How “Unified Memory” Speeds Up Apple’s M1 ARM Macs

0
272
Apple

Apple is rethinking how components should exist and operate inside a laptop. With M1 chips in new Macs, Apple has a new “Unified Memory Architecture” (UMA) that dramatically speeds up memory performance. Here’s how memory works on Apple Silicon.

How Apple Silicon Handles RAM

In case you haven’t already heard the news, Apple announced a new slate of Macs in November 2020. The new MacBook Air, MacBook Pro, and Mac Mini models are using an ARM-based processor custom-designed by Apple called the M1. This change was long expected and is the culmination of Apple’s decade spent designing ARM-based processors for the iPhone and iPad.

The M1 is a system on a chip (SoC), which means that there’s not just a CPU inside the processor, but also other key components, including the GPU, I/O controllers, Apple’s Neural Engine for AI tasks, and, most importantly for our purposes, the physical RAM is part of that same package. To be clear, the RAM isn’t on the same Silicon as the fundamental parts of the SoC. Instead, it sits off to the side as pictured above.

Adding RAM to the SoC is nothing new. Smartphone SoCs can include RAM, and Apple’s decision to put the RAM modules off to the side is something we’ve been seeing from the company since at least 2018. If you look at this iFixit teardown for the iPad Pro 11, you can see the RAM sitting to the side with the A12X processor.

What’s different now is that this approach is also coming to the Mac, a full-fledged computer designed for heavier workloads.

RELATED: What Is Apple’s M1 Chip for the Mac?

The Basics: What Are RAM and Memory?

Corsair

RAM stands for Random Access Memory. It’s the primary component of system memory, which is a temporary storage space for data your computer is using right now. This can be anything from necessary files for running the operating system to a spreadsheet you’re currently editing to the contents of open browser tabs.

When you decide to open a text file, your CPU receives those instructions as well as which program to use. The CPU then takes all the data it needs for these operations and loads the necessary information into memory. Then, the CPU manages changes made to the file by accessing and manipulating what’s in memory.

Typically, RAM exists in the form of these long, thin sticks that fit into specialized slots on your laptop or desktop motherboard, as pictured above. RAM can also be a simple square or rectangular module that is soldered onto the motherboard. Either way, RAM for PCs and Macs have traditionally been a discrete component with its own space on the motherboard.

M1 RAM: The Discrete Roommate

Apple

So the physical RAM modules are still separate entities, but they are sitting on the same green substrate as the processor. “Big whoop,” I hear you saying. “What’s the big deal?” Well, first of all, this means faster access to memory, which inevitably improves performance. In addition, Apple is tweaking how memory is used within the system.

Apple calls its approach a “Unified Memory Architecture” (UMA). The basic idea is that the M1’s RAM is a single pool of memory that all parts of the processor can access. First, that means that if the GPU needs more system memory, it can ramp up usage while other parts of the SoC ramp down. Even better, there’s no need to carve out portions of memory for each part of the SoC and then shuttle data between the two spaces for different parts of the processor. Instead, the GPU, CPU, and other parts of the processor can access the same data at the same memory address.

To see why this is important, imagine the broad strokes of how a video game runs. The CPU first receives all the instructions for the game and then offloads the data that the GPU needs to the graphics card. The graphics card then takes all that data and works on it within its own processor (the GPU) and built-in RAM.

Even if you have a processor with integrated graphics, the GPU typically maintains its own chunk of memory, as does the processor. They both work on the same data independently and then shuttle the results back and forth between their memory fiefdoms. If you drop the requirement to move data back and forth, it’s easy to see how keeping everything in the same virtual filing cabinet could improve performance.

For example, here’s how Apple describes its unified memory architecture on the official M1 website:

“M1 also features our unified memory architecture, or UMA. M1 unifies its high‑bandwidth, low‑latency memory into a single pool within a custom package. As a result, all of the technologies in the SoC can access the same data without copying it between multiple pools of memory. This dramatically improves performance and power efficiency. Video apps are snappier. Games are richer and more detailed. Image processing is lightning fast. And your entire system is more responsive.”

And it’s not just that every component can access the same memory at the same place. As Chris Mellor points out over at The Register, Apple is using high-bandwidth memory here. The memory is closer to the CPU (and other components), and it’s just faster to access than it would be to access a traditional RAM chip connected to a motherboard via a socket interface.

Apple Isn’t the First Company to Try Unified Memory

An NVIDIA diagram from the early days of the company’s Unified Memory feature. NVIDIA

Apple isn’t the first company to approach this problem. For example, NVIDIA started offering developers a hardware and software solution called Unified Memory about six years ago.

For NVIDIA, Unified Memory provides a single memory location that is “accessible from any processor in a system.” In NVIDIA’s world, as far as the CPU and GPU are concerned, they are going to the same location for the same data. However, behind the scenes, the system is paging the required data between separate CPU and GPU memory.

As far as we know, Apple is not taking an approach using behind-the-scenes techniques. Instead, each portion of the SoC is able to access the exact same location for data in memory.

The bottom line with Apple’s UMA is better performance from faster access to RAM and a shared memory pool that removes performance penalties for moving data around to different addresses.

How Much RAM Do You Need?

Apple’s solution is not all sunshine and happiness. Since the M1 has the RAM modules so deeply integrated, you can’t upgrade it after purchase. If you choose an 8GB MacBook Air, there’s no increasing that device’s RAM at a later date. To be fair, upgrading the RAM hasn’t been something you could do on a MacBook for a while now. It was something previous Mac Minis could do, but not the new M1 versions.

The first M1 Macs top out at 16GB—you can get an M1 Mac with 8GB or 16GB of memory, but you can’t get any more than that. It’s no longer just a matter of sticking a RAM module into a slot.

So how much RAM do you need? When we’re talking about Windows PCs, the general advice is that 8GB is more than enough for basic computing tasks. Gamers are well-advised to bump that up to 16GB, and “prosumer” activity likely needs to double again for tasks like editing large, high-resolution video files.

Similarly, with M1 Macs, the base model with 8GB should be enough for most people. In fact, it may cover even the most hardcore of day-to-day uses. It’s hard to say, though, as most of the benchmarks we’ve seen take the M1 to task in synthetic benchmarks that push the CPU or GPU.

What really matters is how well an M1 Mac handles keeping multiple programs and a slew of browser tabs open at once. This doesn’t just test hardware, mind you, as software optimizations can go a long way toward improving this kind of performance, which is why there’s been such a focus on benchmarks that can really push the hardware. However, in the end, we’d guess that most people just want to see how the new Macs handle “real world” usage.

Stephen Hall over at 9to5 Mac saw impressive results with an M1 MacBook Air with 8GB of RAM. To get the laptop to start faltering, he had to have one Safari window open with 24 website tabs, another six Safari windows playing 2160p video, and Spotify running in the background. He also took a screenshot. “Only then did the computer finally grind to a halt,” Hall said.

Over at TechCrunch, Matthew Panazarino went even further with an M1 MacBook Pro rocking 16GB of RAM. He opened 400 tabs in Safari (plus he had a few other programs open), and it ran just fine, without any issues. Interestingly, he tried the same experiment with Chrome, but Chrome flamed out. But, he said, the rest of the system kept performing well despite the issues with Google’s browser. In fact, during his tests, he even noticed the laptop using swap space at one point, with no noticeable dip in performance.

When your PC runs out of RAM, it carves out available SSD or hard drive storage as a temporary pool of memory. This can betray a noticeable slowdown in performance, though not with M1 Macs, it seems.

These are just casual day-to-day experiences, not formal tests. Still, they are likely representative of what to expect for intense day-to-day use, and given the tweaked approach to memory, 8GB of RAM should be just fine for most people who aren’t opening browser tabs in the hundreds.

However, if you find yourself editing large, multi-gigabyte images or video files while also browsing a few dozen tabs and streaming a movie in the background all on an external monitor, then perhaps choosing the 16GB model is the better choice.

This isn’t the first time Apple has rethought its Mac systems and moved to a new architecture.

RELATED: Deja Vu: A Brief History of Every Mac CPU Architecture

READ NEXT

  • › How “Unified Memory” Speeds Up Apple’s M1 ARM Macs
  • › How to Change the Theme and Accent Color of Instagram DMs
  • › How to Record Shows and Movies on YouTube TV
  • › How to Get Cycling Navigation Directions in Apple Maps
  • › What Does “ITT” Mean, and How Do You Use It?