I really like that Ali isn‘t stretching his videos, if it‘s not necessary, Straight up Information and that’s it! And as always great shots and editing :)

Relevant video from Gamer's Nexus also showing performance improvement of 4 DIMMS versus 2 DIMMs. ( kzfaq.info/get/bejne/Y7ubethnppOao4U.html ) For reference, both the 2133 and 3200 configuration here were tested with 4 DIMMs, the 4000MHz kit was tested with 2 (4000MHz C15 is only available in 2x8). So potentially there would be a slightly larger performance difference of 4000MHz if across 4 DIMMs. The recommendation still stands though, 3200-3600MHz kits around C15 will get you most of the way there.

Thanks for Also showing some manually tuned secondary timings, I got a few percent improvement in games (like 4-5% or so) when going from 3000CL16 to a new kit 3800CL15 with manually tuned subtimings. For me tuning ram is fun, so good to see it'll have some effect as well for Ryzen 5000!

This is what I have been waiting for, thanks Ali 👍

Thank you for running these tests, this will help me immensely when purchasing new memory.

The information I was looking for from start to finish, thank you!

Apparently you can get (up to) ~8% more performance in games when using 4 RAM sticks or 2 dual-rank sticks (most 16GB sticks are dual-rank).

that B roll of the RAM holding one another up by the notch gave me a n x i e t y

i like the fact you put 2600 there, so i have really accurate feeling of the scale. love my cpu.

Thanks a lot. I've been waiting for this

First one to the memory speed test on Zen 3, well done and thank you!

Nice! A video showing a step by step memory overclocking would be very helpful for nubes like myself! As well as overclocking the infinity fabric!

Love the content, keep it up! Love your approach for all your recommendations, having price/performance being very important 👌👏👍

These are the tests we're looking for. Well these and maybe ones that go even deeper into RAM confugurations (2x vs 4x, single vs dual rank, GDM on and subtiming tweaking) and the Dynamic OC feature from ASUS. Thank you!

By God EXACTLY the answer I needed, instant sub.

You need to test memory configurations again when the 6000 cards come out and you can run a smart access memory system.

This such a useful video. A potentially complex problem boiled down to an easy answer, and a succinct explanation as to why.

Thanks so much, this is the exact thing i was looking for!

The content I was looking for, thanks!

Just wanted to say thank you for the info offered once again on your channel. You strike an excellent balance between the info and down to earth aspects of value for money related to costs benefit and returns.

I've been researching this and an aio cooler for 5950x ncase build. Your channel is a goldmine of quality content!

I discovered the performance gain while building a Ryzen 2200G PC back a few years ago. It was a budget setup with only DDR4-2400 and when tested with two sticks and four sticks the FPS went up around 5 FPS which was huge for an APU. The ram was also ran at 2666 later on which pushed it to around 8 FPS total gain.

Answering so many questions in 5 minutes thank you so much!

Hey Ali, a lot of people (myself included) are building or upgrading around now, and with the new AMD CPUs out, I was wondering if we could see another motherboard roundup? You make some of the best itx roundups around, and I'd love to hear your thoughts! Cheers and thank you for all your amazing content.

Much needed review - thanks!

Dude, your videos are the most meaningful pc videos for me, no kidding, every other guy deviates left and right, showing shit no one wants to see. Great job, keep it up, you just saved me $311

Great video and interesting results. Thank you!

Thank You, this is what I waiting for.

Thanks for the straightforward great video!

Thanks for this review. Been a lot of talk about this since 5000 release and considering the cost of 4000 MHz I will stick with some slower ram.

I been trying to find out lmao, thanks

this exact video is just what I've been looking for, thanks!

Your channel rocks.! Always straight to the point, clear and concise.

Respect to you sir, for covering this info! You are the most useful tech channel.

thank you so much. I was wondering the same thing.

How about testing more usable memory sizes? I'd like to see tests that are more useful for workstations, such as having a total of 128 GB ECC memory.

5:25 saw the new pump and the 3080 now, can’t wait for the new vid

Perfect explanation. Significantly important for the moment.

i think the main difference is real time response from when the application opens and how fast your actions are registered. i used to have ddr4 2400 cl14 ram and now have 3200 cl14 ram and it seems that actions are just slightly faster on the new ram to where i can notice a difference. both kits were 16gb as well and i didn't think there was a major difference back then but now that I've actually felt the change i can say that it does actually matter. maybe not as much going beyond but it did at the lower end of frequency speed.

And the 3200 cl14 kit I bought in 2017 for Zen1 keeps on truckin'.

Thank you it was just the video i needed to begin looking for compatible kits for my next 5900x :)

very interesting! how would this fair though on applications like after effects and 3d modelling software, where you are constantly shoving things in and out of the ram?

For this test did you use one of the 4 approved mobos that AMD has everyone using for CPU benchmarks? I know many of the boards currently out require a BIOS upgrade to get optimized CPU performance from the Ryzen CPU's

TNX man. really useful straight up information without too much of the annoying loud Trap beats music on the background. Although it would be nice to see the red cat more often :)

Straight to the point. Thank you!

Ali, love the videos, best itx channel on YT. One small nitpick with this vid though, at 2:12 your explanation of why the fclk is half of ram speed is wrong. It's not because of dual channel but because ddr4 is double data rate, meaning information is sent to memory twice per cycle. Dual channel has more to do with the bandwidth than clock speed. Not a big deal but it might lead people to interpret things wrong if they ever get into 4 or 6 channel memory configs. Keep the itx content rolling, I have an ncase m1 because of your channel.

wow, you explaning is actually so relaxing to listen to, loved it! Very clear explanation!

this is very good, simple, to do point and informative as usual, nice

Man that’s such a helpful video like always.

I'm a huge fan of this video, when I was looking late october I was first looking at 400$ royals, then really thanks to people on facebook groups and discords really homed in on 3600 Tridents with 16-16-16-36 timings and c16. Back then we didn't have videos like this :)

Man this was such a relief to hear. I ended up purchasing 16gb Ripjaws 4000Mhz C15 and returning that kit for 32Gb Ripjaws 3600mhz C16. Same price, but with my ITX motherboard I’m limited to two Dim slots. Good to know it’s not a significant difference.

Very, very informative. Thanks for the info sir.

what a short and helpful video! :-D

Excellent review. Thank you very much!

Yeah I still use my 3200MHz CL16 boosted a bit to 3333MHz CL16 and it works like a charm. The boost gave me a hundred extra points or so in Cinebench.

Just subbed off this info, straight forward and to the point

Great video! thanks for testing.

THE review I wanted to see ! Finally someone that tests these new chips to the max of their ability !

Would you please do a studio tour? I'm fascinated by your great aesthetics in your videos and want to see where you record etc.

I think it is a good think that zen 3 cpus are not as effected by ram speed as older parts. It is good to know that even a 3200mhz ram will give you 95% of the performance compared to faster speeds. Some of the older parts, especially zen and zen+ ones, had some issues reaching higher speeds while also being very memory and latency-sensitive. Mine 3200mgz hynex kid could not reach advertised speeds on B450m-pro4 with 2600 and I had to settle for 2667mhz with tighter timings.

THANKS!! SHORT AND DIRECT!

Can you do another build with all the new hardware that has been coming out?

Awesome! Many thanks ali!

THIS is the information I was looking for and that no reviewer was giving ! Absolutely nowhere in amd's promotional material did they say anything about increasing the Fclock and I was going insane.

As someone who has 4x8GB 3200MHz CL14 already, this makes me happy.

Thanks for the video! I bought 16Gb of 3200Mhz memory when I built my Ryzen 2700 system on a Gigabyte x470 MB. Nice to know I can drop in a 5600x - 5900x without concern for memory!

The man has spoken 👍🏼 nice video like always bro

Ali, will you be taking a look at the JOAT 14.9 SFF case by SpyLabs? I'm interested in what you think about it - an ITX case that allows for an NH-D15 to fit seems like a perfect candidate for a build I want to do early next year. (The other option I was considering was the Lian Li O11-D Mini, but that's a hell of a lot bigger, and not exactly the direction I want to go.)

Just watched Gamer's Nexus testing different kits. Steve's also talking with Wendell from Level1Techs. The best option seems to be 2x16GB Dual Rank or alternatively 4x8GB Single Rank kits. In contrast, with a 2x8GB Single Rank kit, no matter what speed and latency, you'll going to be relatively slower.

What about in productivity applications? Most, if not all of the benchmarks used in this video were gaming-focused, and it would have been really interesting to see if something like a Blender render, or Firefox compile would have benefited significantly more from the FLCK improvements.

Thanks for this video. Glad I just went with a 3600mhz CL16 kit. (Still went for the Trident Z Royal Silver RAM though 😅)

Great video as always!

Built a Ryzen 1700 with 3200mhz c14 in summer 2017, just threw in a 5600x into the old mobo yesterday for £200 thanks to a bios update and now lets me run the same ram kit at 3666mhz c16 nice and stable, awesome value upgrade for a 5 year old build!

Thanks for theminfo, but you only did optimizing for the 4000 kit. Maybe you should have done the same for the 2133 and especially the 3200 as well. How much could you gain from optimizing the 3200 kit?

Thanks for the great video. Could u please test 2 to 4 ram sticks? So 2133mhz 2x vs 4x vs 3200mhz 2x vs 4x 4000mhz etc etc. Would help a lot

Great info, thanks Ali! Now to try and actually buy a 5600x...

Picked up a 5800X at launch and some 3600mhz dominator sticks (cos budget). As soon as I saw that thumbnail I was all "oh what have I done....." That AMD slide about "4000 being the new 3600" had me worried. Well, worry no more, you're here to put our minds at ease! Thank you, this is just what was needed.

Not bad getting your fix of OT on a Sunday.

You make so good videos so you convince me to enable adverts in your channel :)

wanted to get the 4000Mhz ones, but could not find them available anywhere, so i went for some 4x8 Gb gskill 3200 C14. really glad that i did so :)

Such a relief that my newly bought 3200 cl14 ram will do the trick. Thank you!!!!

Not even 1 second into the video and all i see is vascularity... Ali jacked as ever!

@op The current theory is that memory rank is just as and potentially even more important than memory clock. All the kits you have linked in the description are single rank. With two sticks, that's 2 ranks of memory used when people have reported a boost of nearly 10% when using 4 ranks. I would be very interested in seeing a followup video with double the amount of sticks to see how it compares with 4 ranks. Something else I have noted in the video is that your benchmarks shows 32GB of memory but the B-roll shows only 2 sticks of 8 GB on the board. What's up with that? Did you actually test with 2 or 4 ransk?

Great topic, thank you 👌

I saw the GN video talking about 2x 8 vs. 4x 8 for performance, but haven’t yet seen any comparisons with memory intensive loads with larger configurations of RAM such as 2x 16 or 2x 32 (the latter which I’m considering for an editing rig), and would love any comments or thoughts on that. It seems that doubling the capacity to 64GB total makes sense as it’s less than double the cost of a 32GB kit and could save on upgrading later (cheaper to buy once than twice, even if the price of a 64GB kit drops later, it won’t go *that* cheap).

Brilliant!! I was about to OC my ram. But after watching your video, I won't. Thanks!

subscribing very straight forward -_- unlike the other tech vloggers ehem....

I have wanted to do a b550/ryzen 5000 build and have decided to go the high end route because of total cost of ownership. To me it makes sense that reaching the 1:1 ratio with fclk at 2000mhz gives more futureproofing to a build even if the benefits are slim at the moment. Any thoughts?

Thanks for the information really looking at new faster memory once I get my hands on a 5900x or 5950x but also as odd as it may seem the video by gamers nexus actually does show something strange happening with 4 sticks vs 2 sticks ram just as a thought.

I rlly do appreciate including the 3600

Curious about some other tests than standard render benchmarks, don't really have an idea of what they would be but I imagine RAM in general isn't super impactful on render times/game play past a certain amount. Maybe something that captures more everyday use speedups? It's not very l33t but browser speed tests/ application switching speed tests or something like that could also be indicative of increased RAM performance

Got a G.SKILL TridentZ Series 128GB kit on my current X99 machine, which I'm replacing for a 5950x. I think I'll just grab 64GBs of that one and put them in there, with timings of 14-14-14-34, I expect it's not worth spending ~1,000 CAD $ for what I would suspect to be a minimal 1-3% increase.

Thanks for testing

i see the graph, wait, where is int.. oh right...

I decreased my latency from 60ns to 57ns just by lowering the trfc value from default 600 to 350 with my bdie kit and 5900X. I should be able to get it even lower. Great vid!

verdict barely 1% unless you want the best, use what you want, very nicely explained

Nice, Thanks for making this video! I`ll hang on to my 3200mhz ram.

Modeling my next build off of the great content provided by Optimum Tech. Right now there is a lot of confusion about memory it seems. Keen to know what kits / factors cause these figures to pull ahead several percent as other reviews have demonstrated. like: 1)Are the 4 vs 2 DIMM variances actually just a matter of weather they are operating in single / dual rank. 2)Does ram on b-die offer potential for >7 FPS gains in some titles (assuming the "fabric" in the CPU can handle being pushed). 3) what can be achieved by combing the perks from the factors mentioned, with manual tweaks / timing optimisations. If I can spend an extra 100 / 150 on the right kit and manually tune it for up to 10fps more, I'd do it. This is because I have many, many problems.

This literally answered the questions I had about the new ryzen, I wanted to buy exactly the Gskill 3200C14 2x16GB Rams

There is a dual rank factor also, that would seem to more than negate the difference between low latency 3200 and expensive 4000 mhz memory. I think I'd rather not push the infinity clock and have to do a ton of micromanagement and lose stability for such a minimal gain. Glad I came here to see others discuss it. I just ordered my 5900X with some good Gskill Trident 3200 C14. (2 x 16GB)