- Deepseek R1’s parameters of 671 billion run smooth on the M3 Ultra’s Unified Memory
- Apple’s Mac Studio shows that AI workloads do not require expensive, power-hungry GPU clusters
- M3 Ultra consumer under 200W, far less than traditional multi-gpu AI setups
Apple’s Mac Studio with M3 Ultra Chip has shown a capacity that no other personal computer can match, and runs the Deepseek R1 AI tool with 671 billion parameters completely in memory.
A test of YouTube correction reader Dave2D showed despite using a 4-bit quantized version of the model, it preserved its full parameter number and performed smoothly.
The Deepseek R1 model, a hefty 404 GB of storage and high-bandwidth memory, typically found in GPU VRAM, is usually run on multi-GPU setups that distribute treatment on multiple high-end graphics cards.
A unique holding: runs Deepseek R1 in memory
However, M3 Ultra’s overall memory system instead of relying on external GPUs uses its 512 GB of Unified Memory to store and process the AI model in a way that no other personal computer can.
Although MACOs impose a standard VRAM limit, Dave Lee manually increased it through the terminal to allocate up to 448 GB to AI treatment, eliminate memory bottlenecks and reduce the need for multiple components to streamline AI performance on a single system.
One of the most striking aspects of this test was the M3 Ultra’s power efficiency when it consumed less than 200W while driving Deepseek R1.
The ability to run such a demanding AI model without a multi-GPU setup challenges the industrial standard that is dependent on advanced NVIDIA and AMD graphics cards, as the best workstations and server farms typically use GPU clusters that consume large amounts of electricity.
Apple’s overall memory architecture enables significant power savings by sharing M3 Ultra’s memory pool across CPU and GPU workloads, as opposed to conventional PC setups where VRAM is separated from system memory that maximizes bandwidth while minimizing energy consumption.
Apple’s Mac Studio, launched with M3 Ultra Chip, has up to a 32-core CPU and an 80-core GPU, making it one of the best LLM work stations and one of the best video editing computers.
Via WCCFTERCH
