News Empty RAM slots can harm DRAM performance — Asus NitroPath slots curb electrical interference, gain 400 MT/s, and are 40% shorter

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
So why not use termination modules in current boards?
I can think of three reasons: cost, potential for user error (i.e. putting them in the wrong slots), and it's not obvious to me the additional load of terminating those extra slots won't impact the memory controller or affect the top performance achievable vs. ASUS' solution.
 
This phenomena got me wondering what what happen in overclocking with 4x8GB modules populating all slots of course. With only 32GB for the IMC to grapple with it will be easier to aim for high RAM clocks in theory at least & in 1:1. However I've only seen Kingston make 6000MT/s 8GB modules rated at 1.35v with CL of 36. It would be nice to see some experiments with this on current AM5 chipset boards.
 
Last edited:
This phenomena got me wondering what what happen in overclocking with 4x8GB modules populating all slots of course. With only 32GB for the IMC to grapple with it will be easier to aim for high RAM clocks in theory at least & in 1:1. However I've only seen Kingston make 6000MT/s 8GB modules rated at 1.35v with CL of 36. It would be nice to see some experiments with this on current AM5 chipset boards.
8GB modules aren't common because that limits them to 4 memory chips per module so there won't be any high speed ones.

Intel's specifications for 13th/14th Gen also indicate it's the act of having 2DPC which is the problem rather than the capacity or anything else which makes sense as you still have two termination points per channel.
YP8q0AC.png

EVcoYvJ.jpeg
 
Last edited:
8GB modules aren't common because that limits them to 4 memory chips per module so there won't be any high speed ones.
Yeah, don't they have to do some weird dance to multiplex twice the data across what the DRAM chips natively support? I'm a little foggy on the details, but 8 GB DDR5 DIMMs are supposed to have an additional performance penalty beyond what you get with 16 GB at the same speed.

Intel's specifications for 13th/14th Gen also indicate it's the act of having 2DPC which is the problem rather than the capacity or anything else which makes sense as you still have two termination points per channel.
YP8q0AC.png

EVcoYvJ.jpeg
The table you quoted does indeed go beyond number of DIMMs per channel and further differentiates whether they're single-ranked or dual-ranked.

Right now, all 8 GB, 16 GB, and 24 GB DDR5 DIMMs are single-ranked. It's also a pretty safe bet that 32 GB DIMMs are dual-ranked, but higher density DRAM chips are coming and we should eventually see single-ranked 32 GB DIMMs.
 
8GB modules aren't common because that limits them to 4 memory chips per module so there won't be any high speed ones.

Intel's specifications for 13th/14th Gen also indicate it's the act of having 2DPC which is the problem rather than the capacity or anything else which makes sense as you still have two termination points per channel.
YP8q0AC.png

EVcoYvJ.jpeg
That's Intel, I should have been clearer because I was referring to AM5 platform.