• Acolyte of the Butterfly
    Acolyte of the Butterfly
    Posts: 104 from 2009/3/20

    Btw, the PowerXCell 8i wasn't supposed to power the PS4 either, yet it got R&D'd. Care to explain?

    Well, sure, PowerXCell 8i has 5-8 times *more* double precision floating point math performance (actually that's the theoritical limit, usually it's more like ~4x faster), and allows up to 16GB of RAM. It's *targetted* for scientific simulations, where floating point is almost useless, and lots of RAM is of vital importance. The plain Cell is fine for games (32-bit floats are fine for a game, but they just don't cut it when it comes to scientific calculations, they're almost useless there). But I guess you never had to do that so you wouldn't know.


    Maybe they had plans for their Cells that never came to fruition, who knows. Fact is, you (and me) know nothing about IBM's internal plans for the coming PowerXCell 32. At least, there's not one single document out there proving that R&D of PowerXCell 32 has anything to do with PS4. So you're obviously talking out of your behind again.

    Well, sure, let's see. Amidst the insults, I see a challenge. So, if we have an 8-SPU Cell used in PS3 good at 32-bit fp math and shortly after a HPC-targetted PowerXCell 8i, still 8-core but ~4x faster at 64-bit fp math, used in huge clusters and supercomputers, and we have an announcement of a 32-SPU Cell successor (2 versions, 32-SPUs each, 2/4 PPEs resp.). Now, why does that sound familiar? Oh wait, it might be because one model, one possible explanation might be that the 2-PPE one would be strong only on 32-bit fp and the 4-PPE strong on 64-bit fp as well, just like the Cell and PowerXcell 8i. Of course that's just a guess, because not much specs are available on either CPU model, in fact, it's only a rumour that there are two versions. Of course you wrote it as a fact, falling in the same trap you accuse me of.


    No, the 450 is not sold on the free market. It's only used within BlueGene systems. So, they indeed have done that before already in an even stricter way.

    Oh, now you have inside info of the IBM CPU sales? Fantastic. However, I think this is funny:


    check the part where it says "A third synthesizable processor core, the IBM PowerPC 450, targeted for the networking and communications segments, is expected to be available in 2006."

    Availability for what? demos only? And you would know this, how exactly?


    The PowerXCell 8i is not only used in their RoadRunner but also on server blades and PCIe boards for the free market. Why shouldn't they do the same thing they already did with the PowerXCell 8i and (in an even stricter way) the 450 with the PowerXCell 32 as well?

    Because the 450 is cheaper to R&D and produce and sell than the PowerXCell 32 by a big factor? Especially, when Sony or someone else is not interested in funding the R&D? What I mean is, if Sony is not sharing R&D costs, of what benefit would it be to IBM to take it up on itself to develop a 32-SPU Cell, only to put in a few super computers and sell a few thousand blades/boards? If you add up the costs, it doesn't make financial sense.


    From the PowerXCell 8i being a better version of the Cell B.E. you conclude that the PowerXCell 32 was intented for the PS4? What's that kind of logic called?

    Please read above.


    Really? The PowerXCell 8i didn't end up in a Sony's gaming device, yet it got R&D'd and gets produced and is sold on the free market as part of certain IBM systems today. Care to explain?

    Please read above.


    You're tired of me posting tons of links? Let me recommend you to get some sleep helping your tired and confused mind then. Else your delusions might reach a dangerous state.

    Thank you that was a nice advice, I can see more clearly now that my tired and confused mind wasn't playing tricks on me. You firmly established my opinion of you.


    As for the unrelatedness of embedded CPUs: Why is the QorIQ related, but other future e500s and future 460s are not?

    And other future e500s and future 460s certainly won't fit netbooks?

    Sure, when they're here, they might, "future" being the key word here. QorIQ at least has already been shown and demoed, and available in mid-2009 -or so they say anyway. Same argument used as for the Titan, let's see it first, even as a demo and then we can talk all day about the merits of each CPU.


    Most important regarding which aspects exactly?

    Brand recognition. Most people know Power6/7 and its -allegedly insane- performance, but very few know names and specs or even existence of the other models. I guess you don't disagree there.


    To sum it up: You claimed the QorIQ and POWER7 to be the only future Power Architecture CPUs and now justify your denial of the other lines with something like thread context, which is ridiculous especially regarding your mentioning of POWER7. You could as well have mentioned the e200 line instead for it's probably the most used Power Architecture CPU line of all. That would have been as much within this thread's context as POWER7. Your justification is composed of inconsistent reasons you now come up with to conceal your claim being false.

    You know, I've never met anyone so obsessed with proving the other one wrong. You remind me of http://xkcd.com/386/. SO WHAT if I neglected a few unrelated CPU models? I mentioned real and existing CPUs, which are the main focus points of the two biggest PowerPC companies, and you mentioned future ones. And you call me inconsistent? Let's wait first ok?


    Btw, you still keep failing to back up your claim that Sony "first [...] started with a Cell 3 design" for the PS4.


    These are ...

    Bah, I just felt an strong earthquake now and I realized just how much I'm wasting my time replying to you. Last message on thread, enjoy your meaningless triumph.

    [ Edited by feanor on 2009/4/12 2:59 ]

    [ Edited by feanor on 2009/4/12 19:42 ]
  • »12.04.09 - 00:58
    Profile Visit Website