Broadway Complete!

[Link]

RevoGaming has heard from sources inside the development community that IBM has finalised the Broadway processor and is showing it to developers. Details inside.

The details come from two different sources inside the development community. One source revealed the information to RevoGaming around three weeks ago, whilst the other revealed it to us a few days ago. We were waiting for a word from another development insider before we would report it.

Apparently, the "Broadway" processor is finished, and IBM has shown plans and a prototype to Nintendo. After some discussion, it has been shown to select developers. One source said that the processor was a dual-core one, whilst the other would not reveal any technical information.

Both sources would not talk about ATi's "Hollywood" graphics chip or any other internal Revolution components. The names of the sources cannot be revealed at their request, nor the developers they work at.

There is no official word from Nintendo, so this should still be classified as a rumour.

RevoGaming will keep you updated.

http://www.revogaming.net/html/modules/news/article.php?storyid=145

[Konfucius]

Good to hear the processor is complete but why can't they leak some specs? :)

[system_error]

A friend of mine has a uncle who has a cousin and the brother of him works at a store which sells Del Monte bananas and he said that the IBM Broadway chip will feature IBMs new RubberD uCK technology.

It enhances the plastic look of a given object by 1 and it can float. It looks very realistic in water areas especially in a bathtub

[Link]

A friend of mine has a uncle who has a cousin and the brother of him works at a store which sells Del Monte bananas and he said that the IBM Broadway chip will feature IBMs new RubberD uCK technology.

It enhances the plastic look of a given object by 1 and it can float. It looks very realistic in water areas especially in a bathtub

why are u being an ass?

[YenRug]

why are u being an ass?

Because he's a developer and doubts the likelihood of this news being true, at least with regard to the possibility of it leaking like this.

[system_error]

@Link: Don't feel insulted so easily I was making fun of that article. My intention was clear - posting a sarcastic reply to an article which is horrible. I think it is good that you posted it but please allow me to express my opinion about it.

Look why would IBM need to show the chip to developers? The architecture and the custom features Nintendo wanted have been finalised surely months ago and developers don't need to see a chip but rather a good SDK with the needed documentation. "Look Mr. EA this is the Broadway chip it is silver and very shiny ... look at the little golden pins on its back aren't they beautiful?"

Furthermore one developer says dual core, the other says "I don't say anything" and both don't comment on the ATI chip or anything else Revolution related? Those must be insiders for sure ...

[DCK]

system_error can be so undeniably right sometimes :D

[Hellfire]

And isn't dual core kind of hard to develop for? That seems to defeat the purpose of Revo being simple.

[system_error]

Well since Sony and Microsoft use even more cores developers have to get used to it and at least dual core CPUs are "easier" to code for than 3 or more. Moreover you gain (IF you have the neccessary skills) much more power from a dual core CPU with less MHz than a single CPU with a lot of MHz. Temperature is definitly a issue for the small Revolution case so dual core means less heat (although it takes a bit more space depending on the architecture) but a good amount of power.

[DiemetriX]

Well.. if Coding for a singel Core is a 1 (in terms of difficulty) then coding for a dual core is a 10. With revolution already using technolgy that makes games run better (singelpass) Singel core is the way to go IMO.

[Gaijin von Snikbah]

Call me cwazy, but doesnt the DS have 2 processors? One ARM9 and one ARM7. Im also thinking it could be the physics engine, wich is almost like a second CPU. I dont believe dual core is 10 times harder to develop for.

[Jordan]

Call me cwazy, but doesnt the DS have 2 processors? One ARM9 and one ARM7. Im also thinking it could be the physics engine, wich is almost like a second CPU. I dont believe dual core is 10 times harder to develop for.

No no, the ARM7 is waaaay to weak to do any kind of physics calculations. Same with the ARM9 actually.

Basically, the DS has two processors so that the CPU's can share out the calculations or so that they can display two different things on two screens and also for GBA Emulation.

[DCK]

Yeah, the ARM7 is just a nice chip to dump annoying stuff on you don't want to bother the ARM 9 with. It's mostly used as a sound processor and network manager, so that the ARM 9 can focus on what's more important. Dual core works a bit different.

Besides, if Nintendo uses dual core, then PC programmers that are getting used to the new PC dual core generation (Pentium D and Athlon 64 X2) can use their experience for the Revolution.

I'm not worried too much about the CPU, it's the GPU Nintendo needs to do a good job on, seeing how much the 360's GPU costs.

[Schpickles]

For starters, neither of the DS chips do floating point maths :) Not great for physics :)

Back on topic:


Well.. if Coding for a singel Core is a 1 (in terms of difficulty) then coding for a dual core is a 10. With revolution already using technolgy that makes games run better (singelpass) Singel core is the way to go IMO.


There's no reason why developing for parallel processing should be more difficult. Most of the scare stories come from the fact that development teams currently have highly optimised physics and graphics libraries built around single threaded game apps, which need substantial re-writes in order to work on multi-threaded systems. As most new libraries weren't ready in time for development of early X360 titles, most developers stuck with what they knew, and used single threaded libraries for the launch titles. Dual core CPUs needn't even be an issue if compiler support is good enough - you compile the code the same way as always and the compiler does the work for you. However, that's a huge, huge if... normally the compiler is the "weakest link" in the multiprocessor code, as things stand. that's why developers want to get low level with the CPU...


Look why would IBM need to show the chip to developers? The architecture and the custom features Nintendo wanted have been finalised surely months ago and developers don't need to see a chip but rather a good SDK with the needed documentation. "Look Mr. EA this is the Broadway chip it is silver and very shiny ... look at the little golden pins on its back aren't they beautiful?"

Picking up from the above, the compilers need to be optimised for the processor to make the SDK worth its salt. Furthermore, in the absence of a decent compiler (which will be a straight gcc compiler, lets face it - no SNSystems for this Nintendo console - Sony bought them out) developers will want to get "down and dirty" with the CPU, and use some ASM to get the most out of a multicore architecture. At that point, a full ASM, register and pipeline understanding of the CPU is absolutely paramount.

[system_error]

Metrowerks CodeWarrior might be the alternative to GNU GCC or even a special compiler from Nintendo itself.
Of course a good compiler is needed but so is a fundamental understanding of a parallel architecture. Even the best compiler can't help if your calculations are data dependend. Mapping and scheduling is quite important too. The PowerPC has a rather big pipeline which can speed up things (Fetch, Decode, Get Operands, ...) if the branch prediction works out well.

The problem is with multicore development that everything has to be adjusted to the new plattform. You need the right software and a different design because with a lower clock rate (assuming that the Revolution will have a rather low clock rate with a dual-core layout) you don't want to miss more cycles than neccessary. I am quite sure that at the start of the dual-core era much power is wasted because I doubt that the average developer had to care about such things before.

[Schpickles]

...or will have enough time to care for them in order to make a launch game...
Doesn't CodeWarrior IDE (shudder) use GCC anyway? It really isn't a great IDE, imo, and inevitably if you are using visual studio, you'll be using GCC. Even the SN compiler is only a modified GCC compiler, effectively.

Remember that the coders are often dependent on middleware tools as well... if these aren't ready to go with multithreading, they'll either not be used (forcing developers to write their own solutions) or they'll be used, forcing games into single-threaded versions of themselves (you can't mix and match single- and multi-threaded libraries).

You're absolutely right - the early games won't be close to maximising performance. But then, that's true of any new console hardware to be fair (and pretty much always the case for PC development). Parallel processing is nothing new though - developers have had to wrap their heads around it for the Saturn, GameCube (to a slight extent) and especially on the PS2. It's a steep learning curve for PC-centric developers, but for the oldschool guys... it just takes time.

It's time that's the main problem - launch titles are tight as it is, without writing new libraries, re-writing in-house engines, or relying on buggy early implementations of 3rd party software. Not an easy job :)

[faz99]

why are u being an ass?

I thought what he said was funny. :laughing: