AMD Ryzen 3rd gen. ZEN 2 processors for audio PC

[Pictus]

https://youtu.be/W872lQcy65I

[Caroozo]

I'd like to ask, if someone tested mobile 4600H and 4800H vs Intel CPUs (8th/9th gen). I'm actually using i5 8300H, but want to change notebook at the end of the year. My audio interface is Zoom UAC2 (I'm worried about compatibility with USB-controller...) and I'm using CPU-heavy plugins on live gigs, so main priority is overall latency and stability. Desktop Zen2 is quite powerful, but I don't know if there are anyone who has experience with mobile processors .

[kvotchin]

https://www.techspot.com/review/2027-amd-ryzen-4600h/

Some useful comparisons, re: the 4600H. In my opinion, from what I’ve been seeing, AMD is absolutely killing it on both fronts, but especially with laptops if anything.

Note that some of the benchmark charts are headed with “higher is better” when what they mean is “lower is better”, and possibly some are the inverse. Anyway, watch out for those. Otherwise, not a bad review.

[Caroozo]

4600H/4800H are excellent CPUs, I'm well aware about the synthetic tests. But I'm curious about real-life test of those CPUs in DAW - especially under heavy load in live scenario vs Intel CPUs. 1st gen Ryzens had drops under 70+% load (Intel well over 90%), Zen+ and Zen2 were better in that matter - especially in CPUs using 1 CCX and thanks to faster Infinity Fabric. I'm wondering if the same can be told about mobile Zen2 - my notebook (8300H, 12GB RAM, GTX 1050 4GB) has the equal power to run the VSTi I want, but I'm afraid to change it to "better" Zen2 only to find out that there are compatibility problems, or utilization of the CPU isn't "maxed" as it is on the Intel mobile CPUs. With no possibility to test them out, I'm asking on this forum if anybody checked them properly .

[kvotchin]

Hopefully someone else can help in that way.

But fwiw, some of those benchmarks should translate fairly well, AFAIK.

[Caroozo]

I'm waiting for the tests on scanproaudio.info or from some users on the forum - they seem to reduce latency by a great margin, but the configuration of the mobile variants vs desktop are different. The only professional tests I'm aware of are there:
http://www.scanproaudio.info/2019/07/12 ... ic-number/

But as we know, DAW benchmark is about running a lot of smaller plugins, rather than several "behemoths". Nonetheless, it seems AMD is almost on par with Intel when it comes to "utilisation" of the CPU in DAW (drops start appearing at sth like 90%). IPC-wise, AMD is ahead of Intel - lower clock is not a problem in mobile CPU. I'm rather concerned about power states, peak performance and such mechanism in mobile CPUs that give real difference in contrast of desktop CPU. Zen2 is already a beast (and TSMC's N7 node seems to be much better - early R5 3600 OC'ed to 4,1 at most, when newer ones can OC easily to 4,5-4,6). I'm assuming H-variants (45W) are not much different from the U-variants when it comes to quality of silicon - so it's really interesting to test those CPUs .

Hope someone will post his/her thoughts and experiences about Zen2 mobile .

[oobboo]

Hello guys,

it is so hard to find something about the low latency performance of Ryzen-CPUs.
A week ago I upgraded from an Intel i7 2600 to a Ryzen 5 1600 and the general performance increased a lot. Demanding office tasks and gaming (Cyberpunk runs pretty good) is much better with the new system but the audio performance is poor.

I could not use my Behringer UMC404HD, I had to unplug and replug it every few minutes to hours and could not use Ableton simultaneousley with Windows Audio. So I switched to the MOTU M2. Now the system is stable but the latency is pretty high. With a buffer size of 256 I have a roundtrip latency of 17ms and I can run multiple instances of demanding vsts without crackling. Switching to 128, I get 10ms but it crackles directy with one instance of f.e. the Neurdal DSP Archetype Plini guitar vst. The Behringers latency was lower with the i7 2600 than with the Ryzen 5 1600.

It is okay, I am just a beginner in audio production and I can play guitar with 16ms but it is hard to accept.
I can switch the CPU to a Ryzen 5 3600 on the same motherboard without much problems. But how does it perform when I want low latency performance? At the moment I cannot use buffer rates below 256 but with 256 everything is okay. I´ve read that the first gen Ryzens have poor audio performance but does it get better with the newer gens?

(sorry for my english, I am from germany )

[alberto_balsalm]

So I switched to the MOTU M2

Those MOTU boxes have pretty high latency figures, it's entirely possible that your sound card's driver simply can't deliver the performance needed with your current CPU.

Also try to measure DPC latency https://www.resplendence.com/latencymon it's something OS-related, some badly written drivers can cause latency spikes that will affect audio performance. Make sure to rule that out, because if this is the root cause of your problem, you need to treat this problem instead.

Did you make sure to install MOTU's official drivers and use their ASIO protocol in your DAW? Not ASIO4ALL, because this one has nothing to do with real ASIO, but a lot of newbies don't seem to understand the difference.

[fedexnman]

@oobboo yes 1st and 2nd gen Ryzen are bad for latency or audio in general , the 3600 was what I was going to do a build with but my music room is now my wife's workplace . You have to check your motherboard model number to see if it's compatible with the 3600 you may still need to update the bios of the mobo .

[oobboo]

@alberto_balsalm: Sure, I use the MOTU driver from their website. And I also used latencymon. There are some spikes caused by the driver of my gtx 1070ti gpu. But these spikes already existed with my old system and didn´t cause any trouble. The spikes are still green anyway.

@feexnman: So you have no experience with the ryzen 3000 series? my motherboard would be compatible with the 3600. But it is so hard to find real life experiences from users like me. I do not run 100 instances of vsts, just a few but however, the latency has to be under 10ms.

It seems pretty common for ryzen 1st gen that everything below a buffer size of 256 does not work.

[egbert]

Earlier gen Rizen processors have reduced performance at low latencies due to the design of the cache system - you end up with some of the downsides of NUMA systems baked in even though you only have one CPU. With successive generations, the CPU design has been improved and low latency performance is better.

[Pictus]

Hello guys,

Hi,
Some teaks may help
viewtopic.php?p=7624448#p7624448
It may be better if you also connect to the motherboard USB ports that
are directly connected to the CPU, some are connected to the CPU
and others to the chipset, check the manual or ask the support.

The Ryzen 3600 is a good CPU and probably will improve things, with a
stronger CPU you can also use the MOTU at higher sampling rates like
88.2 or 96KHz to lower the latency even more...

[oobboo]

@Pictus:

Some of these tweaks really seem to work. Actually I can reduce the buffer size to 64 (rtl 7ms) with a few instances of the Neural DSP Plugins and 128 (rtl 11ms) with many instances of these plugins without crackling. But I don´t know which in particular is responsible for my results.

My question is: If I load these plugins to individual tracks, route them all to my guitar input and then mute them except for one track, the cpu load is the same as unmuted, right?

If that is the case, I am very happy with the results and can create music with this computer . With a Ryzen 3600 it would only be better, I guess.

[Pictus]

@Pictus:
My question is: If I load these plugins to individual tracks, route them all to my guitar input and then mute them except for one track, the cpu load is the same as unmuted, right?

Good to hear!
I do not know, but I guess yes...