you are viewing a single comment's thread
view the rest of the comments
[+] -17 points 4 months ago* (last edited 4 months ago) (12 children)

There's a sweet spot for RAM efficiency, it's not actually a more = better thing. In many cases more RAM results in a worse bottleneck.

They have calculators for it:

https://bottleneckcalculator.tech/ram-bottleneck-calculator/

  • source
  • hideshow 12 child comments
  • [–] 13 points 4 months ago* (8 children)

    Please explain how more ram can cause a greater bottleneck then less ram of the same sku

  • source
  • parent
  • hideshow 8 child comments
  • [–] 12 points 4 months ago (3 children)

    Have more ram then the cpu can address

    What worse bottleneck then not being able to boot?

    (If anyone has this problem i will gladly take the ram)

  • source
  • parent
  • hideshow 3 child comments
  • [–] 4 points 4 months ago (2 children)

    I don't think that results in a failure to boot. Not to mention 64bit addressable space is like 16 QB (though I'm not sure all bits have traces since we're still orders of magnitude away from being able to use all those bits).

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 4 months ago (1 child)

    It had reminded me of my old laptop which wouldnt boot if it had more then 4 gb of ram

    It may have been 32 bit (dont remember the exact model and no longer have it) or it may have been due to mismatched sticks

  • source
  • parent
  • hideshow 1 child comment
  • [–] 3 points 4 months ago

    4GB is the 32-bit address space. You could be right. My thought was that there wasn't really a technical reason to not boot with more RAM than is addressable, but it's certainly possible that BIOS/firmware or even the OS treated it as a fatal error.

  • source
  • parent
  • [–] 3 points 4 months ago

    Some systems are less stable with 4 dimms populated instead of 2. This is the only real valid one I can think of.

    Or if you have an algorithm that tries to use all available ram, it will spend more time filling up more ram. Though that's the stupid algorithm, not the RAM.

    Or if you add virtual ram and run programs to the point where it needs to constantly page data out and in. Though that's running more programs than you have ram for and it suffers from a lack of RAM, not the other way around.

    Maybe with bad RAM refresh settings where all RAM access is paused during refresh will slow down the system with a sufficient amount of RAM if it needs to be refreshed in series. Though I'm pretty sure I've already seen UEFI settings to do that dynamically over sections of RAM, plus I think that RAM already parallelizes it inside the dimms because it's an obvious limitation for them.

    Oh, another real one, though I don't think it has a huge impact, but the amount of available RAM can affect how many bits are used in the data structures used to manage/track memory allocations, and the number of bits could determine the size of the structures, though those could also be dynamic and depend on memory used rather than available, but I'm not familiar enough with memory allocators to say for sure (both whether it would be a factor at all and how well current managers would handle it). Though even if it does make an impact, each bit added means double the RAM handled, so it doesn't even scale that badly, and could be optimized to that "used" version if it is the "available" version.

    So yeah, without a better mechanism to create bottlenecks, I'd call BS on that statement.

  • source
  • parent
  • [–] 1 point 4 months ago (2 children)

    I guess bigger ram might have slower speeds and vice versa

  • source
  • parent
  • hideshow 2 child comments
  • [–] 2 points 4 months ago

    Are you pointing to a price-performance degradation or a measurable degradation on hardware with all other factors controlled? The latter doesn’t exist.

    All the SEO slop that appears similar to your link (that I’m definitely not clicking) are for PC builders and nobody is making these outrageous claims for server builders.

    There are workloads that demand a higher memory-to-compute ratio than others. Web server? Throw them a bone sure. But they mainly want computing power. Big database? Can’t have enough, and can’t have them fast enough. Especially when a single index can be multiple gigabytes big.

  • source
  • parent