Mark Cerny Shares PlayStation 5 Boost, SSD Speed, Tempest Audio Details

Sony's Mark Cerny, lead system architect of the PlayStation 5, has shared more details on the new console and some aspects of its subsystem behavior. Some of the new details confirm our speculation from March, while others shed light on previously unexplored aspects of the platform.

How the PS5's Boost Mode Works

One of the major distinctions between the Xbox Series X and the PlayStation 5 is how they handle clocking. Microsoft has made much of the fact that its parts run at static clock speed, while Sony wants to talk up its atypical boost scheme. Where standard boost modes vary from part to part, the PlayStation 5 doesn't work this way -- Sony doesn't want different consoles to offer different levels of performance.

On PCs, boost is controlled by temperature and current readings off the SoC, but these are, in turn, impacted by factors like part binning and ambient temperature. Sony's boost mode doesn't take these factors into consideration. Instead, CPU and GPU clock rates are determined based on monitored system activity. Instead of monitoring the performance of your specific SoC, the boost model Sony deploys is based on a model SoC. Every console uses the same model for how and when to boost frequencies, keeping performance identical even when air temperature or power draw might allow some chips to push higher. This keeps performance identical, regardless of ambient temperature.

Cerny specifies that developers don't want boost frequencies that are available for a short period of time before the system drops back to lower frequencies. Sustained targets are much more useful, and programmers will have detailed reports on CPU / GPU power consumption and frequency. They'll also have the opportunity to improve performance by optimizing for power efficiency -- being able to hit the same frame rate on less power will allow the GPU to maintain higher clocks for longer periods of time. Cerny also emphasizes a point that we've discussed several times in the past, namely that the relationship between frequency and power consumption is non-linear. According to him, dropping the CPU or GPU frequency by 10 percent reduces power consumption by an average of 27 percent. Radeon Nano is the clearest example we've ever seen of how binning and frequency reduction can dramatically improve power efficiency, and that GPU didn't achieve a ratio as high as what Cerny claims for the PlayStation 5.

Sony's Proprietary SSD

The Sony SSD design details confirm some of our previous ideas and actually speak to an underlying concern about how much RAM is available on these new consoles. In the past, consoles have typically dramatically increased RAM loadout. The PS1 had 2MB of RAM, the PS2 had 32MB, the PS3 had 512MB (split 256MB between CPU and GPU) and the PS4 had 8GB. By historical standards, the PS5 should be a 128GB console. Instead, it packs a relatively svelte 16GB.

The function of the onboard SSD is key to how Sony is driving performance forward without relying on huge amounts of RAM to do it. According to Cerny, one reason why console storage requirements have ballooned is that assets are copied to the hard drive in so many different locations to speed the data loading process. For the Spider-Man PS4 game, Insomniac copied any asset used more than 400x into RAM, which obviates the load issue but increased RAM requirements.

One subtle point Cerny hasn't addressed is whether games will run better on Sony's custom SSD as opposed to any third-party storage the end-user attaches to the system. The company has spent time talking up its custom firmware controller, but it can't have that much special sauce baked into it. What we know is that Sony will validate specific PCIe 4.0-compatible SSDs to work with the PS5, but this suggests drives will need to hit fairly advanced performance targets rather than being built with specific Sony-compatible controllers.

Tempest Audio Processing

Finally, there's Tempest engine, which drives the PlayStation 5's audio system. I'm not even sure how much detail to devote to this conversation because, if I'm being honest, it doesn't seem as if very many people care about audio in gaming all that much. Those of you who actually do, please don't beat me to death with $10,000 audio cables -- the fact is, we're a minority. Most people don't experience gaming through expansive 5.1 or 7.1 speaker setups with a top-notch sound system. AMD has talked up advanced audio processing before, both with its Hawaii and Fury line of GPUs and as part of the Sony and Microsoft console launches back in 2013, and for all the text lavished on it, the number of players who tap these capabilities seems to be low.

Sony is making a huge push into positional audio with the PlayStation 5, which expands on the advanced capabilities enabled by PlayStation VR. The PS5 will use Head-Related Transfer Functions for positional audio. The sound you hear won't be impacted solely by the quality of your headphones -- the PS5 will calculate an HRTF for delivering sound based on the specific size and shape of your head and ears. Improved 3D spatial algorithms allow for sounds to be positioned far more precisely in space. All of this is handled by the Tempest engine, which is actually a repurposed AMD compute unit designed specifically for audio processing.

These features are also supported on external speakers to some degree, though how much will depend in part on what kind of sound system you use. Eurogamer has a great deal more information on the Tempest engine, and some details on the other points that we only lightly touched on.

Thus far, the Xbox Series X looks to be more powerful graphically than the PlayStation 5, but we don't know how large the gap will be or what the price difference between the two consoles is. Without that information, it's hard to say much about how they'll likely compare with each other.