.

Bullshit. If you use consumer versions of the OS Microsoft walks over your settings as it wishes. Besides, nobody should have to do special tricks to perform a simple task of configuring when your updates are going to boot your computer!

And while the chimes can be muted (it was a metaphor, duh!), it's extremely uninspiring to work on Windows. It's just so revolting in every way. He should stay on mac (or he'll regret it). You'll have to spend more time fighting against Windows 'features' than doing productive work.

Not to mention the stress of being attacked constantly, you basically can't do anything with the computer without the antivirus slowing you down or having to worry which site you opened or file you downloaded will infect you next (and possibly encrypt your hard drive). I would imagine that a producer will have to use a lot of outside sources for sound files and any of them potentially carries a Windows attack payload.

Windows is just stressful and bad for work use.

 

[B00nie]

Nice example: Today I booted up my Win10 box after many months to test a quick thing on it. Naturally I couldn't use the computer at all due to windows updates taking forever. That's not all, after the update the sound which worked prior to reboot now no longer works. My daughters Win10 got also an update today. After that the realtek gigabit chipset no longer finds an IP. After several attempts on reinstalling the driver and using windows restore to no avail I formatted C: and installed Win10 from scratch. Now it works again. Probably until the next update.

If this is the experience you want with your production computer, go ahead and set up windows 10.

Oh I almost forgot: after the reinstall there was yet another glitch. This time with windows activation. 'There was an error'. Troubleshooter showed that my product ID was in use not just on my computer but on some computer in Germany. I was lucky to be able to activate the machine at all.

 

[Pandur]

Bullshit. If you use consumer versions of the OS Microsoft walks over your settings as it wishes. Besides, nobody should have to do special tricks to perform a simple task of configuring when your updates are going to boot your computer!

Here's how I perform this very tough and special trick so that everybody can try and see if they can manage it :

- Open Settings by clicking on the cog in the start menu or by right-clicking the start button and clicking Settings from the list that opens up.
- Click the Update & security icon to open the Windows Update settings.
- Click on change active hours and set it to any range up to 18 hours long. Mine is set to 8 AM to 2 AM. Windows will never reboot your computer automatically within this time period.

When it comes to the rest of your OS-phobic drivel I'm not going to touch that with a 10-foot pole!

Not saying Windows is anywhere near perfect, but this is not something that requires a computer science background to take control over.

 

[HammerSandwich]

Unless you're still on 1607, which remains the case for some people with no special settings.

 

[B00nie]

Here's how I perform this very tough and special trick so that everybody can try and see if they can manage it :

- Open Settings by clicking on the cog in the start menu or by right-clicking the start button and clicking Settings from the list that opens up.
- Click the Update & security icon to open the Windows Update settings.
- Click on change active hours and set it to any range up to 18 hours long. Mine is set to 8 AM to 2 AM. Windows will never reboot your computer automatically within this time period.

When it comes to the rest of your OS-phobic drivel I'm not going to touch that with a 10-foot pole!

Not saying Windows is anywhere near perfect, but this is not something that requires a computer science background to take control over.

How gracious of Microsoft to let you actually define a time period where the OS shouldn't reboot against your will. Then you're just going to wait when the nasty surprise is going to come in forms of windows update downloading updates at times you wouldn't want to use your bandwith to that - or updates forcing themselves in when you're in a hurry and want to boot/reboot your computer.

Both OSX and linux handle updates a million times better. They inform you that an update is available, clearly explain what the update is going to do and let you choose if and when to activate it. The OP will regret the move if he's used to OSX lol. He may not know it until it's too late, though.

 

[IdiotInCharge]

Don't mind the Linux trolls. They get testy when they run out of stuff to act superior about in their dedicated subforum.

So I think it'll be a job for the Intel Coffee Lake 8700k to slice through when it (and an accepting motherboard) become available later this year.

This is probably the best bet. The 8700k will likely be the fastest consumer CPU available for nearly every task upon release, and especially so for latency-sensitive applications. Highest clockspeed, highest IPC, and largest cache, combined, with six cores available.

 

[B00nie]

Heh Midiboss is about to make the mistake of his life. Wouldn't want to be him...

 

[MIDIBoss]

Thank you all for your input! I can see both sides of the coin, there are valid points each way. I have had my fair share of major problems with Apple products over the years, too. The cool thing about Apple is that if it is a manufacturing defect that can be remedied, they will usually fix it for you for free.

Anyway, I think what I am leaning towards doing is building a PC and dual-booting with Windows on one drive and Hackintosh on another. I have done the research and picked out a compatible system. There are some music .vst's that are only available on PC I want to gain access to including an 8-band parallel delay unit with separate panning & modulation available on each band which is an all-in-one effect emulating the hallowed 1980's Yamaha UD Stomp. The only way I've been able to emulate that effect accurately on Mac OSX is to use 8 instances of a modulated delay plugin and pan them manually. Lots more time involved, clutters up your workspace in a DAW, and in some cases it sucks up the max amount of AUX sends available on a channel for example in Logic X where you only are allowed 8 AUX sends per channel. This is just one example of things I'm looking to have other options for, there are more...

However, I am well aware of the advantages of Core Audio over ASIO. I often do want to run two interfaces at once for various reasons (one for input, one for output), and although there are digital multi-interface conections that allow you to do the same thing externally, it is nice to have the option to do it internally. ASIO doesn't allow that AFAIK.

My Mac is also incredbly stable. I rarely have a freeze or crash.

So I'm going to build a Macbook-slaying PC based around the upcoming 6-core 8700k possibly delidded and modestly overclocked, 32gb of RAM, and a hi-end GPU for gaming (why not?) all inside the Dan V2 case which will fit into a backpack and still spend only about half what a new top of the line Macbook Pro would cost. It will dual-boot on Windows and Hackintosh with two separate 512gb Samsung 960 Pro SSD Nvme drives and the two OS's will share a partitioned 1 or 2TB 7,200rpm HDD for extra storage space. I will probably try to keep the Windows side off of the internet, downloading any needed files for the Windows side to the shared drive from the Mac side and running them on the Windows side.

The latest build of Sierra allows the OS to recognize non-Apple Nvme SSD's so where in the past when you had to run a patch to get OSX to recognize a 3rd-party SSD Nvme, now they will run natively.

To say I am excited is an understatement. It is a great time to be alive for general computer technology and especially electronic-music making.

Cheers.

 

[Sycraft]

Just be aware with regards to stability you may not see that from a Hackintosh. Apple makes stable systems because they control hardware, software, and drivers. They have a very limited set to work with and support and can make sure it works well. With a Hackintosh, you've changed that so you may have issues. There's no magic sauce in OS-X that makes it stable, it is just another modern OS like any other. You can get a similar experience from Windows if you buy something like an enterprise (Optiplex or Precision) system from Dell. They control what hardware they choose, certify all the drivers, and the systems are rock solid. As a simple example we have two Hyper-V servers that run most of our infrastructure and they've run 24x7x365 for 5 years now with not a single crash and that is the level of reliability I'd expect, even one crash would be a huge failure in my eyes. Problem is, of course, you can't customize the system for super-low DPC latency like you can a custom build, and you pay more for less.

However you don't get that level of stability with something you put together yourself necessarily. You can, don't get me wrong, but it isn't guaranteed. Depending on the hardware choices you make, the tuning you do, the drivers available, and so on your system may not be as stable as you are used to. That'll be the case for any OS. Not trying to talk you out of it, I think your idea is solid, just making sure you have appropriate expectations. When you build a custom system and hotrod it for performance, it may not be as stable as you are used to, no matter what OS you stick on it.

So all that said you also will likely find you get better low latency performance in Windows, if that is your goal. I don't know of an article with more recent tests than this one but it likely is still this way and is if anything more skewed towards Windows as Microsoft has made some notable kernel improvements for audio since then. YMMV of course, but the best data I can find, as well as anecdotes from people on the Cubase forums, is that you get more VSTs at lower latency in Windows. Now all this depends on how hard you push things if it even matters. You may well discover that at whatever latency you choose you are totally fine in any OS, any DAW, for any work you do. I'm just harping on this since you've expressed interest in doing super low latency (64 samples or less) stuff and that takes some real tuning to make it work well in most cases.

With regards to multiple interfaces ya, ASIO doesn't allow that and it is on purpose at they can potentially conflict in terms of getting CPU time. Not a big deal and unlikely to be an issue, but that is a big reason why it is limited to one at a time. In Windows if you want multiple interfaces you can do it using native WASAPI 2 or WDM/KS unfortunately few DAWs support those APIs and few pro cards have good drivers for them. Again RME is the one that does a good job, their WDM driver being in every way as good and functional as their ASIO driver. The only DAW I know that does a good job supporting it is Sonar. With Windows 10 Sonar can get 2ms of latency using WASAPI mode, which translates to around 96 samples, on a stock consumer integrated soundcard (Realtek chip).

At any rate, something you can look at if you want to do multi-card stuff, though likely not a realistic option as you'd probably have to change DAW. Just more stuff to be aware of.

 

[cageymaru]

Can't he build a Hackintosh with an Intel chipset? That way he has really awesome hardware and OSX still? Never done it myself, but I've seen videos and random guides on the net.

 

[Sycraft]

He can, it'll work, it just won't necessarily have the stability of an actual Mac. Highly stable systems are where the hardware and software is highly controlled. When you do your own thing, stability is much more of a crap shoot, particularly if you do things like tune the UEFI for low latency performance. That's all I meant, just wanted to make sure he didn't think he'd get a rock solid system making a Hackintosh, or that he couldn't have a rock solid Windows system if he wanted to go the Precision route.

Just trying to make sure his expectations were set accordingly. What he's doing will totally work.

 

[MIDIBoss]

He can, it'll work, it just won't necessarily have the stability of an actual Mac. Highly stable systems are where the hardware and software is highly controlled. When you do your own thing, stability is much more of a crap shoot, particularly if you do things like tune the UEFI for low latency performance. That's all I meant, just wanted to make sure he didn't think he'd get a rock solid system making a Hackintosh, or that he couldn't have a rock solid Windows system if he wanted to go the Precision route.

Just trying to make sure his expectations were set accordingly. What he's doing will totally work.

I appreciate your insight and thanks for sharing all that information, it's very helpful. In a way I don't mind the challenge of trying to get it all to work over a period of time. I am a tinkerer by nature. I have had to do a lot of software experiementation on my current Macbook Pro setup to figure out which software conflicts with other software and trying both .vst and .au versions of every plugin I use (which adds up to a lot of trial and error) even just to get my current system to run smoothly. I still have minor issues with software compatibility. Usually I can work around them or experiment with different combinations of plugins in different DAWs until I find something that works.

Computers are never as trustworthy for live music performance as plugging a midi controller into a hardware synth (or just playing an acoustic instrument, lol), but the flexibility and power of what you can do in the computer makes it worth the frustrations of sticking with it, to me.

I will definitely install Windows on the new customized machine and might even try Linux, which my DAW is reported to work with. I want to have at least two options for OS available at all times in case one isn't working after an update or something, so that maybe the other one will be working at any given time. I don't make a habit of updating my OS unless it has an added feature I really want and there are no reports of issues with my software. Sierra, for example, destroyed audio production software for several months after it was released until developers caught up.

My DAW, Usine Hollyhock 3, is cross-platform. The Mac version is 32 bit only. So I'm already using clunky and semi-stable 64 bit plugin to 32 bit intermediary bridge software just to get things running fairly smoothly on my Mac. As it is now, some of the GUIs are not fully functional. In the end though, it works and doesn't usually crash, so as long as I don't mess around with things too much after they're working. I can reliably get through a gig with this setup. Usine HH3 is a lot more stable than Reaper, my old DAW, and even Logic X crashes more than Usine HH3.

The computer, at this point, rules for synth stuff! Dune 2, Reaktor, Dexed... I don't feel that I'm missing anything with softsynths like these compared to hardware at this point, and the flexibility is incomparable especially in the fully customizable and modular HH3, you can make the host do pretty much anything you can dream up.

The only thing I'm missing is the ability to run really complex patches at low latency. I can actually run my simple solo sounds from Dexed and Dune 2 on my 2010 Macbook at like 32 i/o buffer, but any complex patches throw the CPU under the bus!

 

[daglesj]

"My computer is a 2010 Macbook Pro 2.6ghz dual core machine. It does ok at 128 samples i/o buffer with most simple .vst tasks, but I have to lower the performance level of some of the synths I use and I can't layer very many synths at once to create multi-patches without getting the infamous nasty crackles & pops as the CPU in this humble but faithful machine tries to keep up with what I'm asking it to do."

Erm how much more powerful do you really need to go if a 2.6GHz dual core will get you 'kinda there'?

Are you creating tunes or looking for the God Particle?

Just trying to bring this topic back to some kind of reality here.

 

[MIDIBoss]

"My computer is a 2010 Macbook Pro 2.6ghz dual core machine. It does ok at 128 samples i/o buffer with most simple .vst tasks, but I have to lower the performance level of some of the synths I use and I can't layer very many synths at once to create multi-patches without getting the infamous nasty crackles & pops as the CPU in this humble but faithful machine tries to keep up with what I'm asking it to do."

Erm how much more powerful do you really need to go if a 2.6GHz dual core will get you 'kinda there'?

Are you creating tunes or looking for the God Particle?

Just trying to bring this topic back to some kind of reality here.

I get what you're saying. Obviously my main goal is always creating music!

Well, some of my patches are totally stable at low latency on my current system but then others completely obliterate my CPU even at much higher buffers like 1024 or 2048 and they still don't work even there! I have to thin out the more CPU intensive patches (less unison voices per patch in Dune 2, no more than 2 instances of Reaktor at a time, etc.) in order to get them to play.

I figure if I'm going to buy a new machine, I should just try to put together a system that will have no problem doing basically anything I'd typically dream up for practical performance purposes.

Cost factor is also an issue. I had some help when I bought the Macbook. It's 7 years old and looks brand new. No plans to get rid of it, it's a fine machine. However, buying a new one with the specs I'd want would cost over 3 grand!!! The best specs on a MBP also pale in comparison to what you can do with a custom mini itx build.

I have thought about it a lot and it just seems like the right move to build an insanely powerful machine for around ~$1500 give or take a little which will be much faster than what I have already. I can scratch together that amount of cash for the build. $3k is another story altogether, though. Whew!

 

[MIDIBoss]

At the tail end of all this I would like to ask: Are custom built PC's known to be unstable? Like, regarding Sycraft's comments, I was just wondering if there is a general vibe of instability among custom PC builds. Obviously since I'm gigging with my setup it needs to be stable! Any thoughts on that? Planning to try Windows first and possibly Mac/Linux afterwards.

 

[Sycraft]

No they aren't known to be, they just can be. The reason is that you are choosing the hardware, you are doing the install, you are doing the support. So you can make mistakes or have support issues that you wouldn't with an enterprise system. With something like Dell's business line, they've carefully tested all the hardware they sell to make sure it plays nice, and they've vetted the drivers and so on. Thus there's a real good chance everything is rock solid. You haven't done that with a custom build, so there is more possibility for issues. Particularly if you start OCing and stuff like that. You play around with all kind of UEFI settings, you can cause some issues

They can be very stable though. My custom build at home hasn't crash since... well I can't remember when, probably since I was tuning the OCing 2.5ish years ago. It runs solid and fast and doesn't give me any issues. However, I've seen other people who aren't so lucky. Also I'm willing to throw money at it to make it that way. If I have a component causing instability, out it goes and a new one comes in to replace it. I'll spend the time, effort, and money to make sure my custom build is solid.

That was why I pointed to Studiocat since you are basically paying someone to do that for you. It is a custom build, but Jim tests it and makes sure it all works well.

I don't want to give the wrong impression that custom builds are bad, I always have a custom build as my desktop at home. Perhaps I didn't phrase things well and made it look more problematic than I actually meant. I was mostly just cautioning you since you are a Mac user. You are coming from the most controlled, locked-down, environment and going to the most open and free-form. I have seen Mac users who've done that get all pissed off at Windows as being "unstable" because they have a custom build with issues and blame it on the OS, rather than the change in environment.

Basically if you do a custom build you just need to shake it down and test it well before you use it for anything important.

 

[MIDIBoss]

No they aren't known to be, they just can be. The reason is that you are choosing the hardware, you are doing the install, you are doing the support. So you can make mistakes or have support issues that you wouldn't with an enterprise system. With something like Dell's business line, they've carefully tested all the hardware they sell to make sure it plays nice, and they've vetted the drivers and so on. Thus there's a real good chance everything is rock solid. You haven't done that with a custom build, so there is more possibility for issues. Particularly if you start OCing and stuff like that. You play around with all kind of UEFI settings, you can cause some issues

They can be very stable though. My custom build at home hasn't crash since... well I can't remember when, probably since I was tuning the OCing 2.5ish years ago. It runs solid and fast and doesn't give me any issues. However, I've seen other people who aren't so lucky. Also I'm willing to throw money at it to make it that way. If I have a component causing instability, out it goes and a new one comes in to replace it. I'll spend the time, effort, and money to make sure my custom build is solid.

That was why I pointed to Studiocat since you are basically paying someone to do that for you. It is a custom build, but Jim tests it and makes sure it all works well.

I don't want to give the wrong impression that custom builds are bad, I always have a custom build as my desktop at home. Perhaps I didn't phrase things well and made it look more problematic than I actually meant. I was mostly just cautioning you since you are a Mac user. You are coming from the most controlled, locked-down, environment and going to the most open and free-form. I have seen Mac users who've done that get all pissed off at Windows as being "unstable" because they have a custom build with issues and blame it on the OS, rather than the change in environment.

Basically if you do a custom build you just need to shake it down and test it well before you use it for anything important.

Hi Sycraft, oh ok cool! Thank you so much for taking the time, this is exactly the kind of helpful info I was hoping to get here.

All that makes sense to me, and I expect to have to do some trial & error + tweaking to get things running smoothly. May not even have to overclock, in the end. I'll have to get it all running and test it first to find out.

Personally I've always been the problem solving type and actually enjoy a good challenge like this. I've got the old system to use until the new one is up and running smoothly anyway.

Studiocat looks great and thanks again for the recommendation. His prices come out pretty steep compared to putting together the system yourself and he lists almost all the parts in the builds, many of which are the parts I had in mind already. The only thing he doesn't list is the motherboard but I'm sure a little research on my part can help figure out which one would be a good choice.

It's strange to me that most of the mini itx boards don't have thunderbolt. My Apollo interface is Firewire. I don't think the case I have in mind for this build could afford the space of a Firewire PCIe card. The Apollo can have a Thunderbolt conversion card installed in it, but that's pointless if the mobo doesn't have a compatible port...

Anyway, more research to be done! Cheers and thanks for the help.

 

[daglesj]

I'd say look at picking up a Dell dual or single Xeon CPU workstation on Ebay.

 

[SolidSnake3035]

Then like I said, have to configure the system right.

Do you happen to have links to resources that explain these configurations and optimizations? I'd love to have something that I can keep and reference for audio and general latency stuff, as it's something I've had issues with in the past, and you certainly seem to know your stuff.

Thanks!

 

[MIDIBoss]

Do you happen to have links to resources that explain these configurations and optimizations? I'd love to have something that I can keep and reference for audio and general latency stuff, as it's something I've had issues with in the past, and you certainly seem to know your stuff.

Thanks!

Well, here is a great resource about that: https://support.focusrite.com/hc/en-gb/articles/207355205-Optimising-your-PC-for-Audio-on-Windows-10

This might contain some repeated information but it's another link with similar ideas: https://us.focusrite.com/optimising-a-pc-for-low-latency-audio

On my Mac, the other side of the pond, I have found that things like shutting off GPU switching, shutting off Wifi/Bluetooth, and performing routine cache/preference cleaning with an app like Onyx help Mac OSX run smoother for low latency audio stuff.

 

[Sycraft]

Do you happen to have links to resources that explain these configurations and optimizations? I'd love to have something that I can keep and reference for audio and general latency stuff, as it's something I've had issues with in the past, and you certainly seem to know your stuff.

Thanks!

No, I don't have any resources that I know of. All the ones I've found are in my opinion deficient and I've not taken the time to write my own. Basically what it comes down to is shutting off anything you don't need, both hardware and software, as it all competes for system time, and shutting down any power management stuff, since that takes time to change state. It is also something that isn't a hard and fast kind of situation as depending on your needs and your system, you can reach a "good enough" state. Like if 256 samples of latency is your goal, and you find you can reach that without any optimization then ok, cool. No reason to go hard on this and limit functionality if you have what you want out of the box. This is the kind of stuff you do when you are trying to hit a low latency and it isn't working smoothly.

Basically the steps I would take are, in this order (these apply to desktops in particular):

0) Get a good soundcard. Seriously, can overemphasize this enough. The soundcard drivers and their interface implementation (be it USB, FW or PCIe) make all the difference in the world. RME is amazing in this arena, getting latencies via USB that others have trouble with via PCIe. However even with them, if low latency is the order of the day, PCIe is the way to go. It just has less latency than USB/FW.

1) Shut off all unneeded devices in BIOS. Things like onboard sound, parallel port, extra SATA controllers, WiFi, etc. Anything you aren't using, turn off. Since you aren't using it you won't miss it .

2) Shut off low power states. Tell your processor to run at full speed all the time. No no c-states, no speed step in BIOS. Likewise tell your GPU to run at full power all the time (nVidia GPUs have an option in their control panel for that for example), and then tell your OS/other components to do that by setting the high power profile. Your system will guzzle more power and generate more heat but it will help. This alone may be enough to do the trick.

3) Disable VT-x and VT-d (sometimes called VMX for whatever reason) in the BIOS. This shuts off the ability to do virtual machines using Hyper-V, makes VMWare slow as dogshit, and can slightly decrease security but it helps with latency.

4) Disable any background software you can. Things like Steam or what not. Try to have the system running as few things as possible.

5) Minimize connected USB devices. Have yourself a keyboard, mouse, and that's it. Unless you have something you need, get rid of it. The less that your system needs to talk to, the better.

6) Disable networking. This is a pain I know, but NICs are bastards with DPC latency. You can reenable it when you aren't using it. Also in line with that, use wired not wireless. Wireless is way worse for problems than wired.


Past that it starts to get system specific and a lot of trial and error on things. In general if you've gone this far and can't get the latencies you want, you may be asking for something that isn't possible with the hardware you have.

Two other things of note:

--I find that the "set Windows to optimize background processes" advice is wrong, but you can try it (it is easy to change). What foreground and background mean is "give more CPU time to the program that has focus" and "give more equal time to all programs". Given that you are focused on your DAW and nothing else, foreground generally will give more time where you want it. Your synths aren't normally "running in the background" they are loaded as threads in your DAW and thus are part of the application in focus. Lots of people seem to recommend this though, so maybe I'm wrong, but my testing does not indicate it, nor does my understanding of how the Windows scheduler hands out quanta. Perhaps different DAWs work different with that (I use Sonar personally).

--Don't get too obsessive with low latency. It can become an obsession to its own end and gain you nothing. Keep it in perspective. For example: If you are running 48KHz audio then 48 samples = 1ms, 96 samples = 2ms and so on. So think of it in those terms. Also consider that sound travels about 1 foot per ms in the air. So if your speakers are 6 feet away from you, that's 6ms of latency. Thus if you are fretting about taking your audio interface from 128 samples (2.7ms at 48KHz) down to 64 samples (1.3ms at 48KHz) ask yourself if that really matters in light of the delay from your speakers to your ear, or if just moving the speakers closer to you or wearing headphones might be easier. 256 samples of latency with headphones is less effective latency than 32 samples with speakers 6 feet away. Also remember that real instruments have latency as well. If you play an electric guitar you probably have 10ms or so of latency between a pluck and you hearing it because of where you place the amp.

 

[MIDIBoss]

No, I don't have any resources that I know of. All the ones I've found are in my opinion deficient and I've not taken the time to write my own. Basically what it comes down to is shutting off anything you don't need, both hardware and software, as it all competes for system time, and shutting down any power management stuff, since that takes time to change state. It is also something that isn't a hard and fast kind of situation as depending on your needs and your system, you can reach a "good enough" state. Like if 256 samples of latency is your goal, and you find you can reach that without any optimization then ok, cool. No reason to go hard on this and limit functionality if you have what you want out of the box. This is the kind of stuff you do when you are trying to hit a low latency and it isn't working smoothly.

Basically the steps I would take are, in this order (these apply to desktops in particular):

0) Get a good soundcard. Seriously, can overemphasize this enough. The soundcard drivers and their interface implementation (be it USB, FW or PCIe) make all the difference in the world. RME is amazing in this arena, getting latencies via USB that others have trouble with via PCIe. However even with them, if low latency is the order of the day, PCIe is the way to go. It just has less latency than USB/FW.

1) Shut off all unneeded devices in BIOS. Things like onboard sound, parallel port, extra SATA controllers, WiFi, etc. Anything you aren't using, turn off. Since you aren't using it you won't miss it .

2) Shut off low power states. Tell your processor to run at full speed all the time. No no c-states, no speed step in BIOS. Likewise tell your GPU to run at full power all the time (nVidia GPUs have an option in their control panel for that for example), and then tell your OS/other components to do that by setting the high power profile. Your system will guzzle more power and generate more heat but it will help. This alone may be enough to do the trick.

3) Disable VT-x and VT-d (sometimes called VMX for whatever reason) in the BIOS. This shuts off the ability to do virtual machines using Hyper-V, makes VMWare slow as dogshit, and can slightly decrease security but it helps with latency.

4) Disable any background software you can. Things like Steam or what not. Try to have the system running as few things as possible.

5) Minimize connected USB devices. Have yourself a keyboard, mouse, and that's it. Unless you have something you need, get rid of it. The less that your system needs to talk to, the better.

6) Disable networking. This is a pain I know, but NICs are bastards with DPC latency. You can reenable it when you aren't using it. Also in line with that, use wired not wireless. Wireless is way worse for problems than wired.


Past that it starts to get system specific and a lot of trial and error on things. In general if you've gone this far and can't get the latencies you want, you may be asking for something that isn't possible with the hardware you have.

Two other things of note:

--I find that the "set Windows to optimize background processes" advice is wrong, but you can try it (it is easy to change). What foreground and background mean is "give more CPU time to the program that has focus" and "give more equal time to all programs". Given that you are focused on your DAW and nothing else, foreground generally will give more time where you want it. Your synths aren't normally "running in the background" they are loaded as threads in your DAW and thus are part of the application in focus. Lots of people seem to recommend this though, so maybe I'm wrong, but my testing does not indicate it, nor does my understanding of how the Windows scheduler hands out quanta. Perhaps different DAWs work different with that (I use Sonar personally).

--Don't get too obsessive with low latency. It can become an obsession to its own end and gain you nothing. Keep it in perspective. For example: If you are running 48KHz audio then 48 samples = 1ms, 96 samples = 2ms and so on. So think of it in those terms. Also consider that sound travels about 1 foot per ms in the air. So if your speakers are 6 feet away from you, that's 6ms of latency. Thus if you are fretting about taking your audio interface from 128 samples (2.7ms at 48KHz) down to 64 samples (1.3ms at 48KHz) ask yourself if that really matters in light of the delay from your speakers to your ear, or if just moving the speakers closer to you or wearing headphones might be easier. 256 samples of latency with headphones is less effective latency than 32 samples with speakers 6 feet away. Also remember that real instruments have latency as well. If you play an electric guitar you probably have 10ms or so of latency between a pluck and you hearing it because of where you place the amp.

Man, you really know your stuff. Thank you for sharing this knowledge. I'm sure we can all reference this as a starting point if any of us are aiming for low latency setups.

Cheers.

 

[daglesj]

I see those optimization lists for audio and it takes me back to the 1990's where a good portion of those optimizations or the concept behind them came from. Most of them were necessary back then due to single core CPUs, low ram and often HDD systems that shared the same IDE cable. One little blip of something else ruined your work.

It's like when audiophiles say you should burn CDs at 1x and disable a load of stuff. Again old outdated folklore and totally ignorant of the fact that CD/DVD dye formulations have changed so much that you should actually burn much faster. Not to mention they forget a lot of them were due to not having 'burn-proof tech' that came in at the turn of the century.

I bet on a modern $3000 machine the likes of what is talked about on this thread, most of those optimisations might improve things by a negligible % at best. The fact the machine will also be highly hamstrung for anything else useful is neither here nor there.

 

[Sycraft]

I see those optimization lists for audio and it takes me back to the 1990's where a good portion of those optimizations or the concept behind them came from. Most of them were necessary back then due to single core CPUs, low ram and often HDD systems that shared the same IDE cable. One little blip of something else ruined your work.

It's like when audiophiles say you should burn CDs at 1x and disable a load of stuff. Again old outdated folklore and totally ignorant of the fact that CD/DVD dye formulations have changed so much that you should actually burn much faster. Not to mention they forget a lot of them were due to not having 'burn-proof tech' that came in at the turn of the century.

I bet on a modern $3000 machine the likes of what is talked about on this thread, most of those optimisations might improve things by a negligible % at best. The fact the machine will also be highly hamstrung for anything else useful is neither here nor there.

There is some of that for sure, however you still have to optimize when you get crazy with low latency. Like think about if you have a 32 sample audio buffer, and you are running at 48kHz. That means that the DAW has just a little over half a millisecond to receive a command, send the command to the relevant synth, get the sound back, do any additional processing (you often have chains of effects) and get it to the soundcard. Take any longer than that, you drop samples. Also remember in doing that the system has to a couple thousand times a second drop to kernel mode, switch to a driver, service that piece of hardware, and switch back. It is fairly intense. It isn't the sheer amount of computation that needs to be done (though that can be high with certain synths or plugins) but rather the short timeline it has to be done on. Doesn't take much of an interruption to cause a problem.

Hence why turning up the buffer is the right answer, when you can. On my system I don't do live audio, just mixing. So I run a 20ms buffer because why not? I don't do shit for optimization. I have all the power states turned on, my computer is on the net all the time, it runs all kinds of background processes, etc and I never have a problem. If background processes get greedy and grab things away for a bit, no problem, there's audio in the buffer and it'll have time to refill soon enough. However some people who do live music want really, really small buffers and that is where you can have to start to optimize.

 

[B00nie]

Why can't they just play 20ms before everyone else?

 

[Amblyopius]

However, I am well aware of the advantages of Core Audio over ASIO. I often do want to run two interfaces at once for various reasons (one for input, one for output), and although there are digital multi-interface conections that allow you to do the same thing externally, it is nice to have the option to do it internally. ASIO doesn't allow that AFAIK.

You can do this using ASIO4ALL which essentially sits on top of the respective WDM drivers of the devices and then presents it back in ASIO. ASIO4ALL is often also a good path to low latency.

 

[westrock2000]

He just wanted to let you know that he neither knows how to configure Windows updates properly or turn off the OS sounds. Nothing else to see here.

I will say the ability to direct OS sounds specifically in OSX is really handy for things like Audacity. I just have OSX pump system sounds through HDMI to the monitor and never have to worry about it. I still hear them, they just don't interfere with what I;m doing. Does Windows offer something like that? I

 

[Nenu]

Yes, set the Windows default sound device to whatever you want Windows to use.
Then set up your audio program to use the one you want it to use.