BEST CPU FOR REALTIME AUDIO

[ariajazz]

Hello, I'm about to build a new PC essentially to play my guitar live, using plugins like Amplitube, Guitar Rig 5, Scuffham S-Gear and some Soft Synths triggered by Jam Origin Midi Guitar. All these as VSTs within Sensomusic Hollyhock III as a host.

My low budget options are:

INTEL:

CPU: Pentium Gold g5600 (3.9 GHz - 2 cores, 4 threads)
https://www.amazon.com/Intel-Pentium-De ... ntium+gold

MB: MSI Pro Series Intel Coffee Lake H310 LGA 1151 DDR4 Onboard Graphics Mini ITX
https://www.amazon.com/dp/B07CF3TRFQ/?c ... _lig_dp_it

RAM: Patriot Signature Line 8GB (2 x 4GB)
https://www.amazon.com/dp/B016A29ETY/?c ... _lig_dp_it


AMD:

CPU: AMD Ryzen 3 2200G (3.5 - 3.7 turbo Ghz, 4 nucleos 4 hilos)
https://www.amazon.com/AMD-Ryzen-Proces ... en+3+2200g

MB: MSI Performance Gaming AMD Ryzen 1st 2nd Gen AM4 Mini-ITX Motherboard (B450I Gaming Plus AC)
https://www.amazon.com/dp/B07CF3TRFQ/?c ... _lig_dp_it

RAM: Patriot Signature DDR4 8GB (2x4GB) 2666MHz (PC4-21300) Dual Channel
https://www.amazon.com/Patriot-Signatur ... 4+2666+and


The price difference between these two configurations is about 40 dollars in favor of the INTEL one. My question is: which of these two options would be the most appropriate. And here comes the eternal debate of "more speed or more cores". I've done a lot of reseach and I still do not have this topic clear enough, people do not seem to agree on this and there are very different opinions.


The way I understand it is that to obtain low latencies buffer sizes of 64 samples and lower demands higher CPU speeds, here the Pentium Gold wins, in addition to having a lower power consumption (51W) compared to 65W of the Ryzen, this is an advantage when setting up a MINI ITX machine. But the Ryzen 3 has more cores and apparently has a better architecture than the Pentium. On the other hand, I understand that the greater the number of channels, tracks and plugins, the more cores will distribute the workload better, depending on how the host and the plugins manage the multicore resources, which I do not know.


My regular configuration with Sensomusic Hollyhock host is off 5 or 6 tracks, of which 3 or 4 at the most process audio, the rest only data for the system operation. I find myself in this great dilemma of choosing between the INTEL or AMD configuration. What I'm looking for is the best possible performance in terms of latency, the lowest possible. It would be great if I could get 32 ​​samples (the lowest allowed by my ESI MAYA 44 XTe interface). Currently I can get 128 samples in an absolutely stable way without drops, glitches or any other artifacts with an AMD Athlon II x2 270 processor (3.4 GHz).

Waiting for you responses and opinions. There are many people in this forum with a lot of knowledge and kindness. Thanks in advance.

[lfm]

I does not matter one bit which cpu you choose - if you spend money on the right audio interface.
My old 15 year XP computer with a P4 single cpu core(2.8 GHz) would do the job as described with an RME interface. It was my former daw.
Any modern computer would do the job(>2 GHz clock).

Drivers and brand of audio interface is what matters to get low latency.
When it comes to running loads of plugins at the same time - then cpu matters.
So running a daw with as large projects as possible without freezing/rendering some track, all in realtime - then cpu matters.

RME would be a good choice for audio interface.
Spend money on that rather than computer - for the task you mentioned at least.
If computer is to serve other purposes - increase computer money for more headroom.

[fmr]

I does not matter one bit which cpu you choose - if you spend money on the right audio interface.
My old 15 year XP computer with a P4 single cpu core(2.8 GHz) would do the job as described with an RME interface. It was my former daw.
Any modern computer would do the job(>2 GHz clock).

Drivers and brand of audio interface is what matters to get low latency.
When it comes to running loads of plugins at the same time - then cpu matters.
So running a daw with as large projects as possible without freezing/rendering some track, all in realtime - then cpu matters.

RME would be a good choice for audio interface.
Spend money on that rather than computer - for the task you mentioned at least.
If computer is to serve other purposes - increase computer money for more headroom.

You are right, in general. However, the OP mentions "some Soft Synths". It depends on which soft synths he/she is thinking, because "some Soft Synths" may kill the CPU. It depends which soft synths are we talking about.

[ariajazz]

Hi, Thanks all for replying.

Soft synths that I use are very low CPU usage. My interface is an ESI Maya 44 XTe (PCIe), drivers are pretty decent but I must say that I have achieved better results with ASIO4ALL. I don't have any plans on getting a new one. I'm planning to build thid new system because my current one is faulty. In any case, which one of the processors would you choose'

[fmr]

Hi, Thanks all for replying.

Soft synths that I use are very low CPU usage.

Then I entirely subscribe what lfm wrote

My interface is an ESI Maya 44 XTe (PCIe), drivers are pretty decent but I must say that I have achieved better results with ASIO4ALL. I don't have any plans on getting a new one. I'm planning to build thid new system because my current one is faulty. In any case, which one of the processors would you choose'

The fact you are getting better results with ASIO4ALL means that the original drivers are not good. So, you better look for another audio interface. Remember that a system is as powerful as its weakest component.

Regarding the CPU - buy the most powerful you can get with your budget, no matter what.

[ariajazz]

OK. Let me point out something. I've always known that ESI drivers are very good and stable. Any way this is what I have experienced: If I have a pretty simple setup (few plugins) I can reach 64 samples easily with ESI driver, but if I have a complex setup (Hollyhock is a modular software) with many more plugins, then I have to raise the buffer size because I start to get drops and glitches and there is where ASIO4ALL tends to work better. So that's why I suspect that it is a CPU power issue. My current one is an AMD Athlon II x2 270 (3.4 Ghz).

[fmr]

OK. Let me point out something. I've always known that ESI drivers are very good and stable. Any way this is what I have experienced: If I have a pretty simple setup (few plugins) I can reach 64 samples easily with ESI driver, but if I have a complex setup (Hollyhock is a modular software) with many more plugins, then I have to raise the buffer size because I start to get drops and glitches and there is where ASIO4ALL tends to work better. So that's why I suspect that it is a CPU power issue. My current one is an AMD Athlon II x2 270 (3.4 Ghz).

Hmmm... maybe it's better for you to try another audio interface before moving to something else then. It may be a problem with the drivers (that interface is old AFAIK), the CPU (as you suspect), or you may as well being fooled by ASIO4ALL (I never trusted it). If you can access an RME audio interface, try it, and check how the driver behaves.

Anyway, don't you have (for example, in the DAW you are using) a CPU loader meter to check, while you are experiencing those drops and glitches?

[ariajazz]

Thank Fernando for you interest. I know the whole rig is old (built in 2012).... that's the problem.... when you buy a guitar amp, you spect to have and keep it for a long time, specially if it is a good one. In computer world is different, I can not afford those RME or UA interfaces, I know those are the best drivers available, but I have achived pretty decent results with my current configuration... 64 samples buffer size with simple setups which is 2 to 3 ms. which is not bad at all. you get more latency moving several feets away from a real amp. In any case the fact is that my current PC is experiencing weird behavior with BSODs and freezes no matter what I do at software and hardware level. I suspect something is faulty, due to this I considered that is time for a renewal. Any way what I want from you guys and under your point of view is an opinion of which one of these processors would be your choice accordingly to all that has been stated in this thread.. Than You...

[lfm]

ASIO buffer size is not the full story of latency - it's also about how fast the converters are - both ways AD and DA. This for roundtrip latency to be low enough like for Amplitube. This matters for doing guitar that route. You are most likely not close to any 3ms roundtrip checking that out.

Some daws show inheret delays + buffers separately - like Sonar - if interface has built in converters. Certain pure digital connections interfaces exist - and then only buffers are known to system - so you have to calibrate manually with a loopback thingy. A simple test for real latency is to make connection from analog output to input - turn off latency compensation - and record metronome - and see how it differ from grid once recorded.

My guess is you are more close to 10ms than 3ms experiencing what delay guitar has through.

And as you mention - getting away from amp a bit - 340 m/s speed of sound also adds and probably make delay noticable quicker. 3m is a full ms extra. But that goes for any live band, I guess. But analog gear does not have inheret delay noticable in themselves to start with.

I doubt choice of cpu will make any difference that you notice. Maybe more headroom so less issues if running smoothly at really low cpu.

[ariajazz]

I'm aware of the roundtrip latency facts. I was only referring to the buffer latency.... and you're right probably the real latency is around 10ms which is not bad at all, actually is not noticible, almost imperceptible. ESI Maya 44 AD/DA converters are pretty good, not the best but good enough, also its hi-Z instrument input is awesome. Sensomusic Hollyhock as a host adds no extra latency in normal situations since is designed for live purposes. I haven't done the loopback thing on this equipment but I will using "CEntrance" Latency Test Utility wich is pretty accurate.

I believe the way to go is to get as much headroom as possible CPU power wise... and that's what I'm after. The eternal debate: "more speed or more cores". One question could be what to go for, better single core performance or overall multicore. AFAIK audio streaming occurs in only one core.

[BertKoor]

I still use an Audiotrak (ESI) Maya1010, and can tell you that the drivers this company made can be categorised in the same league as RME: within the best 20% of what's available on the market.

Regarding Asio4All, it piggy-backs on WDM drivers but skips some layers like the mixer that add latency. My guess is that using Asio4All gives the same results as ESI's own ASIO drivers with maximum buffer size of 512 samples.

If you're going to invest in a new computer, you might as well factor in a new audio interface. It's just that the CPU you pick is not really relevant. Other components have much more influence. Wifi and GPU chips, low-level BIOS routines... they think they are the most important and hold the bus hostage, and thus cause dropouts.

My advice is to get a PC specifically assembled for realtime audio. There's just one company that makes these I'm aware of: Scan in the UK. There were partner of ADK in the USA, but I think that disappeared. Also the time of big cases is of the past. Save yourself from lugging stuff around, setting up displays and a keyboard. Just get a laptop. You can fit them in a 2-unit 19" case which flips open for life on the road.

[Kaine]

ADK's still hanging on. Scott's in the process of carving up the company and will continue to look after the video arm. Chris who's overseen support for the last couple of years has taken over the audio side of the company and will be relaunching it again properly over the coming months.

Normally I'd lean towards the higher core speed for live work, which puts Intel at 2200 vs 1850 in CPU benchmark for the single core score. However, if your using multiple channels of instruments as you appear to be, I'd probably say a true 4 core without SMT or HT (i.e. the one you listed there) might make more sense overall if it's spreading the load properly.

Just to mirror some of the other comments, driver is more important for latency as long as you have enough overhead to play with. You can't add performance and expect a latency drop, unless you're already bottle necking. As already noted buffers are no real reflection of real world performance.

You can get 64 buffers that have a total RTL result of 2.5ms and you can get 64 buffers that have a total RTL of 22.5ms and everything inbetween, that all comes down to the quality of the driver and the design of the hardware. In either case, upgrading the CPU isn't going to make any difference to that.

Now, that's not to say you don't need an upgrade, given your workload that's asking a lot of the single core of any vintage and you might well see an improvement to you're currently slamming into a wall with current workloads. I'm just saying that you also shouldn't be surprised if it makes no difference at all to your RTL in real terms.

[ariajazz]

Hi all. Thaks a lot for your replies, a lot of good information here. I'd like to point out some things here.

I use to work with an ESI Audiotrak 1010 (PCI) and it always performed brilliantly, also I've always known that ESI drivers are very good and solid, didn't know they were on the same league as the RME. That's why I chose the ESI Maya 44 XTe sis years ago, therefore I believe this interface works great.

It seems to be clear that single core speed would benefit the specific task the system is supposed to do - realtime audio - but multicore CPU will distribute the workload better. So what would be more important? (the ethernal debate, more speed or more cores).

Taking in consideration everything explained here, which one of these two processors proposed you'd go for?. I'm not planning to get a new audio interface, ESI Maya is just working fine and I'm pretty happy with it. It is not only that I want to upgrade hardware, it is that the current one is acting weirdly (BSOD, freezes, spontaneous reboots, etc,) I've check every possible cause hardware and software wise and I suspect there's someting faulty.

All you guys have put your 2 cents on this thread and very valuable information is here, but is time to vote. I'm slighty in favor of the Ryzen 3, four physical cores, good speed up to 3.7 GHz which is more than I have now, apparently better and newer architecture and Radeon 8 graphics (however I don't need much).... I use a touch screen, a midi footswitch controller and a Behringer X Toucn mini,,, So everything is under control.

Waiting for your vote.... Thank you very much to all you guys

[Kaine]

Take the lowest clock speed as your working speed, not the turbo figure.

Audio drops out when the first channel hits the wall, not the last one, which is why I clock lock as high as possible during testing.

That said, you're right, you've got a far higher IPC result than your current solution no matter whichever chip you go for, and in this instance, I'd probably favour the 4 core AMD.

My caveat here is to check that any hardware you have is fine on the AMD chipset and USB in perticular.
First generation Zen platforms USB integration was ASMedia supplied and could prove picky with some hardware options, largely real-time (i.e. audio interfaces) devices running USB3 ports. Thankfully this is largely cleared up in the current second-gen hardware from what I've seen, as it seemed largely resolved in my last batch test but it wouldn't hurt to check before building the new setup.

[ariajazz]

Thanks for your reply and for your advice. The ESI MAYA 44 XTe is a PCIe interface and it's been working nicely with Athlon II x2 platform, so I think it won't have any problems with Ryzen 2nd generation.

"Audio drops out when the first channel hits the wall, not the last one, which is why I clock lock as high as possible during testing."

What do you mean by that?

Thank You.