3

Someone I know posited an idea that I instinctively think (know?) cannot be true, but I can't seem to figure out how to explain why not. The idea is based on the assumption that more electricity must be used to store a bit value of 1 in RAM than a 0. If that assumption is true (I'm not sure that it is), then the idea is that we could lessen the electricity drawn by systems if we hyper-optimized every encoding system such that less commonly stored values are assigned the heavier byte values (heavier in the sense of having more bits set to 1).

I'm not even sure that's a sense-making endeavor, even if the initial assumption were true. This would all certainly be moot, though, if someone could explain that the byte value x00 does not require less electricity to store in RAM than xFF.

Improve this question
13
  • 2
    Note: Your question includes an error: "bigger values" does not necessarily means that more bits are set to 1: 0x7F has 7 bits set 0x80 is greater but has only one bit set. Hence an optimization would get really complicated.
    – Robert
    Commented May 12, 2020 at 14:14
  • 1
    It's sad to note how incapable of taking a joke some of you guys are here... The only way to do what the OP implies would be a complete absence of power consumption for one state of a bit, be it "0" or "1", it doesn't matter, so that when that state was attained, no electrical power would be consumed at all. This hasn't even been achieved in regular HDDs when storing data in ferrite particles, and we're talking data storage here, not active data in RAM, so I took the OP's question to be rhetorical, and answered in kind... Maybe in quantum computing, or photon-based computing, could it be done?
    – user1019780
    Commented May 12, 2020 at 15:00
  • 1
    Don't forget that DRAM (which is typically "everywhere" these days, aside from the smallest embedded equipment) needs to be refreshed... so it stands to reason that one bit value might take more energy to read and re-assert than the other... but I couldn't tell you which, and I suspect it would be an incredibly small figure - possibly in the order of nano/femto Joules difference or less between a single 1 vs a 0(?)... scale that up to 16-32 GB of RAM and you might make a measurable difference, perhaps in the order of milli Joules.. (note: all figures are guesses).
    – Attie
    Commented May 12, 2020 at 15:13
  • 1
    Here's a write up from 2018 that suggests that storing a single bit on an SSD comes to approx 0.35 femto Joules (x10^-15) vs. a HDD at ~100 fJ (or ~285 times higher)... so my guesses may be an order of magnitude or two off. No mention of a 1 vs a 0 though.
    – Attie
    Commented May 12, 2020 at 15:19
  • 1
    This paper suggests that power consumption strongly depends on the data value being read or written (see Section 8, point 2). There are some words and figures on power consumption during refresh too!
    – Attie
    Commented May 12, 2020 at 15:30

0

Reset to default

Browse other questions tagged .