Fishie on 10 July 2008
| goddog said: back to another problem though. with the cell sony was trying to force programers to work one way, which looked like they could do giving their position in the industry. it has not panned out. so that leaves sony with a system that though technicaly more powerful in theory or for future programers who work at it, for practical purposes no bonuses yet. apple had this problem with alvatec on powerpc platform, though a 2 ghz chip could out perform much faster x86 chips when optimized for the alvatec engine, it required significatly more dedicated code than x86. so few companies out side of apple used it (adobe being one of the few). this is one of the maine reasons apple is x86 now this is the same problem facing sony for cell to make it to the ps4 they must convice developers to code specificly for it, maybe pay them, or discount the license if they take full use of the cell, hold more work shops send out more help. we have seen sony start to do this already. I doubt though it will be worth it in the end, though multicore is the future, cell requires too much extra work to get the results, i for see in the ps4 some sort of comprimise based on ppc arch maybe a 8-core, with one spe per core preprogramed for either vector or floating point. or far more likly some other ppc customized processor with no spe.... the reason i feel spes are doomed, is because you have not seen aroll out by ibm or freescale, both of whom make majority of the ppc chips that are used in everything form tvs to cars |
I wrote an article about what IBM did, how Sony and Apple were screwed over etcetera long before apple switched to Intel and details of CELL became public.
Sadly nobody dared publish it or wanted to fund further research for the article and since you seem to be pretty into this feel free to PM me your email address so I can forward you what I wrote way back.
