8 ms·

Just to inject a bit of history. The previous eras of consoles had a design complexity that made it hard for developers to get anything done or tap the potenti

Just to inject a bit of history. The previous eras of consoles had a design complexity that made it hard for developers to get anything done or tap the potential of the hardware. This mostly includes the fifth through seventh generation of video game consoles, so PS1-PS3 era, roughly 1993-2010. The “radically different” architectures are sure a bit more exciting, but as a developer it’s gonna take a few years to really figure out how to use these different architectures to a reasonable semblance of their full potential. The Atari Jaguar had this crazy bus with a 68000 plus two custom CPUs on it, and to unlock the full potential you have to reduce bus contention which means paying attention to what data is local to the SRAM available to each processor. Maybe you need to rearchitect your game to really unlock the potential of the Jaguar. Fast-forward to 2006 and the PlayStation 3 has this fancy new Cell processor where… guess what… you need to pay attention to what data is on the local SRAM for each core and it was a nightmare to figure out how to really use that. Honestly, I think that console manufacturers were in danger of getting strangled by game developers if they kept up these radical architectures. Exclusivity, on the other hand, is a key tool in Sony, Microsoft, and Nintendo’s arsenal. Good luck prying that away from them.

mdasen 7y ago
I definitely concur that the radically different architectures weren't good for developer productivity. There was a lot of crazy going on. I guess it just seems like we're now wasting money on duplicate consoles. It feels like it would be better if Sony and Microsoft created a "standard" for manufacturers to build to. Zen 2 3.5GHz with a specified level of Radeon graphics and 16GB of RAM (with a bit more specifics). You could still have exclusives to your platform - Microsoft has DirectX, Xbox Live, and other exclusive software features. Developers would still want access to a platform for their games. Don't they usually lose money on the actual hardware? Nintendo is doing a slightly different thing with the Switch, but the PS4/XBONE and PS5/XBX it seems like they're just making people pay twice for the same hardware just because they've locked up some exclusives. What's the business model there? 1) Lose lots of money selling a console; 2) We'll pay to develop first-party games that we'll sell to fewer people because they're exclusive; 3) We'll pay third-parties to make their games exclusive to our platform (as compensation for the lost sales due to exclusivity); 4) charge huge licensing fees for access to the platform for non-exclusive developers. It just seems like there are a lot of money-losing (and consumer-happiness-losing) steps to get to #4. That's a lot of loss to society just to get to #4. I guess the reason why they can charge such high prices for licensing is that developers know that consumers will likely only purchase one system. If people had access to both systems on a single box, then developers could negotiate them against each other. Now, developers can't negotiate them against each other unless they want to lose out on a lot of potential customers.
- ryandrake 7y ago
  > It feels like it would be better if Sony and Microsoft created a "standard" for manufacturers to build to. You're talking about two companies notorious for ignoring existing APIs and standards, coming up with their own proprietary systems, using marketing and FUD to knock down and strangle existing and emerging standards, and then spending the rest of their effort locking customers into their proprietary stuff. And they're not alone. It's the standard tech company mentality: Deliberately make your product different and incompatible with everyone else and compete instead of cooperate. Best case outcome is monopoly and customer lock-in. Most likely case is no one solution gains enough traction and you end up with 15 incompatible chat apps. Worst case is you emerge from the fight as HD-DVD.
- ATsch 7y ago
  This is basically what the "steam machine" model was. Standardized controller, hardware specifications, OS, available from multiple vendors. It was unfortunately a bit of a flop.
  - badsectoracula 7y ago
    Not really, Steam Machines were just a PCs running Linux at a time when Linux was in a far worse state for gaming than they are now (and almost all of Steam Machines produced either had weak Intel-based graphics or AMD GPUs when their drivers were still woefully broken). There was never any standard for hardware or even controller and most machines were overpriced, ugly, huge or all of that. So of course they flopped. Steam Machines should have used standardized hardware (even if it was in terms of "levels" like "entry level" or "high end level"), software and peripherals the developers could rely on and Valve should have made sure that the hardware actually fit their respective levels. But Valve never did that and while they still slowly update SteamOS, they have been long sidetracked by VR so there are low chances for them trying a second time.
    - mjevans 7y ago
      They'd have been better off defining a standard with something vaguely like a year, or pinned against a console generation as a reference point.
leetcrew 7y ago
I always wondered who thought it was a good idea to make a gaming console with seven cores in 2006. iirc, most pc games at the time barely used more than one thread. even in 2020, most games have a main thread that pegs one core at 100% and creates a hard ceiling for fps. what were they thinking?
- klodolph 7y ago
  Maybe they took some cues from the people who released an eight-processor system in 1994? Granted, only two of those are CPUs.
- wmf 7y ago
  The PS3 originally planned to run the game on the PPE and shaders on the SPEs; you could think of it as a single-core CPU with a 7-core GPU on one chip. At some point they realized this wasn't powerful enough so they added the RSX but that left the SPEs with no particular job to do. It was too late to redesign Cell so they just shipped it and punted the problem to game developers. In retrospect, a G5+RSX with unified memory probably would have been much better.
  - stordoff 7y ago
    > In retrospect, a G5+RSX with unified memory probably would have been much better. Which is not _that_ far from what the 360 ended up with. The 360 CPU is basically a triple-core variant of the the Cell without the SPEs, and it has 512MB unified memory.
    - monocasa 7y ago
      I'll throw out there that a G5 was never in the cards from a count count budget perspective.
      - monocasa 7y ago
        *from a gate count perspective
  - 7y ago
smogcutter 7y ago
I hadn’t known about the jaguar’s hardware complexity until that article that was on HN a few days ago, it’s really fascinating. I think a lot of the complexity comes from needing to cram what’s essentially a high end gaming PC into a <$500 price point. You can do it by creating specialized hardware and putting developers face to face with the metal, but that comes at a cost of complexity.
- pjmlp 7y ago
  The complexity is only an issue at the beginning, having standardized hardware without having to deal with hardware fragmentation and driver issues, is usually seen in the community as positive.
hinkley 7y ago
Hasn't NUMA support improved in the last 14 years though? Surely someone will try again.
saagarjha 7y ago
> The previous eras of consoles had a design complexity that made it hard for developers to get anything done or tap the potential of the hardware. I thought the developers that did figure out how to tap the potential of the hardware did quite well?
- rodgerd 7y ago
  Do you want your game creators focuses on twonking around in the guts of an obscure, probably poorly documented and understood one-off architecture, or worried about creating an interested and enjoyable game?
  - taeric 7y ago
    This is a false dichotomy. If anything, I could see arguments that having to squeeze out capability it's a filter to only get things in the game that pass a development budget. Not just all you can throw at it.
    - JohnBooty 7y ago
      I don't agree that it's a false dichotomy. You have a limited budget and a limited number of employee-hours with which to create and ship a game. Every hour spent fighting with the platform itself is very literally an hour spent not creating cool and fun stuff. That time investment would only be worth it if a "novel" platform delivered huge gains in return for time invested. However, IMO we've long passed the point where that's feasible. CPUs and GPUs are very mature at this point: they are the result of decades of evolution and trillions of dollars of R&D. Inventing your own radically-different, competitive, bespoke architecture for a game console at this point would require an investment orders of magnitude beyond the expected earning potential of these systems. It also wouldn't be guaranteed to work.
      - taeric 7y ago
        Agreed I used too strong of language. I should have said it could be a false dichotomy. My perspective is colored by being a huge fan of the demo scene and loving older games. A large part about what made some of them fun was understanding just how hard they were to do. Is also, I feel, why so many games did not have as much in game help. Couldn't put it in. There is something to be said for that. My hunch is there was a selective pressure such that only really amazing things survived. That said, how many potentially amazing things were killed fighting the platforms? Could Daikatana have been amazing with more powerful tools? I honestly don't know.

mdasen 7y ago

I definitely concur that the radically different architectures weren't good for developer productivity. There was a lot of crazy going on. I guess it just seems like we're now wasting money on duplicate consoles. It feels like it would be better if Sony and Microsoft created a "standard" for manufacturers to build to. Zen 2 3.5GHz with a specified level of Radeon graphics and 16GB of RAM (with a bit more specifics). You could still have exclusives to your platform - Microsoft has DirectX, Xbox Live, and other exclusive software features. Developers would still want access to a platform for their games. Don't they usually lose money on the actual hardware? Nintendo is doing a slightly different thing with the Switch, but the PS4/XBONE and PS5/XBX it seems like they're just making people pay twice for the same hardware just because they've locked up some exclusives. What's the business model there? 1) Lose lots of money selling a console; 2) We'll pay to develop first-party games that we'll sell to fewer people because they're exclusive; 3) We'll pay third-parties to make their games exclusive to our platform (as compensation for the lost sales due to exclusivity); 4) charge huge licensing fees for access to the platform for non-exclusive developers. It just seems like there are a lot of money-losing (and consumer-happiness-losing) steps to get to #4. That's a lot of loss to society just to get to #4. I guess the reason why they can charge such high prices for licensing is that developers know that consumers will likely only purchase one system. If people had access to both systems on a single box, then developers could negotiate them against each other. Now, developers can't negotiate them against each other unless they want to lose out on a lot of potential customers.

ryandrake 7y ago
> It feels like it would be better if Sony and Microsoft created a "standard" for manufacturers to build to. You're talking about two companies notorious for ignoring existing APIs and standards, coming up with their own proprietary systems, using marketing and FUD to knock down and strangle existing and emerging standards, and then spending the rest of their effort locking customers into their proprietary stuff. And they're not alone. It's the standard tech company mentality: Deliberately make your product different and incompatible with everyone else and compete instead of cooperate. Best case outcome is monopoly and customer lock-in. Most likely case is no one solution gains enough traction and you end up with 15 incompatible chat apps. Worst case is you emerge from the fight as HD-DVD.
ATsch 7y ago
This is basically what the "steam machine" model was. Standardized controller, hardware specifications, OS, available from multiple vendors. It was unfortunately a bit of a flop.
- badsectoracula 7y ago
  Not really, Steam Machines were just a PCs running Linux at a time when Linux was in a far worse state for gaming than they are now (and almost all of Steam Machines produced either had weak Intel-based graphics or AMD GPUs when their drivers were still woefully broken). There was never any standard for hardware or even controller and most machines were overpriced, ugly, huge or all of that. So of course they flopped. Steam Machines should have used standardized hardware (even if it was in terms of "levels" like "entry level" or "high end level"), software and peripherals the developers could rely on and Valve should have made sure that the hardware actually fit their respective levels. But Valve never did that and while they still slowly update SteamOS, they have been long sidetracked by VR so there are low chances for them trying a second time.
  - mjevans 7y ago
    They'd have been better off defining a standard with something vaguely like a year, or pinned against a console generation as a reference point.

leetcrew 7y ago

I always wondered who thought it was a good idea to make a gaming console with seven cores in 2006. iirc, most pc games at the time barely used more than one thread. even in 2020, most games have a main thread that pegs one core at 100% and creates a hard ceiling for fps. what were they thinking?

klodolph 7y ago
Maybe they took some cues from the people who released an eight-processor system in 1994? Granted, only two of those are CPUs.
wmf 7y ago
The PS3 originally planned to run the game on the PPE and shaders on the SPEs; you could think of it as a single-core CPU with a 7-core GPU on one chip. At some point they realized this wasn't powerful enough so they added the RSX but that left the SPEs with no particular job to do. It was too late to redesign Cell so they just shipped it and punted the problem to game developers. In retrospect, a G5+RSX with unified memory probably would have been much better.
- stordoff 7y ago
  > In retrospect, a G5+RSX with unified memory probably would have been much better. Which is not _that_ far from what the 360 ended up with. The 360 CPU is basically a triple-core variant of the the Cell without the SPEs, and it has 512MB unified memory.
  - monocasa 7y ago
    I'll throw out there that a G5 was never in the cards from a count count budget perspective.
    - monocasa 7y ago
      *from a gate count perspective
- 7y ago

smogcutter 7y ago

I hadn’t known about the jaguar’s hardware complexity until that article that was on HN a few days ago, it’s really fascinating. I think a lot of the complexity comes from needing to cram what’s essentially a high end gaming PC into a <$500 price point. You can do it by creating specialized hardware and putting developers face to face with the metal, but that comes at a cost of complexity.

pjmlp 7y ago
The complexity is only an issue at the beginning, having standardized hardware without having to deal with hardware fragmentation and driver issues, is usually seen in the community as positive.

hinkley 7y ago

Hasn't NUMA support improved in the last 14 years though? Surely someone will try again.

saagarjha 7y ago

> The previous eras of consoles had a design complexity that made it hard for developers to get anything done or tap the potential of the hardware. I thought the developers that did figure out how to tap the potential of the hardware did quite well?

rodgerd 7y ago
Do you want your game creators focuses on twonking around in the guts of an obscure, probably poorly documented and understood one-off architecture, or worried about creating an interested and enjoyable game?
- taeric 7y ago
  This is a false dichotomy. If anything, I could see arguments that having to squeeze out capability it's a filter to only get things in the game that pass a development budget. Not just all you can throw at it.
  - JohnBooty 7y ago
    I don't agree that it's a false dichotomy. You have a limited budget and a limited number of employee-hours with which to create and ship a game. Every hour spent fighting with the platform itself is very literally an hour spent not creating cool and fun stuff. That time investment would only be worth it if a "novel" platform delivered huge gains in return for time invested. However, IMO we've long passed the point where that's feasible. CPUs and GPUs are very mature at this point: they are the result of decades of evolution and trillions of dollars of R&D. Inventing your own radically-different, competitive, bespoke architecture for a game console at this point would require an investment orders of magnitude beyond the expected earning potential of these systems. It also wouldn't be guaranteed to work.
    - taeric 7y ago
      Agreed I used too strong of language. I should have said it could be a false dichotomy. My perspective is colored by being a huge fan of the demo scene and loving older games. A large part about what made some of them fun was understanding just how hard they were to do. Is also, I feel, why so many games did not have as much in game help. Couldn't put it in. There is something to be said for that. My hunch is there was a selective pressure such that only really amazing things survived. That said, how many potentially amazing things were killed fighting the platforms? Could Daikatana have been amazing with more powerful tools? I honestly don't know.