The Lounge is rated Safe For Work. If you're about to post something inappropriate for a shared office environment, then don't post it. No ads, no abuse, and no programming questions. Trolling, (political, climate, religious or whatever) will result in your account being removed.
The world's largest chip, weighing in at an unbelievable 400,000 cores, 1.2 trillion transistors, 46,225 square millimeters of silicon, and 18 GB of on-chip memory, all in one chip that is as large as an entire wafer. Add in that the chip sucks 15kW of power and features 9 PB/s of memory bandwidth:
The system measures a mere 15 Rack Units tall, or roughly 26", so three can fit in a single rack. Given the performance, that's an incredibly compact package: It would take a 1,000-GPU cluster, which consumes 15 entire racks and half a megawatt of power, to match the performance of one CS-1 system. That's because a single Cerebras chip has 78 times more cores, 3,000 times more memory, and 10,000 times more memory bandwidth than a single GPU. It also has 33,000 times more bandwidth (PB/s).
«One day it will have to be officially admitted that what we have christened reality is an even greater illusion than the world of dreams.» Salvador Dali
They compared its 400k cores to a GPU with 5k cores, so they are probably continuing the bad tradition of calling SIMD lanes "cores", IIRC that's something GPU vendors started (specifically nVidia and their "CUDA processors", AMD likes to call them "stream processors" which is fine I guess?). Or maybe it's even counting some more specialized component such as 8-bit multipliers (important in deep learning) or whatever, who knows, actually if someone knows please explain.
Either way it's a beastly processor, actually either way it's the same processor, I just think their marketing material is deliberately misleading about its internal organization.