AI chips vs GPU

Still umming and ahhing over a new laptop.

In the past I’ve built desktops with basic fanless graphics cards for silence so I thought I’d look for a laptop without a GPU. Then I read that GPUs are now being used by some VSTs and AI functions such as stem separation. I don’t expect to do much of that but don’t want to close the option off. Would an AI processor such as AMD Ryzen AI compensate to some extent for not having a GPU on a DAW?

I don’t know (one way or the other) about VSTs using GPUs, but I do know that the song unmix feature in SpectraLayers 12 Pro is MUCH faster when using my GeForce RTX 3060 12G GPU than my Core Ultra 9 285K 3.7 GHz 24-Core CPU for the AI processing. As for an AI processor, I have no clue. It might be worth asking the question in the SpectraLayers forum. Of course, that answer might be specific to SpectraLayers, rather than general for anything that might otherwise use GPU for acceleration.

1 Like

My understanding is that the answer is no. GPU cores and AI cores are not the same thing. Different drivers and protocols are used to access them.

A good GPU is already beneficial for a premium DAW workstation. Even if you don’t use apps like SpectraLayers for unmixing, having a decent GPU can still offer some speed and visual benefits. It’s not unusual for a DAW, or hosted plugins to tap into resources of a GPU where possible. I’m not suggesting that you need a high end gaming or rendering card, but something solid that supports multiple screens at resolutions you enjoy is important. The more memory the card offers the better if you’re considering getting into the apps and plugins that use the GPU for brute force computation tasks.

The AI stuff is still pretty new, and I don’t think any DAW currently accesses these new cores at this time? Some plugins might be using it by now, but not many. Of course this could change any day now. The AI cores will eventually be exploited in the future I’m sure. To what purpose and extent, I have no idea.

Outside the DAW itself, the AI cores probably get plenty of attention by musicians and other artists even if it is still really new and almost uncharted territory. While the DAW itself might not use it much at this time, it might be helpful with a number of mundane tasks that are related to music production. Research and development seem to be where the most AI action can be found presently.

If considering getting or building a machine with AI cores, it might be good to research AI in general, and what sorts of tasks musicians are finding it useful in achieving. It could be a little while before AI is heavily integrated directly in a DAW like Cubase?

1 Like

AI accelerators (NPU) won’t help at all for VSTs, SL, CB, etc.

A good GPU will help sometimes, but think about if it is worth it for you overall.

For me, it sure would be nice to save few minutes a week on SL, but I’d rather keep my laptop smaller, cooler, quieter (and cheaper too) and just take a break when I start a heavy job on SL once in a while.

If you’re frequently waiting after SL or some heavy process, it’s a different story.

1 Like

That’s pretty much the way I see it and has worked for me so far on a desktop that’s probably way less powerful than whatever I end up with.

If streaming video works fine without a GPU I would have thought any visuals from a DAW should be a breeze. Just a question of whether it’s worth future proofing for increasing AI in plugins. Hence the AI CPU thought.

Thanks all for input.

I could be wrong, but in addition to having to code for an AI processor I think any software wanting to spread the load would have to be coded for that as well. And with for example the recent AMD 395+ APUs the AI processor does 50 TOPS, and the total including that, the GPU and the CPU is about 125. I have no idea how much of the 75 is iGPU and how much is CPU, but clearly that particular solution seems…. not necessarily better than just getting a beefier dedicated GPU in the laptop (for that specific solution I think the massive memory is really the benefit).

1 Like

It 100% depends on what the plugin is coded for.

NPUs / AI Processors, from each brand, and tech like CUDA on NVIDIA GPUs, all have to have the specific models optimized to run on them. Then the plugin needs to be coded (or use a library that is coded) to load the correct execution provider for that chosen NPU or GPU.

So, in short, “it depends”.

Pete

2 Likes

For the most part that’s true, but aside from using the GPU for brute force parallel calculations for things ‘audio’ DAWs and plugins do sometimes take advantage of a GPU if it’s there for UI scaling and some of the drawing tasks. Quite a few plugins do manipulate 3d objects in the GUI rather than just flip through bit images, or make use of the GPU for other image manipulation abilities. I.E. things like vector graphics/fonts and image scaling. These days even the integrated GPU cores on some models of CPUs can handle that stuff, but for what it’s worth, DAWs and plugins do exist that can and will use the GPU for that stuff when at all possible.

Where a dedicated GPU might well be most important with a DAW is connectivity with screens, the resolution you want supported, and how many. Having a really large monitor, or several more moderate sized ones is a big help. So consider how many screens and what sizes/resolutions you want and pick a graphics option that fits your needs with a little room to grow. That’ll be the most important thing.

You can always buy or build something that has the ‘option’ to add or upgrade dedicated graphics options later if you decide you want/need that. With an ATX or larger desktop build, you’ll almost always have easy routes to pop in a GPU (or upgrade an existing one) if you decide you want it. With laptops and mini-PCs, it’s not as easy, but then again, some of those, the entire system is cheaper than a mid range GPU….so…I guess it all depends. Yes, there should be plenty of laptops and mini PCs or Macs out there that’ll run a DAW well without a discrete graphics chip if you’re only running one, maybe 2 1080p or so screens, and don’t need major brute-force GPU parallel computation abilities.

1 Like

Yeah. Much as I’d like to keep it minimal I’m aware that laptops (which I haven’t used for decades) are less upgradeable so if I’m getting something to last a while I might as well spec it higher for future developments. Yes, it’s not easy to see right now which software will actually benefit from which hardware. I had in mind for the laptop a Thunderbolt dock to hang the bigger, immovable stuff, including large screen(s), so I could tuck the machine somewhere the fans can’t be heard - as with the desktop now. So I won’t get too hung up on noise which was my main reason for the question. Away from the studio I can put up with it.

Thanks all for input