Hacker News Reader

Comment by ppseafield | Hacker News Reader

For the PlayStation 3, another thing driving Linux's removal was that organizations were buying PS3s to use in data centers. But Sony sold them at a loss (IIRC over $100/unit loss, even being sold at $600 twenty years ago) and intended to make it up on revenue from games sold. They also didn't have nearly enough at launch because they were using a brand new processor (designed with help from IBM) and a brand new Blu-Ray disc format. That slowed down their initial production, and it had to launch in time to compete with the Xbox 360 (Pre-Christmas 2005), which also had a HD-DVD attachment.

Notably, the PS3 at launch was actually cheaper than some dedicated Blu-ray players, so some home theater enthusiasts bought them (few to no games), eating into Sony's revenue even more. It took several years for the PS3 to be profitable, but it won the format war.

View on HN

They should have sold a datacenter version, without the optical disc, with a copy of AIX and/or BSD, with an OS license of $200 (or at least under $1,000).

It doesn't look like a workstation was ever produced, just blades and embedded.

https://en.wikipedia.org/wiki/Cell_(processor)

A hypothetical Datacenter version might have skipped the optical drive, perhaps cutting down on their losses. At launch I doubt they would have had enough spare processors to build them without cutting into their PS3 production capacity.

Additionally many game studios reported difficulty in writing for the PS3's processor compared to the Xbox, which had gone with a processor pretty similar to most PCs at the time. So by the time they had ramped up the capacity to make enough to meet PS3 demand, there were several years of worse/less performant PS3 versions of games developed for multiple platforms, making Sony's console look worse. It would have been a tough sell internally for the PS4 to use a custom processor architecture again, so they opted not to. This must have soured Sony's opinion of the Cell processor overall.

Sony should have realized that the Cell/Power was needed in great quantity, and insisted upon a second supplier. Motorola started making the PowerPC 601 in 1992, so a secondary foundry was absolutely available.

From the wiki: "The introductory design is fabricated using a 90 nm SOI process, with initial volume production slated for IBM's facility in East Fishkill, New York."

AMD did this for Intel up to the 286. Sony should have insisted upon a tidal wave of wafers, if needed.

Cell used the PowerPC isa as did the Xbox 360. Both were designed in the same IBM facility but separated by a floor. IIRC the Xbox team indirectly learned from the Cell team's mistakes at the process/microarch level.

Cell was definitely more weird to code against and Sony put max theoretical perf above Xbox's approach to be more general purpose chip architecture. So strictly speaking it wasn't like most PCs at the time in the x86 sense but in the three mostly same cores for Xbox vs custom Cell and special ways to squeeze out performance.

Sort of. The xenon cores are pretty damn close to cell PPE cores, just with VMX-128 strapped to them. They even share some taped out blocks, and have almost all of the same microarchitectural issues like the load hit store penalty.

Thanks for the clarification. I must have conflated "more general purpose" -> "more like PC".

There's a book about the development of both of them, The Race for a New Game Machine: Creating the Chips Inside the XBox 360 and the Playstation 3 by David Shippy & Mickie Phipps. [0]

It has some details on the awkward position the IBM developers were put in.

Rodrigo Copetti's excellent articles on the PS3[1] and Xbox 360 [2]

and Ars' Hannibal's article on the Xenon Chip. [3]

[0] https://www.goodreads.com/en/book/show/6056634

[1] https://www.copetti.org/writings/consoles/playstation-3/

[2] https://www.copetti.org/writings/consoles/xbox-360/

[3] https://arstechnica.com/features/2005/06/xbox360-2/

IBM sold that as BladeCenter QS series. Reportendly there even was an prototype version that had essentially same hardware as PS3 including the optical drive (production ones are dual-CPU without GPU and optical drive).

FWIW, early designs for the PS3 also were missing the GPU in lieu of another Cell chip. The end of Dennard scaling meant that they didn't hit the clock speed they originally intended (close to 5Ghz). It was originally supposed to do rasterization in software on that second cell, but relatively in the last minute they needed to strap a dedicated GPU to it. That's why Cell reads out of VRAM are incredibly slow, around 16MB/s.

While true, Its important to note that the second Cell wasn't gonna be the same type as the main CPU cell. The cell they intended to use as the GPU would have only 4SPEs(vs 7 in the main Cell)along with various rasterization components. Would have been cool. Settling on the Cell was compromise after compromise.

Iirc the PS2 also got a huge boost by being a cheap alternative to designated DVD players, but it eventually obviously paid off. I wonder how that factored into their PS3 strategy

I purchased my XBox One for the same reason (cost-effective Blu-Ray player with the possibility of playing games from time to time).

Indeed, although Sony hadn't created the DVD format (Phillips), having it did boost sales of the PS2. With the PS3 Sony was pushing its own format in direct competition to HD-DVD (also Phillips), and I'm sure losing Beta max market to VHS was still in their minds. So they decided to sell PS3 consoles at a much greater loss. It did pay off as well, but the first few years cost Sony a lot.

Reply on news.ycombinator.com