Thinkpads Forum

As I continue my journey down the path of hardware hacker, I'd like some help from someone who knows more about RAM than I do (which.... is probably most people!)

Given a system with two channels of DDR3, is there an appreciable difference in power draw when populating them with modules that run at 1600, 1866, and 2133? If so, any ballpark figures? (Assume the same size for each, let's say... 8GB per module?)

Thanks in advance. I feel a bit silly bothering you guys over something so trivial, but I'll be darn if I can actually find a straight answer to this anywhere on the web. (Most of the RAM freq. discussions are more desktop-oriented, and the laptop ones quickly devolve into people arguing over whether the higher freqs are necessary.)

Ray is probably the best one to answer this one, maybe he'll chime in.

My thought is that for a given voltage you'll see more power draw and heat on the higher clock speeds, in much the same way you see that with CPU frequencies. It's probably negligible with RAM, but it's true... I've never seen it discussed or written about before!

Even better. When you have run the DDR3 2133 or 1866 at 1600, will produce less heat and even more better when notebook (ivy bridge, Haswell and later processors) using DDR3L which is less heat even more running at 1600 using 1866/2133 DDR3L at 1.35V vs 1.5V. Combine that both, you have double bingo on less power use. Dual channel is best especially with integrated GPU.

What is your notebook?

Cheers, thinkpadcollection

My thought is that for a given voltage you'll see more power draw and heat on the higher clock speeds, in much the same way you see that with CPU frequencies. It's probably negligible with RAM, but it's true... I've never seen it discussed or written about before!

But that's not strictly true with CPUs though, is it? The TDP is often the same across different frequencies of the same model CPU (ex: the i5-3320M and i5-3340M have the same TDP despite the latter having a base freq. of 2.7 GHz vs 2.6 GHz). It's just a bit more confusing with CPUs since they're not usually running flat-out (so the "hurry up and sleep" approach means that a faster chip can get into a deeper C-state faster, and thus use less power over a given period.)

I know that for RAM the difference *should* be minimal, but as you pointed out, there's not really any discussion of it anywhere.

TDP isn't power draw, it's an indicator of heat dissipation.

Quality of same silicon varies across same slice of dies. They are tested and programmed to best use most of same slice. For example fewer crop of cream processors for xeon and ones that does not make the cut for xeon gets i7 and plenty of i5 that does not make it to i7 status. So on. Now, new dimension to this is all the processors also had to be tested again for target TDP at different frequency for all of them. The best die tend to be running cooler at higher frequency anything else get sorted on different frequencies while meeting target of wattage but not exceed TDP. Overclocking is possible because of margin is calculated during design phase of new silicon spin during design work of specific percentage. I had little luck with overclock back then so I didn't bother.

Cheers, thinkpadcollection

Here's my take on this, from years (decades, even) of a memory test engineering background. Given that the memory clock speed is kept constant in Thinkpads (I'm assuming that all current and past designs do NOT adjust their clock speed when presented with memory modules that have a higher speed rating than what was initially available when the laptop motherboard was designed AND that no hardware tweaks have been done, read: BIOS does not adjust memory clock given the module's SPD information and there's no overclocking), then the "newer and faster" modules will probably draw less current. However, to be sure, you'll need to carefully consult the published datasheets of the modules that you intend using. Note that the current consumption limits in datasheets are MAXIMUM values, although in some cases you'll be given "typical" values. This means that the memory module manufacturer guarantees that the module will not draw more than the specified maximum value, but it will often operate at a somewhat lower value. Power consumption of memory modules does depend primarily upon the clock frequency it runs at. Given a PC3-17000 (2133Mhz) rated module, it will consume less current than it's typical value if put in a system that's speed locked at PC3-12800 (1600Mhz). How much less can sometimes be determined from the datasheet.

Is there an appreciable difference? Probably not. We're talking a couple of mA here and there. That might equate to a few minutes of runtime when operating on a battery and probably no discernible difference in the amount of heat that pours out the side.

I'm running coreboot, so yeah, it actually should handle higher clock speeds just fine. coreboot's RAM init will handle higher frequencies (and IIRC can even be made to ignore what it reads via SPD...)

But thank you, that answers my question nicely!