G5 64-bit Vs. Alti-Vec

Discussion in 'General Mac Discussion' started by shadowfax0, May 4, 2002.

  1. shadowfax0 macrumors 6502

    May 2, 2002
    Ok, here's the idea, I hear all this discussion about how programmers are unwilling to add the code to optomize a program for Alti-Vec, so what's going to persuade them to convert the whole thing into 64-bit for the G5 (if it's 64-bit that is)? Ok, I mean I can see that happening, but I guess this is what I am getting at...isn't Alti-Vec supposed to run at 128-bits in the first place? i.e. so the computer can handle '128 bit chunks compared to 32 blah blah' so isn't stepping up to 64 just inbetween and we'll see better performance with Alti-Vec optomized programs? I'd really like to hear some opinions/facts on this, because I am DEFINATELY skipping over something, I just know it :D
  2. Catfish_Man macrumors 68030


    Sep 13, 2001
    Portland, OR
    A better way of thinking about Altivec...

    ...is that it's "32 bit * 4". I doesn't process 128 bit chunks, it just does four 32 bit ones at once. 64 bit chips on the other hand, process 64 bit chunks (which are only used for very precise floating point numbers), and have 64 bit memory addresses (so they can support more than 4 gigabytes of ram). 64 bit itself doesn't add very much performance, except for things that use a lot of double precision floating point math (which is not much), or use more than 4 gigs of ram (again, not much [outside of servers]). The reason why things like the Itanium are faster is because as well as being 64 bit, they're also a complete redesign of the chip which drops all the 20 year old baggage associated with Pentiums/Athlons. A 64 bit G5 would add a bit of performance, but not as much as one might think from the hype. x86-64 (the 64 bit extensions that AMD's upcoming Hammer chips use) mainly provides a performance boost because it adds a bunch of registers to the chronicly register short x86 design.

Share This Page