Ryzen Above: Best Memory Settings for AMD's 3000 CPUs, Tested

[Crashman]

Right, I haven't been able to find ANY sticks over 8gb that is single rank, we just aren't there yet.

And I have only two ranks per channel, (sorry my initial reply says 4x16 but thats referring to ''# of sticks x amount of memory per stick'') they are 2Rx16, so two ranks per.

I've decided I don't need all 64gb (little overkill anyways?) so I'm going to try with these new four 8gb sticks that I ordered that are actually on the QVL- hyper x 3600. I was not aware that the ASRock x570m board was so particular with the QVL, very limited list. I noticed my RAM model # was off by a single letter, does this matter? Probably not... but it's easy to troubleshoot and also best to stick with manufacturers recommendations. PLUS the model number i DID find on the QVL list (that was off by a letter) should only be done with the 2 stick configuration....as I said I installed 4... and no i haven't tried just two sticks of my current RAM b/c i feel like at that point I'd rather just get new RAM thats on the QVL thats also supposed to be faster.

I only know that getting to the speed you desire is easier with two ranks than with four, and easier with four ranks than with eight, which is what you have. Heck, we only even test these boards with four single-rank modules (two ranks per channel).

 

[ldwilliams]

So we should get 4x8GB RAM for Ryzen? I see on the reviews for Ryzen 3, the official support for 4 dimms at dual rank is only 2667. Can anyone confirm is it easy or hard to buy 4x8GB of dual rank and use it at 3600? i will do it but if it is a lottery of working or not I want to know.

I have been building PC's for years but I never really got into overclocking other than XMP profiles. So i am willing to do this if it is easy enough to succeed. Thanks to anyone who helps explain this to me. I enjoyed this article but it was not as noob friendly as I needed.

I have a 3900x on a Gigabyte Aorus x470 gaming 7 wifi with latest Bios of course.
I have 4 x 8GB of Corsair Dominator Platinum 3200 cl16 OC to 3733 cl16 running nicely.
Tested with Prime95, Intel Burn, Aida64 and memtest64, so quite stable.
I know that they aren't dual rank but thought I would mention it

 

[M5C-Kong]

Is Ryzen 3000 optimized for DDR4-3200? What if we want more? We examine everything from frequency to rank count to nail down optimal settings.

Ryzen Above: Best Memory Settings for AMD's 3000 CPUs, Tested : Read more

This article talks to the enthusiast that wanted to get the last 1/10th of 1 percent of performance out of a build, but it does not talk about the value of the chips tested. If the 3600 16 CAS chip is less than 5% better performance than the 3200 14 CAS chip is it worth the price premium to get the higher frequency chip.

 

[cryoburner]

This article talks to the enthusiast that wanted to get the last 1/10th of 1 percent of performance out of a build, but it does not talk about the value of the chips tested. If the 3600 16 CAS chip is less than 5% better performance than the 3200 14 CAS chip is it worth the price premium to get the higher frequency chip.

Actually, looking at current prices, the lowest-cost DDR4-3600 CAS 16 kits cost less than the lowest-cost DDR4-3200 CAS 14 kits. : P

As for whether paying more for faster RAM is worth it in general though, I guess a lot depends on how much more the RAM costs relative to the overall cost of the system, and whether that money might be better put toward other components. If the faster RAM costs just $10-20 more and is being put into a $1000 system, that might be a reasonable option if it offers at least a couple percent more performance in many scenarios. If the faster RAM costs $50 more, for only a few percent better performance, then you are running into diminishing returns, and might be better off putting that money elsewhere in the system.

Another notable thing this article shows is that the capacity of the RAM kit can make more of a difference to performance than frequencies or timings, with a 32GB kit of moderate-performing RAM providing a bit more performance than 16GB of otherwise faster, low latency RAM, even if the extra RAM isn't being used, due the higher capacity utilizing 4 ranks instead of 2. So, given a choice between the two, a 32GB kit (either 2x16GB or 4x8GB) of standard latency RAM makes more sense than paying a similar amount for a 16GB kit (2x8GB) of low-latency RAM. Especially if the system is expected to be kept long enough to make use of that extra memory.

 

[cifroes]

I crunched the numbers for myself, focusing on gaming, specially 1440P and, to be honest, got very different conclusions.

I considered my base 100% numbers the 2x8GB, 3200MHZ, CL16, 2T as that seems like a good 'default' buy with current pricing.

F1 2017 @ 1440p ultra or medium basically doesn't care, everything is within 1%.

Ashes 1440p crazy - everything more or less the same (1% +-) except a positive for 2x8GB 3200MHz CL14 1CR (+3.7%) and a negative for 2x8GB 3600MHZ CL18 2CR (-2.18%).

Ashes 1440p high - A bit more differences here. 4x8GB configurations clearly win here. Going to 2x8GB 2800MHz is a major loss even at CL14. Other than that, at 2x8GB differences are very small.

Just for fun, 3DMark time spy overall score - doesn't matter except for some weird reason 4x8GB 2800MHz CL14 1CR that has the best score.

3dMark time spy graphics score - all the same except 4x8GB 2800MHz CL14 1CR that is 2.02% better.



My initial doubt about all of this DDR talk was to figure out what to buy for my 2400G that will be replaced at the end of year probably with a 3600. Focus is on gaming only, since, to be honest, PCs are already fast enough for other daily tasks (and that's my use case) My GPU is a RX 5700.

I have very similar prices for sticks like this (all 2x8GB):

• 3000 CL16 - 72€
• 3200 CL16 - 74€
• 3000 CL15 - 86€
• 3600 CL18 - 87€
• 3600 CL16 - 96€

It's very hard to get conclusions for gaming. Probably depending on games and your GPU they could get a bit more CPU bound and DDR makes a difference but my specific analysis the main takeaway is that nothing matters within these bounds

For anyone in particular it's 'impossible' to guess if this or that will benefit them. It seems it varies too much for Game, resolution, settings, GPU, probably even CPU. It's perfectly possible that in certain scenarios DDR can make a big diference, but with the published numbers it's impossible to know.

4x8GB seems to be helpful. Other than that, get 3200 CL16. If better ones are available for just a few bucks more (3600 CL18 or CL16) I would get those. Otherwise, just buy something approved for your motherboard, don't worry too much and enjoy them

For my specific case... I have a % discount on that store so the price differences between sticks are lower but I got even more confused and since all 3600MHz are out of stock I will get the 3200 CL16 Aegis with the benefit that it doesn't have tall heatsinks so no worries with cpu coolers. I would get the 3600 CL16 Ripjaws V but they are out of stock with no predicted date of arrival.

 

[cifroes]

Can't believe I missed this article.

If you are testing with DDR4 3200 CL 16 and DDR4 3600 CL 18...you are testing with the wrong RAM. Most informed enthusiasts are going to go for higher quality memory, especially now that prices have improved.

High performance Samsung B-Die or similar really benefits Ryzen and should spec out around DDR4 3200 CL 14 and DDR4 3600 CL 16. They also usually have better sub timings.

With the numbers from the article and I'm focusing on gaming 1440p it seems 3200 CL14 or CL16 doesn't really matter for performance. same for 3600 CL16/18. The differences are incredible tiny, probably within the margin of error.

I agree that it probably shows they are higher quality sticks and maybe the sub-timings are better and could make a difference but for someone that just wants better gaming performance with a >300$ GPU and no OC it just doesn't make a real world difference.

And the prices for 3200 CL14 and 3600 CL16 (samsung) seem to be a big difference, at least here in my country. If it's a few bucks more, fine, get it. If it's >20€ difference I would hardly recommend it.

 

[cryoburner]

With the numbers from the article and I'm focusing on gaming 1440p it seems 3200 CL14 or CL16 doesn't really matter for performance.

Yeah, at 1440p with an RX 5700-level card, I suspect these differences in RAM timings would make no perceptible difference. In most demanding games with the settings turned up, performance is likely to be limited by the graphics hardware more than anything. Even if there were a couple percent difference in performance in some game, that would hardly be noticeable anyway, and probably not worth paying substantially more for.

 

[jpe1701]

We've been doing Intel for years.
https://www.tomshardware.com/reviews/super-talent-project-x-f3000ux16g-ddr4-memory,5038.html
Did you want another article like "Does AMD really get a bigger boost from fast memory than Intel?"

Spoiler, it doesn't. All the myths and rumors about AMD needing fast memory more than Intel needs it are excuses and rationalizations dreamed up by AMD fanboys. Both platforms benefit in a similar manner, and I'm just happy that AMD now supports fast memory without jumping through a bunch of hoops to run it.

How do you figure it was some AMD fanboy when in essence it's saying AMD needs fast ram to compete with intel?

 

[Crashman]

How do you figure it was some AMD fanboy when in essence it's saying AMD needs fast ram to compete with intel?

They both benefit similarly from faster DRAM, therefore the idea that Intel needs it less is a fantasy.

Sound better?

 

[jpe1701]

They both benefit similarly from faster DRAM, therefore the idea that Intel needs it less is a fantasy.

Sound better?

Much better. Lol. I wish I had read this article before I bought my ram today. I bought the Ripjaws v 3600 c16 for$10 more than the c18 for my 3900x.

 

[maverik-sg1]

I am running 4x8GB patriot viper steel 1T 3600mhz 16-19-19-38-60 1.36v right now with some additional sub timings tweaked using the DRAM calculator.

CPU is 3900X, which via Ryzen master I set at 4.3ghz (forced BCLK at 100.2ish to stop RM from adjusting it) FCLK 1800mhz

Mobo Asus Crosshair VIII Hero WiFi

What I didn't do and what I am looking for is guidance on which sub timings present the best gains - was hoping to see the difference sub timings make, essentially to see where I should be focusing my attention for gains.

Any help/guidance/links for reading greatly appreciated.

 

[jrhansen]

Do you think using a motherboard with a T-topology memory trace layout would give different results compared to the MSI MEG X570 Ace board you used (it has a daisy-chain topology)?

Daisy chaining is supposed to work best with 2 sticks of ram. 4 Sticks might not be optimal. Thats what I have read.
I havn't yet found a detailed review of a board that isn't using Daisy chain. Its not something that the companies usually writes in their specs, so it propably takes a person that has the knowledge to see thid on the motherboard layout, unless they can get these informations direct from the companies.
I would asume that the daisy chain topology propably is cheaper and easier to implement, and its propably only something very few enthusiasts will be worried about if they know they want to go all in and put 4x 32gb sticks in their build and really want the absolute best possible performance/stability out of it.

I am building a Ryzen machine now, and I really think that if you just need the extra ram for certain non game situations, and if price still matters, wether you go with 3200 or 3600 and 2 vs 4 ram sticks don't matter that much. I will go with a MSi B450 Tomahawk Max, Ryzen 7 3800x and a single Coorsair CMK32GX4M2B3200C16 kit to start with. That way I can upgrade to 64gb pretty easy, and the board will also support a 3950 when they get cheaper. The board doesn't officially support 4 slots with those ram, but there is like 90% chance its just not been tested by either MSi or Corsair.

If someone actaully know of any mainstream Mobo's that uses a T-topology ram configuration it would be nice to know.

I know that the upcomming MSi X570 Tomahawk also uses Daisy Chaining

 

[Crashman]

Daisy chaining is supposed to work best with 2 sticks of ram. 4 Sticks might not be optimal. Thats what I have read.

That type of discussion leads to "maybe this design allows both two and four stick sets to run at DDR4-4266, while that one can run two sticks at 4400 but four at only 4000"...which according to the test findings is a mere distraction, unless your entire effort is to maximize frequency regardless of performance. Best performance is typically found at the highest clock for which you can run the memory controller and infinity fabric synchronously, and most users are finding that to be somewhere around DDR4-3733 to DDR4-3800.

And that's not to forget that two ranks per channel significantly outperform one rank per channel, or that you can get two ranks from either a dual-rank or two single-rank DIMMs.

 

[jrhansen]

What I mean is, if you are GPU bound for gaming with any current AMD 5xx or 55/5600 and your goal isn't gaming with that exactly game that gains 10-12% .. whats the point .. most people will be perfect happy with a 3200cl16 T1 memory config weather its 1,2 or 4 sticks of single or dual ranks, and I really believe that a pro user of eg Video editing or a Music dev system will most likely get his system prebuild with the fastest stock cpu/ram the MB will support. Pro users don't normally play "what if i can boost my perf. 2-3%" and risk my workstation becomes unstable, and Really. If you arent up at a Ryzen 9 and don't need 128Gb of ram for video editing, most users won't be able to tell a static test difference that shows under 10% on an avg Ryzen 3/5/7 gaming system. I have been there so many times with OC'ing different platforms, and I almost allways after trying how far I can push the hardware end up with running at somewehere around little over stock speed. It was a whole other thing in the old days when you could OC a Celeron 300 to 500Mhz if you got a good cpu. Today's OC margins are so low that its only for those that invest in serious cooling and primarily do it for fun. Don't think many of us more normal users want to spend the money and space on a liquid Nitrogene tank just to push a CPU more than those 10% you might be able to push it by simple bios OC'ing.

I get what you are trying to tell, and as I understand it the optimal possible scenario is a 3600 kit tweaked a bit to get a tad further than 2667

 

[okbud]

sir could you rerun this test with latest zen 3 cpus? hugh thanks

eg: the Why Do I Want Four Ranks? section