If we take a chip from today and start developing a product with it, it will take maybe 2 years and… the chip is no longer the newest! As a R&D engineer for stage equipment, I can assure you - development time of 2-3 years is super fast and sometimes too short to sort out some issues in the final product. So if they started development of something 3 years ago - this year or next year it would be on the market. But as the development takes time, the cpu in this product will no longer be the newest. It is just impossible.
3 years was a very optimistic prediction, and super tight schedule. More complex devices like prime 4 can have up to 6 years development time.
RK3588 was released in early 2022. After proper market adaptation, that chip gain approval from other manufacturers within a year, so the product development on that chip could start end of 2022 / beginning of 2023.
That makes the Rane System One was developed within 2 years - taking in the complexity of the device - that is super fast.
Take also in account, that main boards are not off the shelf. They are custom made for inMusic and to make that right - it also takes time.
This still makes a blazing fast development of a new product as Rane System One.
So yes, development and R&D time must be taken into account. But how do smartphone, smart TV, tablet, and even laptop manufacturers manage to integrate the latest chips into their products? Of course, you might say that it’s probably just that InMusic doesn’t have the same resources as smartphone, laptop, or tablet manufacturers, that InMusic probably has a very small team. There’s probably some truth to that, but there’s also probably some truth to the choice of a relatively old chip because older technology is cheaper than newer technology.
They probably could have used a slightly newer chip from 2023 if they had wanted to. But they also probably opted for a Rockchip to maintain backward compatibility with the RK3288, which saved them from having to rewrite a large part of the code for EngineOS.
Currently, the most powerful chip in the Rockchip range is the RK3688, but it is only just entering its launch phase and was only presented in 2024. So indeed, there was nothing more powerful available from Rockchip at the time when product development began for the System One.
So you answered yourself. They did took what was the best option money wise and time wise to bring best possible performance in the given time frame.
How does the laptop or cell phone manufacturer get latest chips? Samsung and apple for example ask the chip factories to make the chips for them in a given spec. But this is super expensive. They need to sell way more products to get the cost back. Denon is not going to end up in every household - there is a lot of djs (potential customers), but not in every one is a dj. But almost every one uses an smart phone. So the cost offset and revenue is more realistic to go and demand your own / latest chip and already start coding for that chip, because you know it’s specification and architecture, before it is even build. You can do simulations and tests together with the chip maker. You can’t do that if you buy a ready made chip. You need to code and develop products according to the specs the chip has, you don’t dictate the specs, pinout, options.
This way - we got a chip that is maybe 3 years old, but still best what was available in the time when development started. So it is best thing on the market at this moment anyway.
Ever since Stems became a thing, several people suddenly seem totally laser-focused on it, proclaiming that any ‘serious’ DJ workflow heavily relies on Stems, calling for ‘MK2’ devices which are desperately needed, and whatever. And any device that doesn’t provide real-time onboard analysis automatically turns into a disappointment. I couldn’t disagree more.
Let me remind you that our main competitor doesn’t even offer any form of Stems playback to this date (when standalone), while Engine OS provides full playback compatibility for any device reaching back to the SC5000 released in 2017 (being based on a 12y old SOC), while also being more affordable. No matter if RK3588 or 3688, any modern laptop/PC/Mac chip will still provide more horsepower to perform the analysis ahead (fast) and simply export the stem files to your drive. Complaining about increased storage need is a questionable excuse, as you can either limit Stems to the tracks that you really need, or just get a bigger drive, if it is really required for your customers - which I strongly doubt - that you need Stems for all of your tracks. Next to my YouTube shenanigans videos, I do also perform on weddings, corporate events and clubs - nobody ever suggested that without Stems my performance lacks any form of ‘energy’. More than enough other tools are still available to be creative, even when entirely ignoring Stems, just as everybody did just a few years ago where this tool simply wasn’t existent (ignoring the exotic .stem.mp4 TraktorDJ file concept).
I regard the RK3588, combined with the doubled RAM, as suitable yet still cost-effective solution for >90% of the targeted user group. The other <10% can still utilize something like Serato or Virtual DJ, which the Rane Performer, Prime 4/+ and SC Live 2/4 fully support. If the RK3688 or even a more recent (and even more expensive) chip would provide, let’s say, 1x (realtime) or even 2x (double) the analysis speed, it would still be too slow for most practical workflows. So you would either have to go for a really powerful (and expensive) chip, or go for a reasonable, pragmatic compromise like the RK3588, that still offers a much smoother EngineOS operation overall (which I can confirm by my own tests). Standalone DJ (=niche) hardware will always require some sort of manageable compromises, and if that means that you have to analyze your Stems ahead on Engine Desktop, heck, simply do it then. One click batch analysis, volia, no drama.
In addition, as Noiseriser stated, economy of scale also applies here.
Exactly. That’s why process intensive functionalities (i.e. Stems) are best pre-rendered / offloaded to another device.
Potentially the ‘stems’ analysis process might see efficiency improvements (looking at the AI developments in recent months/years), but there is way less economic drive in this market. Might eventually end up with an open framework (i.e. what OpenRouter is to LLM) to pick a model to your liking (fast/slow/local/expensive etc), but to be honest most audio related movement or break-through announcements are all about generative audio, voice-cloning or transcription. Not so much the ‘DJ’-sphere. Or did I miss something?
I’m happy with EngineDJ providing the best stand-alone live experience, while preparation or extended live experiences require additional hardware (computers or other devices). Software is key these days, and EngineDJ has been evolving with clear building blocks. I hope they find a way to keep up with the development pace compared to the software-only competitors.
Try storing a library of over 12,000 files on the “AZ” and then scrolling through the playlists, especially if they’re long.
In comparison, the P4 feels like a rocket—and it’s, let’s say, rather leisurely, but okay.
What I’m seeing here on the RS1 (a warranty return) is significantly better.
It’s not the right device for me, but otherwise, it’s really good.
If the P4 or the RS1 could only handle onboard processing, even if it took days, they would be far superior to the competition.
But even that is something they couldn’t achieve on the P4.
Ultimately, not a single customer I know returned their EngineOS device because it didn’t have the advertised onboard STEM feature.
However, it’s a different story when, after the third or fifth email or phone call, I still have to tell the customer, “Sorry, I don’t have any new information for Your problem…”
Then they ask, thankfully usually politely, for an alternative. Guess what happens next…
When I ask how smartphone and laptop manufacturers do it, I’m not specifically thinking of Apple or Samsung, who manufacture their own chips, but rather Chinese brands like Xiaomi, Poco, Realme, etc., which use existing chips from manufacturers like Qualcomm, MediaTek, etc.
I mean, these same Chinese smartphone manufacturers also have to design the motherboard to integrate these batches of chips. They also have R&D phases, but that doesn’t prevent them from releasing products using relatively recent chips. Smartphone, tablet, and laptop manufacturers can use chips from three years ago, but certainly not six years old.
So yes, indeed, it’s not the same type of product, the resources aren’t the same, and InMusic’s R&D team is certainly smaller. But I think there are also choices that are made at some point, such as using older but “good enough” technology to get the job done while maximizing profitability. It’s more of a strategy to put in “just enough” so that the product does what’s expected of it at present while being slightly better than what the competition offers, rather than using a chip truly “designed for the future.”
I think you haven’t understood that those who request real-time stems without pre-analysis do so for reasons of comfort, practicality, and workflow. They don’t want to have to rack their brains wondering which tracks they should pre-render if they improvise a lot during their DJ set. And perhaps they also don’t want to waste storage space unnecessarily storing stems—storage space they could use much more effectively to expand their track collection, for example.
When you have a machine with the processing power necessary for true real-time processing, you don’t ask yourself any questions, you don’t have to think or anticipate anything. You simply take any track, load it onto your deck, and in less than 10 seconds the stems are separated in RAM.
Why would you bother with pre-rendering and storing your stems locally, wasting space unnecessarily, when it’s so much faster with a high-performance chip?
Not so long ago, we often heard the argument, “Yes, but it’s not a problem, storage is cheap these days.” Well, since the RAM prices have risen due to the AI bubble, SSD prices have also increased significantly. Now, 1TB of storage costs the same as the same 2TB SSD did a year ago.
I understand these people very well. The problem is that everyone else would have to pay an unreasonably higher MSRP reflecting the much more powerful chipset (certainly also some R&D/coding, performed by a much smaller team), which is required to achieve results…
and in less than 10 seconds the stems are separated in RAM.
… like this.
In other words: Go all-in (and risk selling far less units overall) or go pragmatic (in this case RK3588 + analysis ahead on laptop + slow onboard analysis for backup/emergencies). I don’t say it’s a perfect solution, but it’s a perfectly reasonable approach catered to our current technology situation. Hardware capabilities and hardware demands rise neck in neck, and the primary market (somewhat) benefitting from economies of scale is the smartphone/tablet and PC/laptop/server sector - but not niche DJ devices. And even on the mass market we face disruptions, as you just stated correctly.
Maybe in 5-7 years, when the next generation of devices follows, the situation might be entirely different and chipsets required for reasonably fast Stems analysis (or AI/LLMs in general) become dirt-cheap. Until then, I think it is not too much to ask any DJ who can afford a >$1500 all-in-one system to spend the extra bucks for a reasonably sized SSD and export the music with Stems on it.
Of course, if some ‘brain racking’ aka basic prioritization and organization is too much I ask for, or if 50k+ tracks in a library are still not enough and need to be expanded even further, well then said user group has to pay the extra for an even larger drive, or maybe stick to a laptop and a DJ controller, as laptops can be replaced easier and chipsets there develop quicker. For everyone else, including myself, Stems pre-analysis provides a manageable solution in my eyes.
We have to start a serious discussion should our competitor come up with a CDJ-4000 or AZ2 that features (fast) on-board Stems analysis. But until we reach that stage, we can probably play GTA 8.
The fact that this topic has reached 10,400 views speaks for itself. It’s a massive signal that the community is hungry for what’s next in 2026. I personally know at least two other professionals waiting alongside me to invest in three new SC LIVE 4 II units the moment they hit the market. We might sound demanding or persistent, but we only hold this standard because we consider Denon DJ the real industry leader in innovation ! Also, big respect to @Skaratak and @NoiseRiser for engaging in this deep dive. Their technical insights into hardware performance and R&D processes have been invaluable, even if we see things from different perspectives. It’s this kind of high-level discussion that makes this community great and shows how much we all care about the brand’s future.
The demand is clearly there, the timing is right. GO DENON GO! It’s time to release new stuff! (8-core CPU RK3588)
Some people have simply googled a processor, read one or two of the specifications like speed or cache, and have jumped to the conclusion that THATS the processor that’s needed for the next DJ deck, without the knowledge or expertise to really know which processor would actually do the required tasks.
I appreciate the concern, but the RK3588 wasn’t just ‘Googled’. It was brought into the discussion by members as a viable path for EngineOS (as InMusic uses this already in the Rane System One). We aren’t claiming to be R&D engineers, we are professional end-users pointing out that 2014 hardware is reaching its limits for 2026 workflows (like real-time Stems). When 10,000+ people are watching, it’s clear we aren’t the only ones jumping to conclusions!
I understand your point of view, but I remain convinced that if the System One, or any other next-generation device, were released with a high-performance chip capable of real-time, on-the-fly stem separation, without prior rendering on the computer or in the background on the device itself, had been sold for $250 more than its current price, you would probably have told InMusic, “Wow, that’s incredible! Take my money!” Because honestly, I think that any passionate DJ who is ready to spend $2500 on an RS1 isn’t going to quibble over an extra $250 to get a sufficiently powerful chip capable of rendering stems seamlessly and achieving a cutting-edge product that’s truly up-to-date, not a generation behind.
Because that’s precisely the problem with standalone units and laptop/controller combos: standalone units will always be a generation behind the laptop/controller combo technologically, mainly due to cost considerations.
Where this would have been more problematic is for products targeted at bedroom DJs, such as the SC Live or Numark Mixstream lines. While a $250 price increase on a premium product like the RS1, a Prime 4 MK2, or even SCX000 players might be acceptable because these are products aimed at a professional segment and therefore easily profitable, a $250 price increase would be harder to justify on an SC Live or Mixstream in the bedroom DJ segment, and in this case, the 3588 remains a good value for money option.
Furthermore, I think this could have been a good way to segment the product line between the pro/prime/premium ranges with a high-performance CPU, and the “bedroom” ranges with the RK3588, to maintain a reasonable and accessible price.
Of course, we can DJ without real-time stems, as we always have, but it’s like how time-stretch/keylock/master tempo appeared on DJ decks; we could mix without them, but once we tried them, we adopted them. The same goes for hot cues; who would want to go back to DJing without hot cues these days?
