What's new

My Ryzen 3700X build for music production

Pier

Senior Member
I've mentioned in other threads that I've been building a Ryzen machine for music production. I'd like to share my specs, comments, and results.

TL;DR: For my use case which is mostly using virtual synths, this thing is a beast.

Specs
CPU: Ryzen 7 3700X
Motherboard: Gigabyte Aorus Wifi Pro B450
RAM: 16GB 2133Mhz (I used 2 8GB sticks from a previous build)
GPU: ASUS Strix 1070
PSU: Corsair RM850x
SSD: Kingston Digital HyperX Predator (also used this from a previous build)
Cooler: Noctua NH D15 (still waiting for this, using stock cooler for now)
Audio interface: Audient iD4 via USB (latest drivers and firmware)

Installation notes
I bought the B450 board because it was relatively cheap and I'm not interested in overclocking. Unfortunately I didn't read the fine print which is: the B450 boards need a BIOS and firmware update to be able to run Ryzen 3000 CPUs, and you need a supported CPU to do this. So after running the machine the first time and panicking I figured it out and had to buy a cheap 2200G Ryzen CPU to be able to update the BIOS. Apparently AMD offers free boot kits for this but since I don't live in the US I really didn't want to wait.

I also needed to install the Windows Firmware drivers from the AMD website since the ones from Gigabyte are not the latest.

Windows 10 installed some old Nvidia drivers. I manually installed the latest ones but did not install Nvidia Bloatware Experience.

About noise and temps
The 3700X is rated at 65W of TDP, but I've found this to be very misleading. I'm sure this technically correct but it's not the whole story.

I have a gaming machine with a quad core i5 6500 which is also rated at 65W and I can run it at ambient temperature on idle with a medium sized cooler (Scythe Kotetsu). I expected the same from the Ryzen. Boy I was wrong.

The first time I ran Windows the CPU was running at 60ºC on idle and the stock fan was screaming. After some googling I found a post on Reddit and they recommended the firmware drivers from AMD which did lower the idle temps to about 35-40ºC. Still too high. After reading some more I learned that Ryzen CPUs usually have higher temps on idle than Intel, much like an engine running at higher RPMs to be able to turbo boost at any moment or something like that. It was frustrating at first, but knowing all this it was just a matter of adjusting the fan speeds on the BIOS to run very low until the CPU was higher than 50ºC similar to what Apple does. After doing that the CPU is almost inaudible when the room is in complete silence.

With the stock cooler the max temps have never gone beyond 77ºC when checking with Ryzen Master.

I'd like to add that the Strix GPU and the Corsair PSU do not turn the fans on until there is a significant work load so the only noise I hear is the CPU fan. I also have 2 Nanoxia Deep Silence fans on the case which are inaudible from the sitting position running at low RPMs.

Special configs
I haven't messed around much with any custom configs:
  • Custom fans config in the BIOS
  • Ultimate Power Plan (added with a terminal command)
  • Disabled the onboard audio chip in the BIOS
  • Uninstalled a bunch of games and Microsoft crap included in Windows 10
Benchmarks
Latencymon is all green and reports a max DPC latency of 248us which is fine for me. I've been using the Wifi chip from the motherboard during the monitoring.

CPU-Z benchmarks report results a bit better than the average for the 3700X I guess I got a bit lucky with the silicon lottery.

At 256 samples of buffer Ableton Live reports 10ms of latency which is totally fine for my use case.

So here comes the most interesting result for me. I run a small benchmark in Ableton Live with Hive 2 in the machines I have. I use the default preset (a bass with effects and a couple of unison voices). So I write 1 note 1 bar long and start duplicating tracks until I hear audio artifacts at 256 samples of buffer.
  • Old gaming machine (i5 6500) 34 tracks
  • 5K iMac (i5 7600K) 40 tracks
  • New machine (Ryzen 3700X) 111 tracks
So almost triple the performance of the iMac which is what's expected by looking at Geekbench and other CPU benchmark sites.

After the initial bumpy road with this build I'm now starting to salivate at being able to throw pretty much anything I can probably need at it. :D

I will run more tests with Diva and other synths but this is all for now!
 
Last edited:
Congrats on the new build! I didn't see it mentioned, what case are you using? and Any plans on upgrading ram in the future?
 
Great specs except there‘s a DDR Sweet Spot Ive read about a few times that really kicks up performance by a measurable difference.
Has something to do with Sync and Speed which is in between 3600-3700MHz.

Its a YouTube video and very well done too since the guy does talk but every word is useful. Unlike some demos of other audio related topics where the person is more interested in hearing themselves speak. Way too many of these imbeciles @ YT.
 
Congrats on the new build! I didn't see it mentioned, what case are you using? and Any plans on upgrading ram in the future?

The case is a Fractal Design Define C mid tower.

Yeah I will upgrade the RAM at some point but it's not a priority. I don't think I will see a dramatic performance difference tbh.
 
Great specs except there‘s a DDR Sweet Spot Ive read about a few times that really kicks up performance by a measurable difference.
Has something to do with Sync and Speed which is in between 3600-3700MHz.

Its a YouTube video and very well done too since the guy does talk but every word is useful. Unlike some demos of other audio related topics where the person is more interested in hearing themselves speak. Way too many ofnthese imbeciles @ YT.

Thanks for the tip, I'll see if I find more info on that sweet spot. If it's a big performance increase I don't think I will be able to resist the temptation and see how much I can push this beast.

Edit: holy shit some games are getting a 50% increase in FPS with a 3600 RAM over 2133...

Amazon here I come again...
 
The first time I ran Windows the CPU was running at 60ºC on idle and the stock fan was screaming. After some googling I found a post on Reddit and they recommended the firmware drivers from AMD which did lower the idle temps to about 35-40ºC. Still too high.
I too was surprised, at first, with this when I did my 3700x build. Chipset driver updates and new Bios fixed most of it. I used to see 60C idle and now I see in the 40's but that's also with a Noctua heatsink with 2 Noctua fans and 6 120mm case fans. If it were my old Intel chip it would be running in the low 30's with that beast of Heatsink and fan setup. I also think the temps show higher than they really are because the way AMD's monitoring system constantly pings the bios for readings so you're getting more of a real time read out of the temps rather than a more averaged out measurement.

I'm using 64GB 2400mhz rated memory that easily runs at 2666Mhz (I think they're actually 2666hz sticks with baked in xmp profile of 2400Mhz) If I were to do it again I would have gone for at least 3200Mhz since the price was not that different but I was anxious and it was the only available memory locally at the time. When I upgrade to 32GB sticks for an total upgrade to 128GB of ram then the only speed available for those sticks currently are 2666MHZ by Samsung and Crucial so maybe it's good I'm not spoiled with higher speed memory:)

Glad to see you're enjoying your system.
 
I too was surprised, at first, with this when I did my 3700x build. Chipset driver updates and new Bios fixed most of it. I used to see 60C idle and now I see in the 40's but that's also with a Noctua heatsink with 2 Noctua fans and 6 120mm case fans. If it were my old Intel chip it would be running in the low 30's with that beast of Heatsink and fan setup. I also think the temps show higher than they really are because the way AMD's monitoring system constantly pings the bios for readings so you're getting more of a real time read out of the temps rather than a more averaged out measurement.

I'm using 64GB 2400mhz rated memory that easily runs at 2666Mhz (I think they're actually 2666hz sticks with baked in xmp profile of 2400Mhz) If I were to do it again I would have gone for at least 3200Mhz since the price was not that different but I was anxious and it was the only available memory locally at the time. When I upgrade to 32GB sticks for an total upgrade to 128GB of ram then the only speed available for those sticks currently are 2666MHZ by Samsung and Crucial so maybe it's good I'm not spoiled with higher speed memory:)

Glad to see you're enjoying your system.

I just bought 16GB of 3200Mhz and will report back in a couple of days. I wanted 3600 but it was too expensive and from the videos I've seen the performance increase from 3200 to 3600 is not that big.

Honestly I'm totally fine with the current performance with the 2133 RAM but since I was going to buy new RAM in a couple of months anyway well what the heck :)

I'm sure you will be alright with 2666.
 
I disabled CPB in the BIOS (core performance boost) to see if it made any difference in audio perf. Same idea on both synths. One long note for 1 bar, and then duplicate tracks until I get crackles.

I got these results (I didn't run both synths at the same time):
  • Diva: 29 tracks (patch: HS Amped Forces)
  • Hive 2: 97 tracks (default patch)
I chose that Diva patch because it has 4 stacked voices and uses 2 effects, so pretty beefy.

I then enabled multicore in Diva (same patch) and I only was able to get 14 tracks. I'm almost certain the problem here is the latency between the core groups which someone explained in another thread (not sure if here or on Gearslutz).

Another interesting thing is that with CPB off I got about 10% less audio tracks with Hive 2.
 
I disabled CPB in the BIOS (core performance boost) to see if it made any difference in audio perf. Same idea on both synths. One long note for 1 bar, and then duplicate tracks until I get crackles.

I got these results (I didn't run both synths at the same time):
  • Diva: 29 tracks (patch: HS Amped Forces)
  • Hive 2: 97 tracks (default patch)
I chose that Diva patch because it has 4 stacked voices and uses 2 effects, so pretty beefy.

I then enabled multicore in Diva (same patch) and I only was able to get 14 tracks. I'm almost certain the problem here is the latency between the core groups which someone explained in another thread (not sure if here or on Gearslutz).

Another interesting thing is that with CPB off I got about 10% less audio tracks with Hive 2.


I noticed the same with hyperthreading on my i7 4790k rigs.
I disabled it and have much better timings and latency is all but disappeared.

Im sure they’ve got better with HT since the 4000 series CPUs.
One of the reasons I try and stay with quad cores for live performance.
 
I noticed the same with hyperthreading on my i7 4790k rigs.
I disabled it and have much better timings and latency is all but disappeared.

This piqued my interest as I'm running an i7 4790k. Are you disabling hyperthreading in Kontakt/your DAW or on the chip via the BIOS?
 
This piqued my interest as I'm running an i7 4790k. Are you disabling hyperthreading in Kontakt/your DAW or on the chip via the BIOS?

Kontakt and the ASRock BIOS. No Turbo, no Hyper, just a 4ghz 4 thread stream.
I can get a pretty substantial boost in layering and polyphony by Overclocking to 4.4/4.6 and I dont need it for my live rig.

Not saying hyperthreading isn't useful for recording, editing or rendering, just seems to add additional latency for the real time work I do.
Its not noticeable except when I do my heavy sustain heavy layer heavy poly tests.

I‘ll take PianoTeq, Kontakt and Keyscape and just drive the layering as hard as I can.
Unfortunately Keyscape drops out due to design, but PTeq and Kontakt just go on and on with zero clicks pops or dropouts.

In all honesty for my live rigs I havent really needed additional IPC or CPU Power since my first i7 Bloomdale CPU.
I only upgrade because after three years even though the rig still works Id rather not experience a crash.
 
Wow, I don't remember you mentioning that before. I may try turning HT off, and see if I get a boost. I'd love to go to 64 samples, which mysteriously worked for one day, and never since.

The 5775c is still going strong, can't complain. Pushing 1K of live gigs/sessions at this point. Looked at getting another one, they are still selling used for what I paid for mine 3 years ago.

To get back on topic: Can you give us some Ryzen comparisons at 128 and 64 samples? I can't go back to 256....
 
To get back on topic: Can you give us some Ryzen comparisons at 128 and 64 samples? I can't go back to 256....

So I enabled CPB again and got these results:

Buffer 256 @ 48Khz
  • Hive 2: 100 tracks
  • Diva: 32 tracks
Buffer 128 @ 48Khz
  • Hive 2: 85 tracks
  • Diva: 32 tracks
Buffer 64 @ 48Khz
  • Hive 2: 64 tracks
  • Diva: 25 tracks

Again Hive on the default patch and Diva on HS Amped forces. The difference with previous tests is because I was on 44.1Khz. I was surprised that 256 and 128 made no difference with Diva.

I'm still waiting for the 3200 RAM...
 
So how does that compare with your old machines? Are you seeing similar gains at lower latencies? Those are encouraging results, especially at 64 samples!
 
The old gaming machine with the i5 6500 is not working at the moment... I used a couple of components in the new Ryzen machine. :P

As for the 2017 iMac 5K i5 7600K:

Buffer 256 @ 48Khz
  • Hive 2: 36 tracks
  • Diva: 16 tracks

Buffer 128 @ 48Khz
  • Hive 2: 28 tracks
  • Diva: 12 tracks

Buffer 64 @ 48Khz
  • Hive 2: 15 tracks
  • Diva: 8 tracks

Overall the Ryzen machine is 3-4 times more powerful and it cost me about 1/3. The previous test with the iMac and Hive were done at 44.1Khz.
 
So I received my 3200Mhz CL16 RAM and redid the tests with Hive and Diva. It made barely any difference with my previous 2133Mhz RAM! In some cases I got 1-2 more tracks but it's pretty much negligible.

Oh well, at least I will get some more FPS when gaming :)
 
So I received my 3200Mhz CL16 RAM and redid the tests with Hive and Diva. It made barely any difference with my previous 2133Mhz RAM! In some cases I got 1-2 more tracks but it's pretty much negligible.

Oh well, at least I will get some more FPS when gaming :)

Just figure you’re now more likely eeking out every bit of performance. If you went to 4000 speed you’d possibly get another track. But loading samples should also be faster, opening a template, etc.

And, yes, FPS!
 
Latency RAM affect only benchmarks. In real performance you can't noticed the difference. But now you have it use it and enjoy enven if your gain are weak! ;)
 
Latency RAM affect only benchmarks. In real performance you can't noticed the difference. But now you have it use it and enjoy enven if your gain are weak! ;)

It depends...

In Ryzen CPUs RAM speed has a great impact on some workloads, specially some games.

For audio it seems not, at least in my use case which is virtual synths and effects which is mostly CPU work. I imagine the results would be different with a heavy kontakt instrument. I can do some tests if anyone is interested. I own Albion One and a few other Spitfire libraries.
 
Top Bottom