[SOLVED] How to tell if RAM is Dual Rank or Single Rank from product description?

[Smarticus]

Hi there - With the new AMD CPUs out, there has been a lot of talk recently about Dual Rank vs. Single Rank RAM.

What I'm having trouble with is being able to tell if RAM I'm about to purchase is Dual Rank. For some reason, they don't clearly label it in product descriptions, even though there seems to be a drastic difference in performance between the two. To figure it out, one answer I've seen is: "Multiply total number of chips (less any ECC chips) by the chip width. If it equals 64 its single rank. If it equals 128 its dual rank." The problem is, most descriptions dont list the 'chips' or ECC chips (that I know of) so I cant figure it out from that. I suspect that maybe all RAM is maybe Dual Rank these now so they dont have to label it? But if there is a way of telling based on the product description, can someone please let me know? Or, am I missing something?

Here is an example of a product description.

Here is an example of benchmarks dual rank vs single rank.

Thanks!

 

[Lutfij]

Dual rank and single rank was an issue with the initial run of motherboard's from AMD's camp. Now, with the latest AGESA microcodes, you're good to go with pretty much any ram there is on the market provided the baord has the support for the frequency and the ram capacity, without the need to look through the QVL of a motherboard.

Even if the part was listed on Amazon as dual or single ranked, they are often times mislabeled on Amazon. Also testing the ram out of the box and returning it isn't possible for all regions.

 

[Smarticus]

Not sure I understand. Are you saying, I should assume that RAM is Dual Rank if it's a relatively recent product? Not sure what you mean by 'your good to go with pretty much any RAM' when there seems to be a big performance difference between single rank or dual ranked. Is single rank not a thing anymore? If it isn't, why do I keep seeing single rank vs dual rank pop up in recent benchmark videos? Another example of 'ranks' coming up is in this recent GamersNexus review of 2 vs 4 sticks. You can see his first comment (and a lot of other comments are about dual ranked ram) and how it affected the results, so it doesnt seem like 'good to go with any RAM' is right, although it's highly likely I'm missing something or not understanding.

At the end of the day, I just want to get the optimal RAM. From what I've gleaned, it seems like the optimal RAM for my AMD 5000 build from a price to performance perspective, is 3600hz, CL16, Single Rank, 4-sticks OR 3600hz, CL16, Dual Rank, 2-sticks - but I'm no expert and it seems complex if you get into the weeds. Just having trouble telling what RAM is dual rank and dont want to buy the wrong product.

 

[jtk2515]

Not sure I understand. Are you saying, I should assume that RAM is Dual Rank if it's a relatively recent product? Not sure what you mean by 'your good to go with pretty much any RAM' when there seems to be a big performance difference between single rank or dual ranked. Is single rank not a thing anymore? If it isn't, why do I keep seeing single rank vs dual rank pop up in recent benchmark videos? Another example of 'ranks' coming up is in this recent GamersNexus review of 2 vs 4 sticks. You can see his first comment (and a lot of other comments are about dual ranked ram) and how it affected the results, so it doesnt seem like 'good to go with any RAM' is right, although it's highly likely I'm missing something or not understanding.

At the end of the day, I just want to get the optimal RAM. From what I've gleaned, it seems like the optimal RAM for my AMD 5000 build from a price to performance perspective, is 3600hz, CL16, Single Rank, 4-sticks OR 3600hz, CL16, Dual Rank, 2-sticks - but I'm no expert and it seems complex if you get into the weeds. Just having trouble telling what RAM is dual rank and dont want to buy the wrong product.

The best ram for zen 3 is the lowest cl ram that is 1 to 1 with your fclk ( infinity fabric) 4 ranks. So if you get some God like bin and can run 2000fclk then 4000mhz is the best.

Most people are only able to get 1900 fclk so most are recommending 3800mhz. Then get the lowest sub timings you can afford. Ryzen still likes low subtimmings so 3200mhz cl15 will be very close to 4000mhz cl20. There is a nanosecond calculator for mhz and cl somewhere if you are compareing two meomory speeds and cl.

I see alot of people getting samsung b die meomory and overclocking it and reduceing sub timings while running 1.45 volts.

It really depends what your budget it.

Almost all 8gb sticks are single rank. 16 gb sticks could still be single rank even with chips on both sides. You will have to ask the manufacturer or someone whos has bought them before.

 

[Smarticus]

From the benchmarks I've seen in this vid, it looks like 4000 and 3800 are better, of course, but only by like 5 frames or so in most cases, so its not worth the cost difference in my opinion. 3600 looks like it makes sense to me. Thanks for explaining about most 8gb sticks single rank and some 16gb sticks. It seems crazy that they wouldnt list single/dual rank on the product description as it can affect performance.

View: https://www.youtube.com/watch?v=AGux0pANft0&t=3s&ab_channel=HardwareUnboxed

 

[alceryes]

Ryzen still likes low subtimmings so 3200mhz cl15 will be very close to 4000mhz cl20.

What performance metric are you referring to here? Because DDR4 4000MHz cl20 will blow DDR4 3200MHz cl15 out of the water in synthetic memory benchmarks.

If you are referring to gaming, then the first thing one should realize is that, with a tuned system (no major issues), the difference will be very minor between the two at 1080p and even less at higher resolutions. You're looking at a consistent 0-3fps difference, maybe peaking at 3-5fps, depending on the game.

However, in the gaming benches that I've seen, DDR4 4000MHz still wins by a few fps.

 

[jtk2515]

What performance metric are you referring to here? Because DDR4 4000MHz cl20 will blow DDR4 3200MHz cl15 out of the water in synthetic memory benchmarks.

If you are referring to gaming, then the first thing one should realize is that, with a tuned system (no major issues), the difference will be very minor between the two at 1080p and even less at higher resolutions. You're looking at a consistent 0-3fps difference, maybe peaking at 3-5fps, depending on the game.

However, in the gaming benches that I've seen, DDR4 4000MHz still wins by a few fps.

That would be the total nanoseconds of the memory module. It is good for a approximation.

4000mhz would be faster if you can get your IF to 2000mhz. If you can only get your IF to 1900mhz then it would be faster to run your memory at 3800mhz instead of 4000mhz, because not running 1 to 1 adds latency.

 

[alceryes]

Contrary to popular belief, total latency of RAM can not be determined by CL alone.
I agree that you'll want to run Infinity Fabric at 1-to-1 for optimal performance, but, taking IF out of the equation, you should always go for faster RAM over tighter memory timings.

Here's a good read on memory latency, cycles, and timings -
https://www.crucial.com/articles/about-memory/difference-between-speed-and-latency

(note that there's an error in the graph that showed up recently. The second and third to last Latency's should read 14.32 and 14.25 respectively. You can see this by looking at older versions of the site on archive.org)

 

[jtk2515]

Contrary to popular belief, total latency of RAM can not be determined by CL alone.
I agree that you'll want to run Infinity Fabric at 1-to-1 for optimal performance, but, taking IF out of the equation, you should always go for faster RAM over tighter memory timings.

Here's a good read on memory latency, cycles, and timings -
https://www.crucial.com/articles/about-memory/difference-between-speed-and-latency

(note that there's an error in the graph that showed up recently. The second and third to last Latency's should read 14.32 and 14.25 respectively. You can see this by looking at older versions of the site on archive.org)

But you cannt take IF out of the equation.
Overclock your IF, see if you can get 1800mhz,1900mhz or 2000mhz fclk.

Then you want your memory to match your fclk with the best sub timing you can afford.

The whole point of using a nanosecond calculator is so you can make a approximation when you are compareing different memory modules to buy if you are taking price into the equation.