What's (physically) Inside a Symbolic Sound Kyma?

Hey all, had a question for users or those possessing technical experience with the Symbolic Sound Kyma hardware. This isn’t about some sort of debate between value and price, I’m not concerned with that at all, I’d just like to know about the architecture and materials (SoC, chips, etc) that make these up. If there’s an NDA or something asking users not to open it up, would be cool to know that too.

The Symbolic Sound website is laughably vague ie.

Pro Sound Computer
for live performance
40 capstones*

4 processors
2 GB sample RAM
Upgradable to 8 GB
Kyma 7.1 software
FireWire 800 cable
FireWire 400-to-800 cable
≥ 6 processors
≥ 3 GB sample RAM
Wormwire cables

Kyma 7.1 software
FireWire 800 cable
FireWire 400-to-800 cable
Basic Sound Computer
for studio or lab use
19 capstones*

2 processors
1 GB sample RAM
Upgradable to 4 GB
Kyma 7.1 software
FireWire 800 cable
FireWire 400-to-800 cable

Information on any/all of these would be super appreciated and I’d owe you a coffee/beer/kombucha/whatever-it-is-your-drink sometime.

Thanks in advance!!

1 Like

This post compares the Pacarana to the Percussa SSP:

Bunch of technical detail about Kyma in that post. Sorry for the extra SSP noise.


Thanks @jasonw22 !!20charsofenthusiasm

1 Like

This is off-topic I know but that linked post is annoying. I’m not going to make an account on their site to correct them, but comparing clock speeds between two architectures that are as different as the 56k DSPs and ARM chip in the Percussa is not serious. And I’d hope they know that, if they’re programmers.

First of all, if the Percussa has four cores and the Kyma system really has 19 chips with their own memory, for some tasks the Kyma is quite likely going to be more performant. In general, probably not, but the 56ks, while old, are optimized for exactly this kind of workload (doing the same thing over and over again with little to no branching) and if you’ve got 19 of them doing separate things with no context switching, that’s going to be pretty good. Modern ARM chips can be quite DSP-like as well, of course, and maybe they win in all relevant cases by sheer brute force and clock speed, but they should present something more specific than just the megahertz. Having four cores capable of however many instructions per second doesn’t mean you get 4x the performance as one core, that’s just not how it works.

Second of all, the ARM A17 isn’t the latest and fastest ARM core like they say. It’s not bad, but it’s not stellar either. Based on the description, it seems quite likely they’re using the same Rockchip processor (RK3288) that’s used in the MPC Live/X. You’ll also find it in media centers and tablets of a few .years ago.

That doesn’t mean the SSP is a bad product. It looks amazing, and the price is fair for a small outfit. So they really shouldn’t need to prop it up with meaningless number games.


Yep, their comparison is shit. But other than agreeing with you I’ll have to keep my further opinions withdrawn. Majority of the hardware designers I’ve interacted with have soft skin, and I don’t have released products like some of them to provide credibility for myself. Don’t want to buy either, curious about how the players are building things. re: original category was tech

Edit: thanks again to @jasonw22 for the super useful source material

1 Like

Fully agree with this, but every video so far of the SSP is underwhelming at best, and the manufacturer keeps saying “but it’s only five months old”. It’s hard to understand how they could launch such an expensive product with such extravagant claims and not provide anything beyond some mundane videos to show it off, so I’m a bit dubious about it.
And I have a Pacarana system, and I don’t care what numbers the SSP people throw around, it’s no Kyma.


Heh, really just wanted to share the most detailed info I had encountered about Kyma hardware. Sorry to bring SSP into it!

1 Like

I appreciate it, knocked out two birds with one stone for my curiosity.
But I fear this topic will devolve into gossipy shit-talking and I feel somewhat responsible for starting it.

1 Like

informed tech specs are important!

i think we can all refrain from shit-talking though.


I didn’t mean that to sound like shit-talking. I’ll rephrase - the Kyma is an incredible system. That’s all I actually meant to convey.


Worth looking up the Art+Music+Tech with Carla Scaletti. Can’t remember if tech specs were part of it, but definitely lots of interesting stuff on the technical choices they made, if my memory serves…

1 Like

In-process, thank you @_mark !!

I hope my post didn’t come across as shit-talking. To be clear, I’m on record as a huge MPC Live fanboy even though it’s quite modest hardware for the price, but as a singular music making object it’s exactly right for me at least. I’ve never use a Kyma but it’s quite obvious from demos that it’s something very special, and again, I do think the SSP looks like it could become something fantastic.

Specs don’t make products! You can have all the number crunching power in the world, it doesn’t matter if the whole thing, capabilities and interface and aesthetics, don’t add up to something that works. And the thing about synthesizers (as opposed to, say, 3D graphics) is that what works hasn’t really changed much over the past half century. Say something like a Moog Model D, Yamaha TG-77, E-mu E4 Ultra… great instruments are still great instruments.

I’m not sure how many of you remember the computer architecture wars of the 90’s. But I was a big Apple fanboy (before becoming a Linux nerd) and really thought that the PowerPC chips were part of what made it special. And while I still love what’s inside computers, as someone with a growing interest in product design I think this essay hits the nail right on the head: “It’s dark in the box”. http://www.josephpalmer.com/view/box.shtml

The Mac serial ports were much better, but the PC ones were capable of everything I actually wanted to do. One after another the hardware “advantages” of the Mac were measured against the PC, and were found to be better but… All that really mattered was the user experience of the software. It didn’t really matter what was in the box, or who it was from, because it was dark in the box and that was that.


this this this! the design of norns was predominantly about use and possibilities, not specs. the fact that we could use a r-pi compute as an industrial plug-in allowed for the price point and also design time to be diverted to actual design

and no i didn’t take your post as negative, i appreciate the time you took to write. it’s important to help people clarify complicated-looking numbers.


I believe many of the CPU throughput gains we’ve seen over the last 10s of years have usually come at the expense of latency (e.g. branch prediction, caches). A lot of the features required for modern secure computers don’t help either (e.g. MMUs). Thus DSPs still have their place even if their MIPS/FLOPS are low, as there is still a place for CPUs where you can look up exactly how many clock cycles a given operation will take.

Probably DSPs will end up being squeezed out between ARM chips (for low powered, when you don’t really need a true DSP), and FPGAs (for when you need to go stinking fast).

Regarding the Kyma/SSP comparison. My gut feeling is that a modern ARM core used in a DSP style (single user, no memory management) will beat a 56k handily. However I suspect the Percussa is probably running a full OS, and it’s that aspect that makes the comparison tricky.

One more thing to think about, consider the Aleph vs Norns. The Aleph trounces the Norns when it comes to audio latency, especially with it’s dual CPUs (one for DSP and one for UI), but it’s much harder to write code for, AFAIK it has a homebrewed, very low-level OS on the DSP, and the UI chip is coded bare metal. Whereas Norns, is running Linux, but it’s way easier to write low-level code for, and we get so much stuff for free1. Plus the higher level Lua and SuperCollider code makes an amazing amount of DSP power easily available, arguably that accessibility is way more important than raw power (it is to me anyway…).

1 libre free too!


I can‘t contribute to tech specs here but i heard the sounds and was impressed by the fft capabillities! A friend has a capybara since 2006 or 7… apparently he gotten software updates to certain engines over the years-but can the hardware be upgraded? Or is my friends kyma computer just audio fidelity folklore?

The only upgrades to the Capybara 320 hardware are extra DSP cards (can’t remember the maximum off the top of my head) and an output expander. Both of which you’d obviously have to get used these days.

While it may not have the power of the newer Paca/Pacarana models it’s still a very capable piece of equipment.

Absolutely! And its stable on low latency live processing. I‘d say my friend upgraded the chips, i cant say. Its definitely very highfidelty sounding and has a clear mojo like eg. the emu 4 series. The emu‘s magic lies in the DAC imo. Kyma>emu-hence folklore :smiley:

One area I have found that exploits the extra processing power is the Multigrid system introduced in Kyma 7. The tutorial video provides a decent overview: https://vimeopro.com/symbolicsound/kyma-7-tutorials/video/117336966

…one key thing to keep in mind is that in Kyma speak a “Sound” is an entire DSP network so the Multigrid is running many different DSP networks in parallel and switching between them.

As far as hardware capability goes I can say that in my tinkering I haven’t found the edge of the Pacarana. I was able to exhaust the 16 DSP Capybara-320 I had prior without too much effort. The lower overhead of fewer DSPs and lots more memory on the Pacarana really does make the current generation of hardware feel substantially more powerful that what at first seemed like a modest improvement in raw compute power.

At the end of the day Kyma is as much or more about the software than the hardware. It is a extremely deep system that has a lot of unique Sounds (aka DSP blocks)…


It’s running raspbian.

1 Like