Killzone 3: A Eulogy to PlayStation as a technology company.

PlayStation was once a technology company. The PS1, PS2, and PS3 all featured custom hardware made in Sony development labs. PS1 and PS2 featured RISC controllers, very similar to contemporary arcades. PS3 featured something different and very special: Cell architecture. 

Sony's proprietary Cell Architecture featured a primary CPU and several secondary CPUs that worked in parallel. Parallel processing is common and what we find in Graphic Processing Units, or GPUs. The PS3 wasn't just a giant GPU though. GPUs are limited in processing capability. The PS3 featured several full CPUs capable of processing simple and complex equations. While it required custom engines to use, it could handle anything from rendering to AI. 

The potential of Cell to do real work manifested in the Fold@Home program. Outside of this very specific research project, the PS3 was unfortunately treated mostly as a giant GPU. The average consumer is not likely to notice skill-scaling AI, but is very likely to notice that something "looks good". Even as a giant GPU, the PS3's Cell had shining moments of true glory. 

It is important to note that the PS3 was running Killzone 3 with 256MB of DRAM and 256MB of VRAM, with the assistance of a 550 MHz custom Nvidia GPU. At most the system was using at 400 watt power supply. Outside of Cell, this machine was barely more than a business class Dell tower. 

The PS3 featured a primary CPU and 8 secondary CPUs. 7 of the secondary CPUs were used due to quality control (not every chip made is an ideal chip, so making 8 allowed Sony to pick 7 that worked the best). One of the seven was devoted to the OS, and the other 6 were open to developers. PS3 Cell had a running efficiency that the common computer architecture of X86 will never achieve. 

Parallel processing allowed the PS3 to reduce heat generation. The choke for current X86 systems is heat. X86 is inefficient because it passes a lot of electrons back and forth to complete a single task. It makes X86 modular, but it also makes it inefficient. As noticable with new high-end GPU cards melting their power cable harness, X86 generates a massive amount of heat because it requires a massive amount of power in relatively small packages. Cell did not have the power requirements. The later PS3 slim models used PSUs as low as 150 watts.

Imagine a 150 watt machine outputing 30 frames like this a second.

If Sony had continued to use Cell in the PS4, they wouldn't have had the chip issues with the PS5. Cell doesn't need the power density that X86 does, so it doesn't need tiny 5nm chip tech that is subject to higher manufacturing errors. With Cell they would have binned less chips. With Cell they could have used older, more stable IBM facilities. The switch to X86 costed Sony a lot in performance and manufacturing. So why did they make the change?

Sony switched to X86 because Cell is difficult for programmers to code for. Sony and IBM could have and should have made better tools. Instead they both ditched the superior technology and fell back on the old stuff that developers were comfortable with. Third-party developers, along with Mark Cerny, convinced Sony to build a budget PC. 

Cell was a solution that would have proven to be far more valuable today than it was when Sony helped make it. Unfortunately PlayStation is a services brand now and no longer a hardware brand. The future of PlayStation looks more Vaio than Walkman now. There is very little chance that Sony Group Corp will go back to Cell, or even go back to RISC-V at this point. We are likely to look back on the PS3 as the system that was ahead of its time and PlayStation as an unfortunate victim of Yoshida's mismanagement. With the division and separation of Sony Corp (the hardware part of Sony Group Corp) (look for the July 27th, 2022 announcement) and the growing competition in the industry, it is likely that the IP house that is now PlayStation will be an acquisition for Apple or Amazon.