Hacker News Reader

Those years were amazing because every new console generation had a wildly different artisanal design. NES to SNES to N64 to Switch, Jaguar, Sega, PS to PS2 to PS3, Xbox... and then at some point they all turned into disguised PCs.

Up until the early 00's specially designed components could get more performance/$.

Now it's impossible to beat commodity SoCs on that front.

The arcade industry also went through a couple waves of that, with one major shift to console-based hardware in the mid-90s and another to PC-based hardware in the mid-2000s. Namco made a PS3-based system board (System 357, notably the original platform for Tekken 6), but I'm not sure if anyone else did.

Hey now, Nintendo’s bucked that trend - the Switch is a disguised tablet instead.

I like to think of it as a disguised original Tesla Model S

not sure which is those that's more of an insult to

Oh you mean the disguised nvidia shield?

IIRC, the original PS3 design was rumoured to have dual Cell processors. It didn't meet performance targets, so the Nvidia GPU was wedged in quite late on in the development cycle.

I think the story was a bit different. Sony had envisioned the the final Cell chip to be more powerful than what they launched (4 PPEs and 32 SPEs @ 4GHz versus 1 PPE and 8 SPEs @ 3.2GHz) and they though that would be enough to render graphics with it like they did on the PS2, and only realized later in development that it won't be enough so they went to Nvidia and asked for a discrete GPU.

Also, I think game developers balked at having to implement their own GPU in a hard-to-program unique architecture.

I wonder how many PS3 games ignored the Cell processor and just developed for the CPU and GPU?

> Also, I think game developers balked at having to implement their own GPU in a hard-to-program unique architecture.

It would have been a library.

> I wonder how many PS3 games ignored the Cell processor and just developed for the CPU and GPU?

Essentially none.

> It would have been a library.

A library can be convenient to use or not. I had one afternoon with the Cell and it wasn't a pleasant one.

I think I heard the games Valve published for PS3 didn’t really use the cell stuff

At lot of early games were something like that. They’d mostly ignore it or maybe use one SPU or shove tasks on them that really didn’t take advantage of their power at all. Just leaving a lot of performance untouched.

They absolutely did. The PPEs were far too anemic to run anything like halflife on their own.

Gabe Newell famously sounded off about how useless it was and wouldn't transfer to any other hardware, but he ended up being wrong. The architectural model forced on you by the SPEs is the model that the industry (not just games but any compute intensive work) has embraced for the heavy multicore but coherent system world.

Rather than the "thread per vague activity" model, we've embraced a one thread per core with a work stealing scheduler walking a DAG of chunked compute tasks.

For a modern view of this, this is exactly how Rust's rayon library works.

The whole thing reminds me of a lot of the complaints about the N64, a lot of which ended up being the fact that DRAM was no longer a single cycle away like it was for the SNES. Yes cache conscious code is more difficult to write, but that memory hierarchy architecture was more a harbinger of the new world rather than a one off weirdness of a single console.

> The PPEs were far too anemic to run anything like halflife on their own

To be fair the PS3 port of the Orange Box runs like absolute crap in many instances. Although I believe it was developed by EA, not Valve.

Some games are built the way you've described with job threads and queues, but regrettably most are built with a main thread, a render thread, and sometimes use other threads for some compute.

> I think I heard the games Valve published for PS3 didn’t really use the cell stuff

Maybe - the Orange Box at least was outsourced to EA and the performance was reportedly not great. I think Portal 2 was done in-house though and I don't remember anyone complaining about it.

There's an interesting book called The Race for a New Game Machine: Creating the Chips Inside the XBox 360 and the Playstation 3 by David Shippy and Mickie Phipps on the development of the Cell processor.

I can't recall if if had all the details regarding adding a GPU to the PS3, but I remember it had stories about the awkwardness of having the IBM Cell team also working on Microsoft's Xenon CPU for the 360, which IIRC used modified Cell PPE cores.

> Microsoft's Xenon CPU

This was another interesting take on the CPU for consoles. The 360 was born with a GPU, however.

AIUI it's basically that the plan was similar to the PS2, where you have a "smart" CPU (Emotion Engine / Cell) and a "dumb" rasteriser (Graphic Synthesiser on PS2), with the Cell SPEs taking the role of the vector units on the EE for things like transform + lighting. But the rasteriser chip project failed.

Then they tried the 2 Cells and software rendering approach, to try and keep it as an "in house" solution.

Finally they went to Nvidia and we got the final Cell + RSX solution.

This is what I've heard as well. They started with one cell, when it failed internal expectations they switched to two cells, then when they heard the rumors of what Microsoft was doing with the Xbox 360 they ran to Nvidia to get a real GPU.

Eh, the Cell would have been a fantastic processor if it weren't for hitting the end of dennard scaling like a brick wall.

In the design phase they were expecting almost 5ghz at launch, and to hit around 10ghz at the end of the product lifecycle. When the process guys came back later to let the architects know the reality we've sort of now all internalized (high end consumer electronics can expect ~3ghz with good cooling solutions) a lot of simple but fast pipeline designs stopped making sense. These are the exact same pressures that killed off the Netburst architecture too, making intel go back to a PIII design and make it smarter rather than clock markedly faster.

Nah, SPEs are a bad idea at any frequency.

What specifically about them?

Mostly not having caches.

The TCMs were cool. But unless your work fits that...

The SPEs only had cache! (And, a ton of registers.) What they were missing was RAM ;D

But, even that wasn’t as bad as it was made out to be. People rightly moan about the awkwardness of asynchronously moving data between main RAM and the SPE memory. What they don’t often mention is that the latency of those moves was about 500 cycles —the same latency as a cache miss on the PPE CPUs!

So, which was worse: implicitly waiting 500 cycles all over the place? Or, explicitly scheduling 500 cycle waits at specific points? Unsurprisingly, everyone preferred the first option :P

Nothing prevents them from having their own RAM - ISAs and implementations can be extended.

The market for the Cell never materialised in the end, but it could have turned into something a lot more interesting.

> but by the time it launched on the market in 2005, X86_64 was the new undisputed king of CPU performance

PS3 Cell could do about 154 GFLOPs across it's 7 cores (6 SPUs and 1 PPU, excluding the GPU), while Xeons were in the low double digits. What really was awkward about the Cell architecture was the DSP-style nature of it, which works great for something like shaders or audio processing with a clear pipeline, but not for general purpose computation. It was really a successor to the PS2 in this regard, where data flowed between specialized and memory constrained processors using a DMA.

> Sony nor anyone else expected the consumer CPU and GPU industry to evolve so quickly in that time

PS3 used an NVidia GPU. Though there was some clunkiness around sharing memory between the Cell and GPU compared to the UMA on PS4, but not especially so compared to a graphics card on a desktop machine.