School me on DDR5 v.Single Rank vs. Dual Rank

[Mister Woof]

hey fellas,

 

was looking at DDR5 kits and I am not sure if this is single rank or dual rank.

 

I know in some applications and scenarios, there is some small (but often meaningless) difference between single rank and dual rank DDR4 - and that when running 4x single rank DIMMS the system often will behave as if the dual channel memory is dual rank.

 

Can anyone tell me if this behavior still exists on DDR5? Specifically, this kit, is this DDR5 single or dual rank? Does it matter for DDR5?

 

https://www.amazon.com/gp/product/B09Y1FMSWS

 

Additionally, other than the RGB (which is priceless!) is this kit worth the extra?

 

https://www.amazon.com/gp/product/B09QS2K59B

 

For more information, thinking of combining 13600KF + Gigabyte Z690 Aorus Elite AX DDR5. 

 

The $50 difference perhaps worthwhile to jump to i7-13700KF instead and keep 6000mt/s

[RONOTHAN##]

With DDR5 at the moment, it's very simple whether a kit is single rank or dual rank: if it's an 8GB module, it's 1Rx16 (kinda like half rank, should be avoided if possible), if it's a 16GB stick it's single rank, and if it's a 32GB stick it's dual rank. This will change in the future when higher density memory ICs release, but since every DDR5 memory IC is 16Gb at the moment those are the only possible ways to organize a DDR5 stick for their respective capacity.

 

DDR5 right now is behaving like early DDR4 with its rank interleaving. Dual rank does have a small performance improvement, but the decrease in max frequency is so severe that single rank will give significantly better performance. If you're only going for a speed that should work at dual rank frequencies (6000MT/s should work on both single and dual rank setups), then yeah, dual rank will have a very small performance uplift compared to a dual rank setup, though realistically it's not worth the extra unless you need the extra capacity. 

 

So to answer your questions in order: Technically but barely, single, and not really. 

[Mister Woof]

2 minutes ago, RONOTHAN## said:

With DDR5 at the moment, it's very simple whether a kit is single rank or dual rank: if it's an 8GB module, it's 1Rx16 (kinda like half rank, should be avoided if possible), if it's a 16GB stick it's single rank, and if it's a 32GB stick it's dual rank. This will change in the future when higher density memory ICs release, but since every DDR5 memory IC is 16Gb those are the only possible ways to organize a DDR5 stick for their respective capacity.

 

DDR5 right now is behaving like early DDR4 with its rank interleaving. Dual rank does have a small performance improvement, but the decrease in max frequency is so severe that single rank will give significantly better performance. If you're only going for a speed that should work at dual rank frequencies (6000MT/s should work on both single and dual rank setups), then yeah, dual rank will have a very small performance uplift compared to a dual rank setup, though realistically it's not worth the extra unless you need the extra capacity. 

 

So to answer your questions in order: Technically but barely, single, and not really. 

Thanks for the response. 

 

My main game is Final Fantasy 14, which although not representative of in-game play, GN did a review of these chips in the built-in benchmark.

 

 

 

 

The 13600K with 6000 CL30 (which I presume would perform similarly to 6400 CL32?) pulls in this test around 43% better performance than my 10900K w/3090ti (which is obviously faster than my existing 3080, but often this game encounters CPU limitations not GPU limitations)

 

Regarding waiting for a Zen4 3D part, in this test, we see the 5800X3D only manages around 15% more performance than the 5800X-non3D; it would seem if the 7800X3D has similar gains then performance should be a wash in this title between a 13600K and a 7800X3D as a guess. Then there's some overclocking still available in the tank for the 13600K, which seems like there are gains on the P-cores since this is mostly a single-threaded game. It seems most of the delta between the 13th gen Intel stack is clock frequency.

 

ANYWAY, what would you buy? I can't be sure if FF14 is memory sensitive, but my gut says the faster memory from DDR5 will help since it's frequently CPU limited. Otherwise I would carry over my dual-rank behaving 4x8 4000/CL18 kit.

 

I'd either get:

 

13600KF/Z690/32gb 6000/CL32

 

13700KF/Z690/32gb 6000/CL32

 

Reusing the DDR4 from my existing system would break it apart from that, so I would probably need more RAM anyway; although if I carried over the 4x8 4000/CL18, I'd end up buying some bottom bin 16b ddr4 kit for the 10900k......

What would you do?

 

 

[PDifolco]

2 minutes ago, Shimmy Gummi said:

Thanks for the response. 

 

My main game is Final Fantasy 14, which although not representative of in-game play, GN did a review of these chips in the built-in benchmark.

 

 

 

 

The 13600K with 6000 CL30 (which I presume would perform similarly to 6400 CL32?) pulls in this test around 43% better performance than my 10900K w/3090ti (which is obviously faster than my existing 3080, but often this game encounters CPU limitations not GPU limitations)

 

Regarding waiting for a Zen4 3D part, in this test, we see the 5800X3D only manages around 15% more performance than the 5800X-non3D; it would seem if the 7800X3D has similar gains then performance should be a wash in this title between a 13600K and a 7800X3D as a guess. Then there's some overclocking still available in the tank for the 13600K, which seems like there are gains on the P-cores since this is mostly a single-threaded game. It seems most of the delta between the 13th gen Intel stack is clock frequency.

 

ANYWAY, what would you buy? I can't be sure if FF14 is memory sensitive, but my gut says the faster memory from DDR5 will help since it's frequently CPU limited. Otherwise I would carry over my dual-rank behaving 4x8 4000/CL18 kit.

 

I'd either get:

 

13600KF/Z690/32gb 6000/CL32

 

13700KF/Z690/32gb 6000/CL32

 

Reusing the DDR4 from my existing system would break it apart from that, so I would probably need more RAM anyway; although if I carried over the 4x8 4000/CL18, I'd end up buying some bottom bin 16b ddr4 kit for the 10900k......

What would you do?

 

 

 

I'd get  13600K/DDR4 Z690/Your existing RAM 4000CL18

You don't need the extra cores of the 13700K, as the 13600K is imo th ebest price/perf choice for gaming

And if you already have decent DDR4 there's not much point in getting still expensive DDR5 for a very slight performance boost (a few %)

[Mister Woof]

3 minutes ago, PDifolco said:

 

I'd get  13600K/DDR4 Z690/Your existing RAM 4000CL18

You don't need the extra cores of the 13700K, as the 13600K is imo th ebest price/perf choice for gaming

And if you already have decent DDR4 there's not much point in getting still expensive DDR5 for a very slight performance boost (a few %)

I'd end up buying some kind of RAM anyway to use with the 10900k, although 16gb 3200 is pretty cheap.

[RONOTHAN##]

1 minute ago, Shimmy Gummi said:

6000 CL30 (which I presume would perform similarly to 6400 CL32?)

Not exactly. DDR5 primary timings are almost irrelevant, so the 6400MT/s will always be faster. Only by about 6% at the absolute maximum, in a purely memory bottlenecked situation, but it will be faster. 

 

4 minutes ago, Shimmy Gummi said:

I can't be sure if FF14 is memory sensitive, but my gut says the faster memory from DDR5 will help since it's frequently CPU limited.

A good rule of thumb for whether a game is very memory sensitive is to see how far ahead the 5800X3D is over the 5800X. Because of the 3D VCache, memory bottlenecks almost never happen, so if it increases performance significantly, the 5800X was memory bottlenecked. Therefore, it looks like FF14 isn't the most memory sensitive game out there, but it's definitely memory sensitive so a higher end kit will help. 

 

9 minutes ago, Shimmy Gummi said:

ANYWAY, what would you buy?

I mean, realistically I'd be buying a 5600MT/s CL40 kit and overclocking it to 6400MT/s or higher (every 5600MT/s CL40 kit will be able to do 6200MT/s on a good board and CPU, and if you get lucky 7000MT/s is possible), but I know that is not a thing that everyone would want to do, and odds are you're just gonna enable XMP and call it a day. If you're going for the Z690 Aorus Elite, I'd stick to the 6000MT/s kit for better memory compatibility (that board was pretty famous for struggling to get memory speeds above 6000MT/s, not sure if it's gotten better with Raptor Lake, but it couldn't even do it with a CPU on LN2, so I don't have much hope for it). If you go for a board with better memory support, I'd spend the extra $50 on the 6400MT/s kit, but I wouldn't blame you saving the $50 instead, it's not that big a deal. 

[Mister Woof]

19 minutes ago, RONOTHAN## said:

Not exactly. DDR5 primary timings are almost irrelevant, so the 6400MT/s will always be faster. Only by about 6% at the absolute maximum, in a purely memory bottlenecked situation, but it will be faster. 

 

A good rule of thumb for whether a game is very memory sensitive is to see how far ahead the 5800X3D is over the 5800X. Because of the 3D VCache, memory bottlenecks almost never happen, so if it increases performance significantly, the 5800X was memory bottlenecked. Therefore, it looks like FF14 isn't the most memory sensitive game out there, but it's definitely memory sensitive so a higher end kit will help. 

 

I mean, realistically I'd be buying a 5600MT/s CL40 kit and overclocking it to 6400MT/s or higher (every 5600MT/s CL40 kit will be able to do 6200MT/s on a good board and CPU, and if you get lucky 7000MT/s is possible), but I know that is not a thing that everyone would want to do, and odds are you're just gonna enable XMP and call it a day. If you're going for the Z690 Aorus Elite, I'd stick to the 6000MT/s kit for better memory compatibility (that board was pretty famous for struggling to get memory speeds above 6000MT/s, not sure if it's gotten better with Raptor Lake, but it couldn't even do it with a CPU on LN2, so I don't have much hope for it). If you go for a board with better memory support, I'd spend the extra $50 on the 6400MT/s kit, but I wouldn't blame you saving the $50 instead, it's not that big a deal. 

I'm not married to that board, but it is o lay $199, and was one of the few z690 ddr5 boards with flashback in that price range.

 

Wanted the Asus TUF z690 actually even if a little more to match my 3080, but it doesn't seem to support flashback

[RONOTHAN##]

2 minutes ago, Shimmy Gummi said:

I'm not married to that board, but it is o lay $199, and was one of the few z690 ddr5 boards with flashback in that price range.

 

Wanted the Asus TUF z690 actually even if a little more to match my 3080, but it doesn't seem to support flashback

Fair enough. If you can afford it, I'd probably go for a Z790 board instead so you get a 2nd generation DDR5 memory topologies, but if you're sticking to 6400MT/s or lower it should be fine no matter which way you lean. 

 

For a board, I'd probably go Z690-A Pro, it's in the same price range (~$180 US), has BIOS flashback, and has a better DDR5 memory topology than Gigabyte on Z690. 

[Somerandomtechyboi]

1 hour ago, RONOTHAN## said:

DDR5 primary timings are almost irrelevant, so the 6400MT/s will always be faster. Only by about 6% at the absolute maximum, in a purely memory bottlenecked situation, but it will be faster. 

A bandwidth bottlenecked application you mean, 6% at absolute max but youll be closer to 2% in gains since ram speed just dont matter that much and latency is also something alot of software cares about

 

Wouldnt bother wasting money on higher bins and just buy a 5600 bin regardless of garbage cl timing so you can oc to 6600+ in the future when platforms start needing faster rams

[Mister Woof]

Well, my better judgment escaped me last night and I ended up ordering the Z690 Aorus Elite AX, the cheaper 6000mt/s G Skill kit....and the i7-13700KF 

 

 

[Deadpool2onBlu-Ray]

On 10/24/2022 at 2:05 PM, Shimmy Gummi said:

Well, my better judgment escaped me last night and I ended up ordering the Z690 Aorus Elite AX, the cheaper 6000mt/s G Skill kit....and the i7-13700KF 

 

 

How did that board work out? There a bunch of horror stories on Newegg/Amazon reviews for that board

[Mister Woof]

12 minutes ago, CHICKSLAYA said:

How did that board work out? There a bunch of horror stories on Newegg/Amazon reviews for that board

We'll find out on Saturday hopefully lol.

 

I have been using gigabyte for a long time with success... currently on the Z590 Aorus Elite AX with no issues 

[Mister Woof]

Well I canceled the motherboard. Seems it has really poor memory support. Looking at the QVL list for my ram, what would you choose?

 

https://www.gskill.com/qvl/165/377/1649665818/F5-6000J3238F16GX2-RS5K-F5-6000J3238F16GA2-RS5K-QVL

 

Not many in-stock options for Z790. It has to have wifi as well.