Gaming Console Chipsets: PlayStation, Xbox, Nintendo Architecture Deep Dive

I used to believe that a console was no more than a box, which played games. After that I spent three weeks in a rabbit hole attempting to comprehend why my PS5 requires eight seconds to boot a game as compared to fifteen on the Xbox of my friend. It turns out that the chipsets in these machines are essentially miniature supercomputers and the differences between these two? More entertaining than any console battle on Twitter.

I discovered this after scouring the spec sheets, architecture diagrams, and more Reddits that I had ever discovered.

What Actually Makes a Gaming Console Tick

Gaming consoles are not mere small-sized PCs. These are personal-built computers in which every chip, every circuit is there to serve a single purpose and that is playing games. And even in the comparison of a PS5 and an Xbox Series X and a Nintendo Switch, you are looking at three entirely different philosophies hardcoded into silicon.

The Brain: Custom AMD APUs in PlayStation and Xbox

Both the PlayStation 5 and the Xbox Series X use custom AMD APUs (Accelerated Processing Unit). APU – The trick is to consider an APU a CPU/GPU co-located on the same chip, just as much, – no latency in communications, no bottlenecks. Sony and Microsoft did not simply subcontract, it also collaborated with AMD design consoles, and came up with their own chips tailored to the consoles.

An 8 core Zen2 processor with a 3.5 GHz frequency is offered as the custom AMD APU in PlayStation 5. It shares the same architecture as desktop Ryzen processors in 2019, only desktop optimized. The GPU side? At that point it gets interesting. Sony settled on 36 RDNO 2 true compute units running at a furious 2.23 GHz, and provided 10.28 TFLOPS of raw graphics horsepower.

What caught my eye here is that PS5 has less compute units than Xbox but runs the units higher. It is as though picking a smaller engine that runs faster as compared to a bigger motor that runs more even.
Xbox series x adopted a different strategy.

Microsoft had equipped it with 52 RDNA 2 compute units at 1.825 GHz and reaching 12 TFLOPS. That would be about 17 percent of put-on muscle on paper. The 7nm TSMC fabrication process the same technology as the smartphone chip is used on the two consoles, but design priorities are entirely different.

The two systems have identical 16GB of unified GDDR6 memory, but it is distributed asymmetrically with 10GB of 560 GB/s in the areas the GPU is required the most, and 6GB at 336 GB/s in those least demanding. PS5 is less complicated with 448 GB/s all the way across. I looked at benchmarks and to the truth? An actual game between the two is almost the same.

The Budget Player: Xbox Series S

The Xbox Series S is the scapegoat of Microsoft that is my weird middle child. It is operating with the same CPU as Series X, but has a 4 TFLOPS graphics core, or one third the graphics power. Else, it nonetheless addresses 4K gaming with aggressive upscaling and reconstruction measures.

I did try it with Forza Horizon 5, and, indeed, you can tell, though, with pixel-peeping. But for $299? It is playing the same games at 1440p or scaled to 4K, and even the vast majority who play on a 55 inch TV at a distance of ten feet will not see the tradeoffs.

The Outlier: Nintendo’s ARM-Based Architecture

Then there is Nintendo Switch that ripped the entire rulebook. Nintendo did not pursue brute strength but instead became portable with an NVIDIA Tegra custom CPU – a quad-core ARM Cortex-A57 CPU with a Maxwell-based graphics core.

The point is as follows, Maxwell architecture was released in 2014. The Switch is based on 20nm production process whereas PS5 and Xbox have 7nm. On paper, it’s ancient tech. In practice? It emulates Breath of the Wild and Tears of the Kingdom perfectly well due to the fact that Nintendo tuned everything to that particular chip.

Portable and docked modes give approximately 0.5 and 1 TFLOP Switch respectively. That is a tenth the power of a PS5. However it is pocket sized and has a battery life of up to 4-9 hours. Different game entirely.

If you’re curious how custom silicon is reshaping other industries, check out our deep dive on Automotive Chipsets for Autonomous Driving, where similar architecture trade-offs define self-driving capabilities.

Why Custom Silicon Beats Off-the-Shelf Parts

Makers of console could have simply used the standard PC parts. AMD sells desktop GPUs. Intel makes CPUs. Why not stick to custom designs, then?

Bare Metal Optimization

There is no operating system bloatware, no background programs, no operating system overhead. The release of a PS5 game gives it physical access to all transistors. Developers are able to push the performance that would not have been possible on PC because they have known the exact hardware they are writing to.

I spoke to one of my friends in the industry who was working on a cross platform title. He claimed that it had cost three weeks to optimize PS5. PC configurations optimization? Three months and they continue to get reports of bugs with given combinations of GPU-drivers.

Custom I/O and Memory Architecture

The key strength PS5 has is not the GPU but the I/O architecture. Sony created a purpose-designed SSD controller that provided 5.5GB/s raw bandwidth (8-9 GB/s compressed). That is technical but this is what it means and how it works in reality as the games load in approximately less than two seconds with grand loads of levels.

I tried Ratchet and Clank: Rift apart, and it transports you through the dimensions in the middle of the game. On PS4, that would have been a cutscene that is actually a loading screen. On PS5, it occurs immediately since the SSD loads assets at a rate that was unattainable to the old hard drive, which was only able to initiate loading.

Similar is done by Xbox, its Velocity Architecture custom NVMe SSD (2.4 GB/s raw, 4.8 GB/s compressed) plus a feature known as Sampler Feedback Streaming, which loads only the texture data a graphic card in fact requires. Clever stuff.

To include the wider picture in the situation of how chipsets make these innovations possible, our Complete Guide to Semiconductor Chipsets further dissects the foundation on an industry-by-industry basis.

The Performance Evolution: Then vs Now

I dug up my old PS4 to compare. The comparison is sincerely absurd.

PS4’s GPU: 1.84 TFLOPS
PS5’s GPU: 10.28 TFLOPS

It represents an increase in raw compute power by 5.6 times. Back but TFLOPS only tell part of the story. PS5 also introduced the use of hardware-accelerated ray tracing (real-world lighting and reflections), variable rate shading (GPU devote power to where it counts), and that preposterous SSD that I referred to.
Load times were reduced to 2-8 seconds as compared to 30-60 seconds.

Frame rates were increased to 60fps (or even 120fps) rather than 30fps. Scene adjustment increased to 4K natural.

Launch Games vs Later Optimization

Here is one of the unexpected details: the games launched at the initial stages of a console barely break its potential. Time is required by developers in getting acquainted with the tricks of the architecture.

Look at PS3’s lifecycle:

• 2006 launch titles looked rough
• The Last of Us of 2013 had the appearance of a different console.

Same hardware. Markedly just improved optimization within a period of seven years.

PS5’s already seeing this. Compare the 2020 launching of the Spider-Man: Miles Morales with that of Spider-Man 2 (2023). Similar console, with more impressive graphics, performance flow. Devs worked out the method of getting the chipset to work harder.

The Trade-Offs: Resolution, Frame Rate, and Ray Tracing

There are compromises of every console game. You just can not charge it all at the same moment, silicon will not permit.

Most modern-generation games provide modes:

• Performance Mode: 1440p-1800p up scale to 4K, 60fps, ray tracing off.
• Quality Mode: Native 4K, 30fps on, ray tracing on.
• Balanced Mode 1440p, 40fps, selective ray tracing.

I did play Cyberpunk 2077 in all the three modes on PS5. Performance mode was more enjoyable to feel, Quality mode was gorgeous but jerky. Picked Performance, never turned his head back.

The bottleneck is now cared of by ray tracing. It is incredibly graphics card intensive It requires a graphics card that is amongst the more compute units that today’s chips can hardly manage at 60fps to generate light rays in a graphics card simulating a three dimensional environment. The added TFLOPS of Xbox Series X are an average aid though both consoles lose frame rate or resolution when using ray tracing.

What’s Next: The Future of Console Chipsets

PlayStation 6 and next-gen Xbox have already leaked, and in the event that the rumors are true, another generational jump is possible.

Process Node Shrinks

PS5 and Xbox series X contain 7nm chips. Next-gen should not be much more than 3nm, or smaller, and it implies:

• Additional transistors in the same area.
• Lower power consumption
• Less heat generation
• Higher clock speeds

The same reason is that your iPhone also gets faster every year, yet it does not get any larger.

AMD’s UDNA Architecture

Both Sony and Microsoft are set to implement the future UDNA architecture in which it is a single design that combines the consumer and professional level graphics technology at AMD. Preliminary benchmarks are showing 20% rebate in rasterization (standard graphics render) and 2 times performance of ray tracing than existing RDNA 2 chips.

In case it is true, it then means that full ray tracing at 60fps will be real. The current-generation consoles have trouble reaching 30fps when ray tracing on, and should be smoother with next-generation.

AI Upscaling Goes Mainstream

PS5 Pro came with PSSR (PlayStation Spectral Super Resolution), Sony-made proprietary AI upscaling. It downscales games to lower resolutions (resolutions of 1440p-1800p) and uses machine learning to recreate almost 4K quality images.

The next-generation consoles will presumably integrate this technology into the chipset itself. 1440p internal, 8K external rendering. Your television is 8K and the graphics processing unit only needed to work with 1440p. Groovy hardware optimization.

Final Thoughts: It’s More Than Just Numbers

Whether it is weeks of comparing specifications, benchmarks, and practical testing, this is my opinion on the issue: raw numbers (TFLOPS, clock speeds, memory bandwidth) are irrelevant, but rather how all that fits together.

Technically, PS5 is weaker than the Xbox Series X in terms of its GPU, however, its SSD is faster and the system works with tight integration, which implies that most games run exactly the same. It is absurd that Nintendo switch is worse than both in terms of power, but it sold more than the two together, because game library and portability are more important than teraflops.

When you are picking a console, it does not matter whether you are obsessed with chipset specifications. Ask yourself:

• What exclusives do you want? (Spider-Man, Zelda, Halo?)
• Am I a portable person or an electric power person?
• What’s your budget?

The cards mounted within these machines are design miracles in-house silicon built to do hardware gaming. However, they are merely an instrument at the end of the day and allow great games.
And honestly? No matter which of these consoles you come to, it will blow your mind with the last-gen hardware. Choose the ecosystem that you like and have the fun.