24

I mean if we increase the size of a sdram with same technology does the response time get slower? If it does, is it about complexity of digital logic?

2
  • 5
    From a PC user point of view, it might have sense to have more memory even if it gets slightly slower, because even memory that is unused by applications is usually used by an operating system as a some sort of cache to optimize performance.
    – user
    Jan 12, 2015 at 20:15
  • 2
    I don't think that this is important for any standard user. The decrease in speed is low (see below) and if you really (think you) need that little bit of speed, you can always go for higher clocked RAM (and DDR4 by now). There may be some cases where it's required, but I guess that most people reading this will not need to worry about this,
    – Sebb
    Jan 12, 2015 at 21:55

2 Answers 2

31

Yes and no. As duDE states, memory will never run faster than the bus/clock speed driving it, but the maximum speed of memory definitely is dependent on size.

As a memory assembly gets larger, the number of levels of address decoder increase (with the log of size) and the load on the drivers increases linearly (producing roughly a logarithmic increase in delay).

So, while it is rarely worthwhile to limit the size of RAM in a off-the-shelf system in an attempt to increase speed (there are exceptions where the box adjusts clock speed based on the size of RAM), if you're a system designer the maximum RAM size is one of the performance tradeoffs you must consider.

14
  • Of course this is one of the last things as a system designer you should consider considering. There are other variables to having additional memory installed in a system. Heat and power usage are just two of those variables.
    – Ramhound
    Jan 12, 2015 at 12:40
  • @Ramhound - Actually, having actually designed a couple of CPUs, I can tell you that memory performance is one of the first things that the designer considers, since much of the CPU logic must be built around that. Jan 12, 2015 at 12:47
  • 2
    @Ramhound - Except when designing for a specific low-power environment, thermal and power limits would be "ballparked" by picking the basic technology, then not seriously revisited until the design was nearing completion. Memory speed, on the other hand, pervades the design. Jan 12, 2015 at 12:57
  • 1
    Since this is the accepted answer now, it would be nice to add a note about dual-channel :) For most home users this is more relevant than the speed decrease discussed here, but that question could appear in the first results of google etc.
    – Sebb
    Jan 12, 2015 at 21:58
  • 2
    @Peter - Get yourself a bunch of gates and build a demux. You will find that the speed is proportional to the log of the number of outputs. Jan 13, 2015 at 1:25
2

No. It does not. As SDRAM is synchronized with the system, their speed depends on the speed of the system. What may affect the speed of the memory access is the configuration they are used in.

If your build already has a dual-channel(or triple channel) configuration, and the increased memory does not use identical modules, then you may slow down to single channel operation. However, this decrease is hardly noticeable, as Wikipedia says :

Tom's Hardware found little significant difference between single-channel and dual-channel configurations in synthetic and gaming benchmarks (using a "modern (2007)" system setup). In its tests, dual channel gave at best a 5% speed increase in memory-intensive tasks.

In this case the "speed" may decrease, but you will experience an overall boost in performance due to the greater amount of physical memory available at the disposal of your OS. This, of course, depends on the OS you are using and how efficient it is at utilizing the available resources.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .