What's wrong with a modern ThinkPad? I know the Lenovo ones are a step down from the IBM days of yore but a T or P series are still sufficiently ruggedized designs that aren't crap by most metrics. They're a bit chunkier than a MBP, but the X series ThinkPad comes close in form factor. I like your Fubuki avatar btw lol
What about framework laptops? have you tried one?
That's because its using what's basically a phone SoC. New window laptops with ARM also have much higher battery, but very little support.
And also very little performance. Qualcomm only make garbage. M-series Apple Silicon is very distinct from A-series Apple Silicon, they are genuinely computer processors rather than just big phone processors. ARM windows laptops on the other hand genuinely do just use phone SOCs, you can often find the same processors used in Android “flagship” phones.
They are definitely not using a phone SoC. The design is similar and they use the same microarchitecture for their processor cores but the M-series chips have substantially more cores, more memory, and run at higher TDP. It's like saying a 12th gen i9 is a 'chromebook chip' just because there are Chromebooks that ship with Alder Lake i3s.
Was, the X Elite is actually beating apple chips at many benchmarks. It hasn't been the exact same than mobile chips since the 8cx, unlike previous models which were literally a rebranded mobile snapdragon, like if apple had simply rebranded an A SoC as a M SoC with no changes at all. The real problem is still the lack of ARM support for windows, thats one advantage of apple's vertical integration, tho I heard ARM mac sales are taking a dive and that could hurt 3rd party support.
So still a phone SoC just bigger, its not a custom design for PC use from the ground up. If you want to see real big ARM chips you have to look at Maia and the datacenter stuff nvidia makes.
Where? all I can find are the small boards and USB accelerator that been out for years, where's the big stuff?
The modern A16 and M-series cores were designed for PC use from the ground up. It's just that you can scale them down just as well for mobile usage. I don't think a phone needs to support x86 memory ordering but that functionality is baked in to M-series SoCs to support x86 translation.
Also this ignores the massive spike in sales that occurred with Apple Silicon Macs. Of course revenues declined - pretty much everyone that could have wanted an Apple Silicon Mac bought one between 2020 and 2022.
I mean only the small m.2 boards. There are a bunch in stock at Mouser. I didn't expect the more expensive stuff to be in high demand among hobbyists.
Was writing something but @LegoTugboat above already did.
Dunno, apple had to cancel mfg orders so clearly they were expecting more sales.
So what can you do with it? can't see it being more useful than a GPU. From what I see you can't even run an LLM since it lacks on-board memory and with only x1 using your PC's RAM its gonna be sloooooow.
Most of that stuff, like the Intel NCS2 I use is for 'edge computing'. In my case I run YOLO to do object detection on the cameras outside my house. So, robotics, vision, maybe some speech recognition.
Noice, so it works better than OEM cameras object detection?
I take you are using the rpi camera modules then?
According to the docs, it's made to work with Tensorflow Lite models so I guess you can train and compile TF lite models and run inferencing on the TPU. I'm simply using it for Frigate NVR.
I have no idea how well the OEM stuff works, I just turn it off, my workflow is:
Is there a list of projects/software build for these TPUs? the list on coral's website its like 4 things and all very basic.
This sounds a tad overengineered, are you doing all this for privacy reasons or just because you felt like it? there are no-subscription OEM cameras with AI now (well, one brand).
Finally, my years of identifying traffic lights for google are bearing fruit.
