AMD's Ryzen kicks serious a**

[dkasak]

Nice CV/Resume... though as soon as I saw the "I built from scratch..." part, I threw it in the bin.

That's a pretty stupid attitude to take ... not that I'd ever be applying for a role underneath someone with such an attitude. We're about to open-source that ETL framework, by the way, and it's been particularly profitable for us up to this point. Anyway, as you're repeatedly demonstrated, you're more interested in people who make MS Access reports and run them on Dell desktops ... a totally different world

[SOKRVT]

So, has anyone actually tried this cpu yet other than base their experience on benchmarks and rumors?

I saw the thread of u-He but that's about it.

[kmonkey]

As someone that uses other people's money to build out and coordinate support for various systems, currently AMD is never considered.

I think you are being cocky as hell. You speak volumes but all you did is exactly what we all did. Read reviews.

I worked exactly seven years in your elite title "IT Director and Database Engineer" - ATM i am building render nodes for 3d studios (both cpu and gpu) and i have my own studio. Plus web hosting company.

I agree about everything you said above and yes AMD is never considered here as well. However if you think that AMD did not made awesome improvement from one processor family to another (unlike anything i saw from intel in the last 15 years) and that blue company is some bathsit from another space you are blindfolded. Intel finally have to rethink their OUTRAGEOUS prices - your attitude is all over the top and there is something wrong with you. You being 20 year in the business are you sure you did so long with that attitude? I don't trust you a thing.

There is no doubt intel is still leading race but if these new ryzen things turn to be stable enough do you really think people are that stupid to give 400$++more to intel for what 4-8% advantage?

LOL

[jonljacobi]

So, has anyone actually tried this cpu yet other than base their experience on benchmarks and rumors?

I saw the thread of u-He but that's about it.

Seriously dude, look back through the entire thread. Short answer: Yes.

[jonljacobi]

delete

[F5DX]

Just received the parts today and managed to put the computer together. Tomorrow I will start installing Windows and test bios settings etc. I should have some results with audio later in the weekend.

[incubus]

Well, despite my skepticisms, I'll be very interested

[F5DX]

Got the system built during weekend. AMD Ryzen 1700X @ 3.7 GHz, ASUS Prime X370, 16 GB DDR4 @ 2666 MHz, RME RayDAT, Win 10 Pro in High performance mode. Bitwig 2.0 works quite nicely with this machine, much smoother GUI than on my old Mac (was to be expected). So far I tried the performance with only two plugins, Relab LX480 reverb and Reaper ReaXComp that many use for benchmarks. I will mostly use the Bitwig with buffer sizes of 256 and 512, but tested also 128 and 64 requested by some. I don't really have any use for such small buffer sizes. The Ryzen does not perform really well with such low buffers, at least in the case of Bitwig and the RME interface. A lot of CPU is wasted in doing so. I cannot hear any latency when software monitoring synths at 256 buffer, so will most probably stick with that. I have been using 512 buffer for years with the Mac. The plugin counts increase dramatically when increasing the buffer sizes from the really small ones. Bitwig was set to host plugins in individual processes, loading all cores equally in Win 10 task manager. Things will possibly still change a little bit, when Microsoft issues a bigger update for Win 10 to better support the Ryzen cores. At the moment, the loads reported by Win task manager are not nearly full when the audio starts to distort, but Bitwig meter reaches full nearly full.

Relab LX480, default patch:
Buffer: plugins (Win 10 CPU %)
512: 220 (72 %)
256: 168 (67 %)
128: 148 (77 %)
64: 100 (52 %)

Reaper ReaXComp default patch:
Buffer: plugins (Win 10 CPU %)
512: 950 (77 %)
256: 690 (79 %)
128: 400 (71 %)
64: 150 (70 %)

[jonljacobi]

Seems adequate.

[incubus]

I do not understand the tech geek.

All I care about is how a project performs vs "whatever"

[F5DX]

NEW RESULTS with Bitwig 2.0. Just realized that the plugin setting in the menu can have a huge difference in how much cpu the plugins need. I had kept the setting as "load plugins as individual processes" that allows the sandboxing of individual plugins. Now, I changed that setting to "32/64 bit". This made quite a big difference. The DSP level meter in Bitwig dropped quite a bit after rebooting the DAW with the new setting and it was possible to go to full dsp load according to the Bitwig meter, especially when testing the Reaper ReaXComp plugin. I ran these tests with AMD Ryzen 1700X @ 3.75 GHz, 16 GB DRR4 @ 2666 MHz, ASUS Prime X370 motherboard, Bitwig 2.0, RME Raydat, Bitwig plugin setting: 32/64 bit, Ryzen Virtualization and SMT enabled in the Bios (disabling SMT = hyperthreading made the results worse). I updated the motherboard Bios to version 0511 just before running the tests. However, the biggest difference was changing the plugin load mode in the settings.

I tried these tests by running both a 2.5 MB stereo drum loop as well as a much longer 25 MB stereo track. They both played in a similar way and did not affect the plugin count.

Relab LX480 reverb, default setting (10 plugins per track)
Buffer: plugins (Win 10 CPU %)
512: 310 (88 %)
256: 220 (67 %)
128: 175 (64 %)
64: 160 (72 %)

Reaper ReaXComp default setting, 4 bands (20 plugins per track):
Buffer: plugins (Win 10 CPU %)
512: 1540 (77 %)
256: 1040 (85 %)
128: 660 (79 %)
64: 180 (70 %)

Just saying that the Bitwig 2.0 works really smoothly with the Ryzen processor. I have not faced a single problem with this setup yet. Everything works, even when overclocking the processor with Be Quiet Pure Rock air cooler. If you work with ASIO buffer size of 128 samples or more, this is a good choice. Personally, I feel that the 256 samples is the sweet spot.

Edit. Just noticed that the DAWBench uses 8 bands of compression. I ran the tests in Bitwig 2.0 again with the plugin, and got the following counts.

Reaper ReaXComp, 8 bands (10 plugins per track):
Buffer: plugins (Win 10 CPU %)
512: 485 (75 %)
256: 450 (82 %)
128: 325 (76 %)
64: 230 (66 %)

[jonljacobi]

Thanks for all the info.

~Jon

[Scotty]

Yes,thanks a lot of testing and posting. It is great to get some real field test data like this. I use a pair of raydats in my rig so this is highly relevant data for me.

Thanks for all the info.

~Jon

[Skrzypiec]

First of all thank you for this thread - it's like a surface breath to read opinions of sane, experienced people after digging through tons of brand warfare propaganda. Is anybody trying this with Reaper? It sounds tempting to do so, but I must admit I wouldn't even consider an AMD if I didn't recently buy Focusrite Scarlett 18i20 and Metropolis Ark which drained my budget already. I read about memory issues to be solved by BIOS update but as a never-before user of neither 2400 nor 3000+ RAM I can't really tell how big the performance difference can be and how much it matters.

I'm not necessarily choosing between Ryzen 7 and i7-7700K but I seek any CPU/SSD/RAM bundle that can make a 30+ Orchestral Tools/Cinesamples tracks run without choking. If you show me an i5 which can do it, I'll take it. If you say get i7-5690X... well that will hurt but I'll take it*.

*Someone above wondered why to pay +$400 for little more speed... I'll clarify for me it's not about speed but about making sure I won't have problems... and believe me, $400 is nothing compared to years wasted with wrong upgrade choices. If I was sure Ryzen is as good as it seems to be, I would just grab it without hesitating.

Thank you for all given and upcoming help, it's really nice to land here,
also happy imminent Easter, I hope he's Ryzen indeed

[Groundhog #31684]

Studiocat did benchmarking with faster memory...

1800x at stock speed with RAM at 2933MHz
528 instances of the multi-band compressor without glitches (CPU load @ 93%)
 
6850k at stock speed with RAM at 2400MHz
434 instances of the multi-band compressor without glitches (CPU load @ 99%)

It's worth pointing out that both of these processors easily beat the i7-7700k in this benchmark. And there's the option of getting a cheap 1700 and overclocking it.