Switching Silicon: How Sketch performs on Apple’s new hardware

Earlier this month, Apple finally announced its new M1 chip, and gave us some mind-blowing numbers to demonstrate how fast its three new Macs really are. Naturally, we were excited to get our hands on a new machine to test these performance claims for ourselves — and see just how Sketch would run on this new hardware.

Measuring performance objectively across machines in an application as interactive as Sketch isn’t easy. Recording side-by-side operations produced some notable differences — browsing through complicated Sketch documents was at least 40% faster in our measurements, for example — and in an unscientific way everything felt faster. But we could never be sure whether other factors were at play.

But then we realized that we had another way to put the new Macs through their paces — a more objective, measurable way.

A Mac mini server farm

At Sketch, the Mac mini holds a special place in our heart. We operate a server farm of 19 Mac minis. Together, they process over 60,000 Sketch documents every day — for developer handoff, asset exporting, web previews, and our upcoming live collaboration feature. And as you can imagine, speed and performance make a huge difference here.

We wanted to know if the M1 chip would speed up our operations and ultimately make the experience of using Sketch even better as a result. To help us figure that out, we conducted three separate tests.

Measuring overall performance

First, let’s set the scene. For this test, we ordered the entry-level 2020 Mac mini, featuring an Apple M1 Chip with 8GB of RAM. Its purchase price is $699.

We’re comparing this machine with those that are currently running our render farm; the 2018 Mac mini, with 3.2 GHz Intel Core i7 processors and 32GB RAM. You can still buy this model today, starting at $1,299 for the same processor and 8GB RAM.

To test performance, we took a number of complex Sketch documents, loaded them up, then exported a few hundred images out of each document. This gives us a nice variety of ‘disk’ activity, memory usage, CPU usage and GPU usage to help us compare the machines.

And how did they compare?

The results were impressive and confirmed our less scientific observations from earlier. The Mac minis currently running our render farm are the top-of-the-line model from 2018, and they were consistently outperformed by 30-40% in our first test compared to the base-line model of the new Mac mini. That’s 30-40% more performance for almost half the price!

It’s worth noting that these tests don’t accurately reflect interactive real-life Sketch use (due to all the disk writing, for example), nor are they designed to test one isolated unit on the chip — like the CPU.

Measuring raw CPU

With that in mind, we decided to run another set of tests to get a better idea of how these two pieces of hardware perform from a raw CPU perspective. In this case we compared the results of a detach operation — we took those same large and complex Sketch files and recursively detached every Symbol instance in each.

Performing this task gave us a better idea of how these two machines compare based almost solely on CPU. And what did we find?

Here, the results are almost

Continue reading

This post was originally published on this site