The Framework desktop can game, but its primary strong suit is crunching numbers within a machine learning task. Furthermore, if you would like your game to be playable by an audience with varying configurations and processing abilities, I don’t believe it is wise to lock yourself to such a specific hardware configuration. If the only way to play your game is to have a beast computer most can’t afford, it wont get played much.
I would suggest you read a bit more up on this. Yes, machine learning is able to run on GPUs quite well, and so the GPU in the Framework desktop is no slouch, but perhaps you should be building a normal desktop instead of trying to get a super compact one, with limited expandibility. (I love the Framework desktop. Just trying to give helpful advice.)
I see, so the AI stuff really does not help in my case then? There’s already a lot of reasons the Framework Mainboard and the new AMD chips are appealing anyway. I’m not only looking to develop video games, there’s many new things I want to do with a powerful computer like this. If anyone is pro right-to-repair, then the computer should be able to be configurable in unique ways. I’ll stay tuned in to find out how GPU’s interact with converters in the NVMe and PCIe x4 slot. I find this technology and scene to be pretty important even if I’m not very knowledgeable on the subject. And of course everyone I know and love uses AI at this point, and so have I even if I’m unaware of it. So I wouldn’t turn down an opportunity to try out some kind of LLM work. I’d be down to check out all kinds of cool tech applications. I really do appreciate the advice and taking the time out of your day to answer some of my questions : P
Sorry, I’m not really qualified to answer that. I just wanted to spare others - who might also struggle to remember the correct meaning of GB/s and Gb/s - a web search.
I can give you my take on the framework desktop though. I am a web developer and could do any of my usual tasks (mostly developing websites, backends, API integrations, html templates, server administration through SSH, maybe developing a simple vuejs/typescript/ionic based multi-platform web/smartphone app) easily on the weakest framework hardware. Even on any of the laptops.
Where the desktop will hopefully shine compared to my current Intel + nvidia RTX 3060 based desktop will be in more resource intensive tasks. Thinking of speeding up video editing + especially rendering, running my own AI stack locally. All doable on my current desktop but i assume much slower and especially with AI very limited by the GPU RAM on the 3060.
Marketing for the framework desktop compares it’s performance to an Nvidia 4070. Which would be better than my current 3060 and with the shared system memory will most likely be a true game changer for AI applications. Even if it’s slower than a 4070 I expect it to at least be on par with the 3060, but much more capable in terms of big models.
The crazy thing for me personally: all that in this awesome tiny form factor. Most of the time will hopefully be almost completely silent - something I would greatly appreciate. I might be even more hyped about that than any performance increases. Removing the chunky desktop casing from under the table and just having a much cleaner work environment. And it’s even portable enough to carry it in a standard EDC backpack.
Not an expert on game development sadly. IMO it’s almost a philosophical question. Success is all in appealing to your players. Awesome gameplay and some unique combination of ideas is bigger than supporting the latest hardware requirements. Look at Minecraft
Agree with @2disbetter that you absolutely can develop great games on the framework desktop. If you want to cater to absolutely beast machines then of course it makes sense that you build on a machine with those specs. But I would expect the framework desktop to cover 80% of the existing global gamers. Percentage completely fictional, but I would assume most people don’t play on killer machines - on a global scale.
The prospects of portability is super awesome, I really want to bring it to friends who also do not have easy access to such technology. The idea I had was to put down $200-300 for the equivalent number of mainboards. Then the skies the limits, especially if I can maybe organize a project that isn’t just solo development. Solo development is insanely difficult and I commend all video game solo-devs. It’d be awesome to organize a project, and develop a portfolio of some sort. I’ll give a small project a shot on my desktop once I fix it lmao.
AI can obviously be used as a tool for e.g. helping with coding or generating game textures, but otherwise the compute unit inside the CPU is more useful for raytracing and rendering photorealistic scenes (as a static image or for a movie) and less as a gaming GPU.
It’s probably more cost efficient to get a decent standard gaming pc, unless you really want to run the ai stuff locally.
You can see it as an upper mid-range product for most applications. It will handle most tasks just fine, but might not be the ultimate and fastest in most tasks.
It does have some unique properties, though:
more energy efficient than traditional / older systems with similar capabilities => saving energy
much smaller form factor in comparison to traditional / older systems with similar capabilities
even with the small form factor it has excellent cooling and should be very quiet when not under heavy load, which traditional / older systems often are not
in this price segment it is kinda novel that you can allocate system ram to be used by the integrated GPU, which is very useful in running local AI models, as they could not be loaded by other consumer cards easily previously
There might be more, but those are some of the main ones. It essentially should be able to handle a lot of different use cases, just not the fastest and best in every application. Peak performance in a very specific area can be easier achieved with purpose built machines that are tweaked that way. Nevertheless I see it as a (hopefully, as I don’t have one yet) very good all-rounder.
Edit: Oh… just remembered another one. The demos often show multiple boxes daisy-chained together for very specific use cases where this is needed. That’s something that is really nice to be almost built-in - again with a small form factor not needing much space. Not sure what the best use cases are. Maybe again AI, but could very well be 3D and video rendering - but not sure about that. Hopefully someone else can list a few of the best use cases where the framework desktop will show it’s strengths.
Edit 2: I think a lot of Gaming hype for this platform stems from the fact that the first available devices which launched recently are a Tablet and Laptop specifically intended for gamers. Again not the fastest and most capable devices, but quite novel performance in that small form factor and low energy consumption on Windows. Those devices are capped around 70W though, as cooling would be an issue in that small form factors. Personally love frameworks idea to just build a desktop and see whats possible. We’ll have IIRC 120W sustained and 140W peaks. That should make the desktop more capable than the already nice laptops.
Allocation of RAM is super cool, which is why I’d shoot for the option with the most RAM so that I’d be free to be creative in how I intend to use it. If it’s not great for rendering, I’m sure an AMD non-gaming GPU (the ones intended to use in rendering and stuff) would be a great addition. I intend to switch to AMD for good.
Depends on software support, not sure. But I guess that for many rendering tasks raw spread matters much more than being able to allocate lots of video ram.
