Comparing Threadripper and multiple ARM-based Mac Minis highlights differences in performance, design, and use cases.
Comparing Threadripper and multiple ARM-based Mac Minis highlights differences in performance, design, and use cases.
Hey there, I’m curious—has anyone tried linking several ARM-based Mac Minis to tackle a big project together? Like creating a detailed 3D render in Blender, as people used to do? I’m not very familiar with Macs but I saw some photos and heard stories about it. And how does this relate to the Threadripper? Let me clarify... In my region (Turkey, currently at the bottom of the global rankings), a full Threadripper 3970x setup costs about the same as four upgraded Mac Minis with 16GB RAM, 10Gb Ethernet, plus a slightly upgraded Macbook Air (16GB RAM and 512GB storage). I was thinking: Have others compared the performance of multiple Mac Minis to that Threadripper? What advantages could there be? For instance, power usage, having several devices ready for different tasks, or using a remote desktop to manage work from home while you’re doing presentations. Or can you combine the laptop’s power with it when needed, and still upgrade later—adding more M1 or newer models to boost overall performance? What are the downsides? The raw power will be lower than a Threadripper, but by how much? Will it balance out energy use, cooling, and maintenance costs? Also, are there restrictions from macOS on ARM, or is Windows unavailable on these systems? Or is it really possible? I don’t have much experience or funds to test this myself. If anyone has insights or tries it, it would be a great experiment!
You'd need a proper setup method to link these devices together, rather than just plugging them in. It’s more complex than simple cabling between a few units.
The 4 m1 mac will feature 32 cores and 64GB of RAM. One mini scores 7401 (geekbench) multiplied by four, totaling about 29.604 points after accounting for any penalties or extra latency. The typical power draw for a mac mini ranges from 10 to 20 watts, making the total around 80 watts. A score of 3970 on GeekScore is impressive, but it supports up to 256GB memory and can handle up to 350W just for the CPU. In terms of efficiency, the minis clearly outperform others. However, depending on the software, Threadripper might offer better results.
Discussing the energy use, running the full power mode on both GPU and CPU simultaneously consumes about 35 W for the whole setup. This means it should use significantly less power compared to just the CPU running at maximum speed. While I’m not an expert in clusters, it seems feasible to distribute a heavily parallel workload across this M1 mini system. You’d need appropriate software—like chunked CFD simulations sent to clusters—to make it work. For suitable tasks, you could achieve solid performance while keeping power consumption low if you manage to assemble a M1 Mini cluster effectively.
the main advantage of threadripper lies in having all cores on the same chip, which reduces inter-die delays significantly compared to splitting tasks across multiple nodes in a cluster with only 10g of communication. this concept is intriguing—using a portion of the cluster as an external compute node would need specialized software or substantial custom development in your clustering setup.
I first considered using Blender and wondered how it manages parallel computing for rendering. It seems Blender and similar modeling tools might already have built-in solutions for these scenarios. Most films are processed on farms, meaning multiple computers connected together in a grid. Threadripper is a high-performance desktop CPU, though it has limitations compared to regular Ryzen processors. I’m not sure about the custom modifications needed. It could be useful, but if you remove the laptop from the workflow—before any changes take place—maybe the workload can be balanced across several Mac Minis. By the way, I’m using an older machine with a 2600K (Sandy Bridge i7) at home and a Ryzen 2600X at work. I use remote desktop to edit the model and render it on my main computer, allowing me to monitor progress from anywhere. For instance, in Blender I render two-thirds of the frames locally while relying on one-third from home. This approach reduces overall time, even slightly. These experiences have made me reflect on this setup and its potential for future expansion with more Mac Minis. Edited June 2, 2021 by Baradh – many typos, sorry!