Announcement

Collapse
No announcement yet.

Allwinner A80 Octa-Core Hardware Coming Next Month

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #16
    Originally posted by Ferdinand View Post
    big.LITTLE is because of energy saving. Using more die space doesn't matter because it gets turned off and doesn't use energy. A simple core like A7 needs far less energy to do the same stuff as an A15. It is only when an A7 doesn't have enough peak performance that the A15 gets enabled. If users were willing to wait for calculations in-order cpus would be the best and most power efficient cpus. But users want peak performance when scrolling etc.
    I think big.LITTLE hasn't been used yet because the implementation wasn't smart enough to provide the possible energy savings. Tegra3 was not very power efficient.
    I fully understand that this is the theory but it apparently doesn't hold up in practice. One only needs to look at Apples A7 to see that the route to power efficiency can be accomplished on a different path. With A7 we have what might be the most advanced high performance ARM core out there running at very low power levels giving Apple fairly incredible battery life in a cell phone.

    I don't mean to reference A7 as an Apple fan here but rather to point that they have had great success doing exactly the opposite of what ARM was pushing. Two high performance cores instead of 4 or more mixed cores.

    Frankly it looks like ARM is slightly back pedaling on big.LITTLE as they are now promoting the idea of two high performance cores supported by four low performance cores. My point is if Apples rush to idle techniques are viable and they appear to be, why bother with the low power cores at all? Save the transistors for other things and run one core for light duty and power up the other under heavy load. From all appearances the transistors used for the low power cores are wasted, especially when you consider the support logic required to manage all the switching and interfacing.

    In the end I see big.LITTLE as a architectural mistake. I will be surprised if in five years it is anything more than an entry in the history books.

    Comment


    • #17
      Originally posted by wizard69 View Post
      Frankly it looks like ARM is slightly back pedaling on big.LITTLE as they are now promoting the idea of two high performance cores supported by four low performance cores. My point is if Apples rush to idle techniques are viable and they appear to be, why bother with the low power cores at all? Save the transistors for other things and run one core for light duty and power up the other under heavy load. From all appearances the transistors used for the low power cores are wasted, especially when you consider the support logic required to manage all the switching and interfacing.
      It seems logical to me that a simple core that doesn't do fancy out of order tricks uses less power than an Apple A7 that does use fancy tricks unless Apple A7 can shut down that fancy part of the core. That would be quite a trick but does that save transistors or power? I find it very hard to believe that the people at ARM are retarded.
      Have there been performance/watt benchmarks of that 8 core A7 mediatek soc and Apple's A7?

      Comment


      • #18
        Originally posted by Ferdinand View Post
        It seems logical to me that a simple core that doesn't do fancy out of order tricks uses less power than an Apple A7 that does use fancy tricks unless Apple A7 can shut down that fancy part of the core. That would be quite a trick but does that save transistors or power?
        Actually just about everybody involved in low power semiconductors is looking at power gating just about all logic that isn't active these days. The more inactive logic that does respond to the clock signals the better. Often it isn't a matter of shutting down a unit but rather simply not allowing it to see the clock signal as most power in CMOS comes from the toggling of gates.
        I find it very hard to believe that the people at ARM are retarded.
        I don't think anyone at ARM is retarded just that companies collectively make mistakes. AMD is a perfectly good example, the did well with BRAZOS and blew it with bulldozer. In the end you have no choice other than to go with what you have, after all you can't sell what you don't have.
        Have there been performance/watt benchmarks of that 8 core A7 mediatek soc and Apple's A7?
        That would be extremely interesting to say the least. Unfortunately I don't know how to do that at the moment, but it would sure be nice to have an A7 based system that could boot Linux! Even if you found the same native apps to run on both piece of hardware, Apple has done much more to optimize for low power usage so I'm not sure it would be a fair comparison. The only thing I know of for sure is that there are few platforms out there with the performance of Apples A7 based systems running well on the relatively small battery in the iPhone.

        That of course isn't the best way to compare systems, but I don't have any other useful metric.

        Comment

        Working...
        X