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

[Admin]

Asus' new NitroPath DRAM slots improve speeds by curbing electrical interference from unpopulated RAM slots, allowing for 400 MT/s increases in RAM overclocks. Der8auer helps explain the tech and why it is not found on some new Asus boards.

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

 

[FunSurfer]

There is a difference in the size of the eye also at 6,400 MT/s, it is bigger with the new connector, but the eye is very small at 10,000 MT/s even with the new connector, so there will be a stability problem at about 12,000 MT/s, so to avoid that for now a dummy RAM is the only option at those RAM speeds.

 

[H4UnT3R]

So it will scratch the RAM pins more?

 

[thestryker]

It would be interesting to see how these perform compared to using dummy sticks (though exact testing isn't going to be possible). I imagine the difference wouldn't be very large, but it'd be interesting just the same. I'd guess that current CPUs wouldn't really be able to push high enough memory clocks to negate the benefits without CUDIMMs. It's also entirely possible that the electrical properties of CUDIMMs would end up being different than UDIMMs as well.

I'm guessing the mounting pressure has more to do with the design and longevity than it being any practical advantage.

 

[cyrusfox]

So it will scratch the RAM pins more?

Agreed, I believe you will get less insertion cycles due to this, not a concern for an end user, but will impact the life of dram test modules

 

[bit_user]

This interference is why many servers occasionally use dummy RAM sticks to populate all empty RAM slots, safely terminating their open connections.

Really? The ones I've seen (which are admittedly pre-DDR5 era) were just plastic and didn't terminate anything. I assumed they were there to keep the contacts free of dust or to retain similar airflow as if the slots were populated.

The slots, engineered by Asus and Lotus

Presumably you don't mean this Lotus??

will become available for use by other motherboard OEMs after one year, meaning enthusiast-grade gaming motherboards may all be seeing a rising tide in 2025.

Very nice. Considering the revenue they'll probably get from licensing fees, I'm almost surprised it's not even shorter, like 6 months. Imagine how much $$$ they could make if even servers use this! If the statement is true about servers needing dummy sticks just for termination, you could charge server makers a per-slot license fee that's only a little bit less than those dummy sticks and they'd still probably pay.

 

[Notton]

Yeah, Buildzoid has been saying this for years about DDR5.
And whenever a maker released a 2-slot DDR5 mobo, so many people would complain.
"waaaaaaaaaaaaaaaagh! I want 4 slots! I want futureproofing when my sticks only hit 5200MT/s anyways! waaaaaaaaaaaaaaaaagh! but I want to hit 6400MT/s with 4 slots populatated! waaaaaaaaaaaaaagh!

Without ever realizing 4-slots populated guaranteed speed is like... 4200-4800MT/s
or that the extra slots was hurting the DRAM stability.

 

[bit_user]

I believe you will get less insertion cycles due to this, not a concern for an end user, but will impact the life of dram test modules

What the article said seems to contradict this:

"Asus's press release claims that NitroPath slots have a 57% higher retention force than standard ones and 20% improvements in lateral latching and release forces. This increased toughness should make a difference for gamers or other enthusiasts who often replace DIMMs in their systems"
​

It would be nice to get some hard stats on how they compare.

BTW, like a decade ago, I heard of some alternative electroplating that provided similar corrosion resistance as gold, but was cheaper and tougher. I wonder why that didn't seem to catch on.

Xtalic makes a better contact material than gold

At MWC SemiAccurate met Xtallic, a company making a better gold for contacts through nanotechnology.

www.semiaccurate.com

I also got some hi-fi cables, like a decade ago, that used some other plating than gold. I wonder if it was the same sort of alloy. Given how much environmental destruction is happening due to particularly the illegal gold mining industry (they use lots of mercury to leech out the gold and it just gets dumped into water ways), I support anything that reduces reliance on gold.

 

[bit_user]

Yeah, Buildzoid has been saying this for years about DDR5.
And whenever a maker released a 2-slot DDR5 mobo, so many people would complain.
"waaaaaaaaaaaaaaaagh! I want 4 slots! I want futureproofing when my sticks only hit 5200MT/s anyways! waaaaaaaaaaaaaaaaagh! but I want to hit 6400MT/s with 4 slots populatated! waaaaaaaaaaaaaagh!

Without ever realizing 4-slots populated guaranteed speed is like... 4200-4800MT/s
or that the extra slots was hurting the DRAM stability.

Among W680 boards, 2-slots is basically nonexistent. Well, you can find them in mini-ITX boards, but I wanted ATX. I'd have definitely gone for one, if I could've. I had to get 2x 32 GiB DIMMs just so I could get dual ranked, yet 64 GiB is already more than I expect to use for the life of the machine. Plus, now 64 GiB UDIMMs are on their way (if not already here), making the case for 4 DIMM boards even weaker.

 

[thestryker]

Really? The ones I've seen (which are admittedly pre-DDR5 era) were just plastic and didn't terminate anything. I assumed they were there to keep the contacts free of dust or to retain similar airflow as if the slots were populated.

The enterprise ones I've seen were all this, but the consumer ones have been different.

Corsair had them for DDR4 (Gigabyte sold a kit which was 2 DIMMs/2 fakes unsure if they still do): https://www.corsair.com/us/en/p/pc-...o-light-enhancement-kit-aca-a-black-cmwlekit2

They were supposed to have something similar for DDR5, but it hasn't materialized. The price, performance limit and restriction of only Intel on these v-color ones for DDR5 may explain why they haven't: https://v-color.net/collections/ddr5-rgb-filler-kit

Presumably you don't mean this Lotus??

Probably these folks: https://www.lotes.cc/en

 

[Hotrod2go]

This reminds me of the dummy sticks used in old Intel 850 & E variant chipset boards that featured RDRAM technology around the early 2000s.

 

[das_stig]

Yep, time to go back to the old RDRAM spacers for empty slots., now thta brings back found memories of my old P4 1.4Ghz what a beast

 

[TheHerald]

Why not just buy a 2dim mobo and be done with this?

 

[bit_user]

Why not just buy a 2dim mobo and be done with this?

It somewhat limits the selection, if you're restricted to just 2 slot boards. Plus, for people who either are stuck or insist on buying a 4 slot motherboard, at least this lessens the downside.

 

[thestryker]

Why not just buy a 2dim mobo and be done with this?

Can you name a single non-ITX AMD AM5 motherboard with only 1DPC?

For that matter any non-ITX Intel motherboards with worldwide availability that are 1DPC and don't cost $600+?

(with the obvious caveat of them needing to be competent for overclocking as there are undoubtedly cheap boards that this was a cost cutting measure on)

Midrange 1DPC boards just don't exist anymore despite it being a better choice for most people buying today. I'm not sure how much market penetration this will end up getting (I believe the cheapest of the Asus boards with it now is $500) in the near term, and I'd rather have 1DPC options, but this seems like a net positive in the 2DPC world we live in.

 

[TheHerald]

It somewhat limits the selection, if you're restricted to just 2 slot boards. Plus, for people who either are stuck or insist on buying a 4 slot motherboard, at least this lessens the downside.

Well, demand drives plurality. Meaning if people were opting for 2 dims there would be more of them. Currently most of them are super expensive with the exception of MSI unifyi (super good memory overclocker) but that's an matx board.

I never saw the point of 4 dims. By the time you are going to need to upgrade your ram your current kit will be so freaking outdated and the prices of ram would be so low that you could just buy a new kit with twice the capacity.

 

[TheHerald]

Can you name a single non-ITX AMD AM5 motherboard with only 1DPC?

For that matter any non-ITX Intel motherboards with worldwide availability that are 1DPC and don't cost $600+?

(with the obvious caveat of them needing to be competent for overclocking as there are undoubtedly cheap boards that this was a cost cutting measure on)

Midrange 1DPC boards just don't exist anymore despite it being a better choice for most people buying today. I'm not sure how much market penetration this will end up getting (I believe the cheapest of the Asus boards with it now is $500) in the near term, and I'd rather have 1DPC options, but this seems like a net positive in the 2DPC world we live in.

They don't exist on amd cause they are pointless, you are not going above 6400 anyways, and 4 dimmers are perfectly capable of those speeds.

Yeah I agree they are expensive but that's because 2dims come with a bunch of other "oc / gaming features" or what have you. Realistically shouldn't a 2 din mobo be cheaper than a 4dim if everything else is equal?

 

[thestryker]

They don't exist on amd cause they are pointless, you are not going above 6400 anyways, and 4 dimmers are perfectly capable of those speeds.

Not exactly true as TPU showed there is potential using low latency kits even with the mismatched ratios. Before AMD unlocked this capability it absolutely was true, but it will take someone testing scaling on an ITX board or one of these to see if even higher speeds/lower latency are usable.

Of course if AMD revamps the memory controller for a future CPU on the platform it'd be beneficial there, but I know it would be in the best interest of motherboard manufacturers to make people buy another new board then.

Yeah I agree they are expensive but that's because 2dims come with a bunch of other "oc / gaming features" or what have you. Realistically shouldn't a 2 din mobo be cheaper than a 4dim if everything else is equal?

MSI has (or maybe had) the Z790 MPower which was supposed to launch for around $200. It didn't have great VRM, but should have been sufficient for a tuned CPU and rather high speed memory. Asus has the Z790 AYW OC which is similar, but also not available. It's absolutely possible to make 1DPC boards and have them be affordable, but they're just not.

 

[TheHerald]

Not exactly true as TPU showed there is potential using low latency kits even with the mismatched ratios. Before AMD unlocked this capability it absolutely was true, but it will take someone testing scaling on an ITX board or one of these to see if even higher speeds/lower latency are usable.

Of course if AMD revamps the memory controller for a future CPU on the platform it'd be beneficial there, but I know it would be in the best interest of motherboard manufacturers to make people buy another new board then.

MSI has (or maybe had) the Z790 MPower which was supposed to launch for around $200. It didn't have great VRM, but should have been sufficient for a tuned CPU and rather high speed memory. Asus has the Z790 AYW OC which is similar, but also not available. It's absolutely possible to make 1DPC boards and have them be affordable, but they're just not.

The problem is with current amd chips you have negative scaling until you hit some extremely high speeds, at that point you might start seeing improvements but it's not really worth the problem, you might as well save the money and the trouble (tuning stability testing etc) and stick to the beaten path. With intel, since you are running 2:1 imc at any speeds, it's a linear increase in performance all the way which makes it a bit more worth it.

 

[bit_user]

I never saw the point of 4 dims. By the time you are going to need to upgrade your ram your current kit will be so freaking outdated and the prices of ram would be so low that you could just buy a new kit with twice the capacity.

In the distant past, I bought a PC with a 4-slot board, only populating 2 channels. Later, when I needed more RAM and it was cheaper, I populated the other two. I think a fair number of people do this, and it holds the promise that you don't even need to start out with the highest capacity DIMMs, thereby enabling further cost savings.

A lot of people aren't performance mavens, like yourself. They just want a machine that performs reasonably well and just add RAM when they become aware that it's needed (i.e. either due to swapping or because new games or software require it).

Then, there are people with a legit need for as much RAM as possible. These days, they might be doing stuff with AI or running lots of VMs.

Anyway, I have more recently opted for 1DPC configurations, now that I'm more informed about the tradeoffs of going 2 DPC. It also helps that RAM capacities have increased faster than my need for it.

 

[Peksha]

Can you name a single non-ITX AMD AM5 motherboard with only 1DPC?

Asus x670E Gene

 

[thestryker]

Asus x670E Gene

You're right there was a single board released that was discontinued shortly after the launch which killed its worldwide availability.

 

[cyrusfox]

What the article said seems to contradict this:

​

"Asus's press release claims that NitroPath slots have a 57% higher retention force than standard ones and 20% improvements in lateral latching and release forces. This increased toughness should make a difference for gamers or other enthusiasts who often replace DIMMs in their systems"​

​

It would be nice to get some hard stats on how they compare.

That would be great, I was basing this solely of the illustration showing the contact points, it appears it is now at the end of the lead rather than the side, and regardless of material hardness, I would expect sharper contact angles/ interfacing on a corner edge rather than the side I thought would reduce the rated insertion cycles for the module.

With some DIMM slots only rated for 15 insertions, the module isn't usually the limiting factor compared to the housing and retention of the dimm slot. I do wonder about LPCAMM2, perhaps that really is the way we should all go...

 

[thestryker]

That would be great, I was basing this solely of the illustration showing the contact points, it appears it is now at the end of the lead rather than the side, and regardless of material hardness, I would expect sharper contact angles/ interfacing on a corner edge rather than the side I thought would reduce the rated insertion cycles for the module.

With some DIMM slots only rated for 15 insertions, the module isn't usually the limiting factor compared to the housing and retention of the dimm slot. I do wonder about LPCAMM2, perhaps that really is the way we should all go...

If you watched the video you'd see that this graphic isn't accurate for the actual plug. Roman suggested that Asus may have purposely obfuscated what the new retention mechanism looks like to prevent imitators from hitting the market early.

Spoiler: video screenshot

 

[r00tb33r]

So why not use termination modules in current boards?