@@blackbirdpctech Informative video; and while I feel the fps/$ metric on the cost of the product is important, the overall cost of the system is as well. If I were building a $2000+ system like what is shown in this video, and I were down to choosing my RAM, it's no longer a 95% price increase for ~4% performance, but more like a 5% increase for 4% performance. It probably still isn't the most efficient use of your money of course. Generally, x3d does more for games VS RAM speed, and core count does more for production. This applies even more so to people who might want more than 32 GB of RAM (of course that is only for crazy people like me who like to have 50 chrome tabs open while they run a video encoder while playing a game and watching a RUclips stream). Much of this comes from the fact that RAM has long since hit a point of diminishing return on raw throughput, and now latency is king (it takes about a 10% or more boost in throughput to match the performance gained from a 2% reduction to total memory latency) and higher speed memory kits don't generally improve much in the relative latency department (CL30 6000 is the same 10ns first word latency as CL40 8000). And now I've spent way too much time rambling off topic to write a nearly 250 word essay/lecture that will be read by like 3 people.
@@credge nice comment and interesting take on fps/$ ... my only concern with using total system cost is that you can usually achieve the same performance with vastly different cost builds, so I don't want to mislead people in to thinking something is better or worse than what it is simply because of the components I chose. For example, if I selected an MSI X870e Godlike the total system cost would be high, versus an MSI X670e Tomahawk which would be much lower ... the performance of the system wouldn't change, but the price would. I need to look at latency again because the last time I looked at it the gaming performance didn't really change.
@@blackbirdpctech I don't mean to imply that total system cost should trump RAM cost when buying RAM, because it shouldn't, and you can't be expected to include it on your video anyway, since every viewer has a different setup. I didn't really mean it as a criticism against your video, but more something that people should consider for themselves on a case-by-case basis. Because if I'm building an already top-tier system, then do I really want to lose 5% performance to save $100 on my $2,500 system? Probably not. But if it's an $800 system, then that $100 should go towards something else, like a better CPU or GPU. But the 6400 kit on a $1,200 system? Now it's 1% cost for ~3% gain (overall, but also only at 1080p). Do I feel the conclusion that you should not buy the more expensive memory based on the fps/$ metric is a bit binary? Yeah, but I would also agree that the 8000 kit would not be worth it for the vast majority of setups; however, the 6400 kit on the other hand would have it's place in some builds (being only $12 more expensive instead of $107. But this assumes you're building a whole new computer, I would never recommend paying full price to replace an existing 6000 kit without a really good reason. I do have to admit that my latency assertion could be behind the times as the time I discovered the latency thing was when looking at DDR4 benchmarks: where a 3800 kit lost to a 3200 kit, but a 3600 kit beat them both, but only beat the 3200 by a small amount; and then when I divided out the access time, I found that the 3800 kit had worse access time, while the 3600 and 3200 were very close on access time (tiny bit in favor of the 3200) but the 3600 got a small edge in overall performance because it had a close enough access time but significantly higher throughput, but the 3800 kit could not compensate for it's inferior access time with even more throughput. It was on that video that I did some calculations and came up with a rough estimate that, on average, it took a 10% bandwidth increase to match the performance of a 2% reduction in latency on DDR4 (it did vary from game to game, with some games actually preferring the 3800 kit). But again, this is DDR4 memory, not DDR5. However, on your DDR5 kits: If you calculate for total latency in time (rather than clock cycles) on the four latency numbers on the 3 kits (I'll label them #1-#2-#3-#4). 8000 is better for #1, all 3 kits are functionally tied for #4; but for #2 and #3 the 6400 and 8000 kits are basically tied but both better than the 6000 kit. And while the 8000 kit wins overall, the 6400 has closer performance to the 8000 than it does to the 6000 kit, even when you include the extreme statistical outlier of MS Flight Sim, this makes me think that total latency is doing more work than the actual throughput of the ram, though the throughput is still relevant. And of course none of my observations take into account the MCLK:UCLK:FCLK relationship. Like how much do you lose when you go from 1:1 to 2:1 MCLK:UCLK; how much do you gain going from syncing up the UCLK and FCLK? It could be that I am completely wrong and that those are making more difference than anything else (though the benchmarks do still support the idea of the 6400 kit having a place in some mid-range or higher builds). Games do the same thing with load times as well (sequential read speeds help, but random reads are more important), where a 7 GB/s PCI-E 4.0 SSD does generally do better than a 3.5 GB/s 3.0 SSD, but a 2 GB/s SSD with shorter seek time (Optane) will usually beat them both (but they're still all SSDs, so you have to be tracking the time very accurately for it to even be measurable. And naturally none of this applies to production workloads or synthetic benchmarks, which give roughly zero shits about access time.
@@credge I didn't think it was a bad idea, I simply have to come up with a way to show it in my videos because it's a great point ... perhaps I could expand my recommendation for a range of system prices if building new.
What a fantastic video! I pulled the trigger on the 9700X (not waiting for 9800X3D) and it smokes. I put in 2x16 KLEVV Bolt V 6000 RAM and it's running expo timings without a hitch. 30-36-36-76 Getting great benchmarks in 3DMark - well above average for my hardware combo. I might have some good silicon. This video has convinced me not to chase faster RAM - I'll be returning the KLEVV 6400 RAM that I never opened. Great price on both sets, but the performance diff will be negligible with my mid-tier GPU, not worth the extra heat/watts and potential instability. Thanks for all the practical and scientific advice - this was a lot of work to put together - very much appreciated!
5:33 WOAH! That's The Talos Principle II!!! I love that game so much. 7:21 I cannot believe you included The Talos Principle II in the benchmark. You have just earned my subscription to your channel.
This is an excellent video to begin with. A clear explanation that does not insult our intelligence or gloss over details. Seriously really well done. Instant sub. Also, I bought a Team Group 2x24 8000 CL 38 EXPO kit because I had a hunch and just wanted to try it out. When my 9800X3D arrives I'll find out if X3D negates the advantage. The best thing is, MS Flight Sim is the main reason I'm doing this upgrade, so it's good to see it benefits a little bit from my choice! One thing that's missing: are you just applying EXPO? I keep hearing that EXPO is "loose" and one should tighten one's timings. More information on how to do this would be useful.
looser means timings are set to high which is heavy on latency its better then to lower the timimg of the CAS and RAS... The actual first 4 timers on the ram which makes a big difference if changed lets say ram has CAS of 40 on DDR5....then bring it down to 38,,and luckil if you get 36 but you need good binning modules DDR4 is popular for speeds of 3000-4000mhz with usual CAS timing of 12-18 high speeds and low timing in any module is a winner
You can take a look at my previous reviews on the 9700X, 9900X, 9950X and the performance prediction for the 9800X3D to see step-by-step tweak guides. I enabled EXPO however I also tightened the timings and increased the tREF to 65535.
I'm just going to get 6000Mhz mem and run it 1:1 with the fclk running 2000Mhz. I think that's the sweet spot of performance, stability, price and no headache.
I have that motherboard and I was testing a 9800X3D on it until I discovered that the performance was lower than on my X670e … I’m in contact with gigabyte to try to understand what the issue is … I would avoid using this board with the current bios. And no, there is no guarantee that 8000 will run stable on any AM5 board, though I’ve had good luck with all of my Aorus boards.
It’s always a tough call, I really wanted to test 3 kits and I didn’t want the video to be too long … will test that in my 9800X3D review and tweak guide video.
OMG! ~8:15 Someone who finally gets it mostly right. That's rare. But for DDR5-6000 you can also go for FCLK 2166+. You will get a latency penalty for leaving the 3:2 ratio but the FCLK increase will lower latency more than the penatly is. But that's more an exception but a rule. Going with small ratios is definetly beneficial.
Yes, pushing FCLK as high as possible is always beneficial ... but you have to be prepared to do a CMOS reset when things go wrong, so I wanted to avoid that.
@@blackbirdpctechYes, but he was saying to go for the fastest UCLCK as that is the bottleneck. Since Zen4 the IF is double speed half width links, it's decoupled not synchronised. If you think about it, you'll realise it must be because DDRAM transfers in bursts of cachelines then has (long) delays, while IF has to transfer i/o constantly between the CCD & IOD and there's PCIE too, not just memory. Probably the cache protocol messaging goes over IF too.
@@blackbirdpctech Generally yes the CMOS Reset problem. But you already had FCLK of 2133 @ DDR5-6400. Since lower SOC Voltage makes FCLK more stable I guess 2166 FCLK @ DDR5-6000 should work quite easily. If not. 2133 definetly would work since it already worked @6400. And since @6400 you need much higher SOC it makes FCLK more unstable. In essence: From the data you provided I guess DDR5-6000 with FCLK2166 @ lower SOC Voltage is not a long stretch :) But just my 2 cents. Have a nice day.
@@TheTiltdJ if you are going for DDR5-6000 CL30 RAM, and you can afford the additional cost, then there is really no reason not to get a 64GB kit ... newer games like MSFS 2024 are supposed to want more than 32GB ... but I've never seen the memory demand in games get close to 32GB. Going dual rank does place a higher load on the memory controller, so if you are shooting for 6400 or 8000, then stick with 32GB or 48GB kits ... it will be tough to get 64GB stable at those speeds.
I got my Aorus X670 Elite AX mb with a G.Skill Trident Z5 NEO 6000MHz, CL36 kit when AM5 first came out. Paired with a 7600x. I read so many people on reddit having boot failures when trying to overclock their ram, so nervously I just let my RAM sit at stock settings at 4800. I've since updated my bios a few weeks back and going to give this another shot. Planning on upgrading my 7600x to a 9700x or 9800x3d in the next few weeks so I suppose I'll just wait til then.
I built my 1440p gaming system the week the 7800X3D launched. Had a quick look around at price/performance and a few videos here and there and decided 6000 CL30 was the way to go. Glad to see an in depth video that shows I chose well. 😀
Great cpu choice! I also got a 9700x and this thing is a beast. Never got all the bad day 1 reviews! Got mine fined tuned now and my cinabench 23 and 24 scores are better than the oc 9800x3d reviews ive seen. Plus i game a 4k so the cache wouldn't do alot for me.
I agree, it's a fantastic chip ... did you watch my 9700X vs 7800X3D review? I include a comprehensive tweak guide in that video ... might help you find something you missed (such as setting the tREF = 65535).
I bought the 9800X3D and paired it with the X870E Crosshair Hero, using a 32GB (2x16GB) 8000MT/s RAM kit with timings of 38-48-48-84. The setup is flawless, with no crashes and stable performance while gaming. The MCLK is set to 4000, FCLK to 2000, and UCLK to 2000. I also tested a 32GB (2x16GB) 6000MT/s RAM kit with timings of 28-36-36-96. While the 6000MT/s kit showed higher read, write, and copy speeds, it had higher latency compared to the 8000MT/s kit, as tested using AIDA64.
@@blackbirdpctech Great! I look forward to your review. I'm thinking of going 2100 on the FCLK but I will have to test to see if worth it. Thanks for replying 👍
@@jeroentjeuh3780 that's simply not true. Due to the extra L3 cache the dependence on RAM is less, but you will still see a benefit to running higher speed memory. Whether it's worth it is another question.
@@blackbirdpctech and only on extra low resolution/quality settings you'll notice it. So in real world, nobody is buying an X3D CPU + 4090 to game at 1080p medium or low quality settings. This makes the case for high speed memory kits even worse, even if you ignore the price of the memory.
@@dawn-moon that’s also not quite accurate. If you use features like ray tracing or path tracing, your cpu will be heavily loaded at 4k/ultra, so you will indeed notice a difference. You can try games like cyberpunk, F1, black myth: wukong, etc. to see for yourself.
As for increasing the memory frequency, it does not bring linear performance improvement. I discovered this when I watched tomshardware’s test of DDR4 memory in January 2017（i7-7700k）. The increase in gaming performance from memory frequency is minimal. Although there are some differences between the hardware conditions back then and now, the data you provided is generally in line with my predictions. Thank you brother
Each platform and CPU architecture react different to this. You cannot just say "the increase in gaming performance from memory frequency is minimal" just because it was like that in 2017 on Intel, even though so far yes, it would be true in most cases.
The thing is this video is comparing what is already considered the sweet spot to "What if you went higher than that?" Of course it's not going to show much, because they've already put in the optimum scenario for most users. If they used 4800 MHz DDR5 you're very well gonna see a difference in performance. The actual point of going past the sweet spot is reducing memory latencies to increase the lowest FPS possible, this helps reduce stutter in games when the difference between the highest FPS and lowest FPS is smaller.
Great video. I'd warn people though that even with the new 9000 series chips and 870 boards you're not guaranteed to hit 6400 with 2133 FCLK. I have a new chip that when run with these settings will throw errors within a few hours in Testmem5 Extreme1anta777. Given what we're seeing here, it's good to know that it's hardly worth the trouble going above 6000 as you said.
You could redo the tests with a 7900x, 7950x, 9900x or 9950x. They have more cores and more memory lanes, much more bandwidth at high clocks. The results would change but not for gaming (those would be the same). If you notice Aida's results there is no improvement in read/write/copy bandwidth at higher clocks due to the 7600x, 7700x, 9600x or 9700x limitations.
@blackbirdpctech one question I still have is if I can use an ddr5 kit that's xmp not expo if it's on the motherboards qvl? (Looking at the aorus x870 elite ice with a 9800x3d, ram kit CMH32GX5M2B6000C30W)
I have wondered how best to tackle RAM timings while making it interesting ... it's an incredibly boring topic with a lot of details ... I would likely have to break it up in to multiple videos ... let me know if you have any ideas because I'm worried that I spend a lot of time on it and 4 people watch it 😉
You can do 1 video explaining what the timing parameters are and their effects on systems. Then another on tuning on Ryzen systems another on Aerolake. I agree it's a daunting task however I'm sure you will make this enlighting and entertaining at the same time for us 😄
Thanks for the info, this is gold. By the way, the best game to test your PC stability is Starship Troopers Extermination on Horde mode. Reportedly, most PC gamers crashed in mid-game and the last game during the heavy scene.
I just bought a 9800x3d and I've Been searching the internet for days in order to find the best ram and this is hands down the best explanation, very clear and easy to digest.
@@JamesC1981 yes, the trends are the same however the magnitudes will be lower since X3D chips have extra L3 cache, so their dependence on RAM is reduced. That said, the recommendation from the video holds true for any 7000 or 9000 Ryzen chip.
Thanks for the effort that you put into these videos to help inform us (me being a gamer and just now getting around to my first ever pc build with little knowledge on the finer details of the components), you helped me decide between a 7800x3d and 9700x. I ultimately chose the 9700x and I am very happy with it thus far. I love your videos and hope more people like me will come across your channel. We need more tech tubers like you!
honestly any gamer would recommend the X3D chips best gaming cpus in the world and if you chose the 7800x3d you wouild have had the second fastest cpu in the world now as the 9800x3d {close in price} beats it with newer tech and faster cores and its a unlocked cpu besides for a normal cpu u need fast memory for x3d chips there is some forgiveness as memory data is stored on cpu cache which gives more fps in all games
@starofeden9277 i never asked for your input nor do I care what "every gamer" has to say, hence the reason i said we need more texh tubers like this guy. None of the other tech tubers with hundreds of thousands of subscribers have went in depth with the explanation of comparison between the 2 cpus like this guy has. If you wanna spend +$100 on the other cpu go ahead lol, I'll use it for games. I will also choose to listen to the rocket scientist instead of these other guys who rush out there reviews so that there subscriber base doesn't get upset with them. Also the 9700x bundle from microcenter was cheaper and the fps between the two was too close to justify purchasing the crazy expensive 7800x3d. Not to mention the 9700x will also be better for productivity from what I have seen.
First time on AMD platform and I've watched so many explainers on this fclk, uclk stuff thinking I made a mistake and should go back to intel. This one finally clicked it for me
Happy to hear that … I try to make videos that I would watch and I was genuinely always confused by this, so I decided to make a video on it to make it as easy as I could.
nice video, currently the only true need for fast tuned ram is Star Citizen, it benefits massively because of how broken the engine is and its massive cpu heaviness.
I tested 6000 CL28 and 6400 CL30 and the difference wasn't meaningful in the games I tested. That said, my top choice would be DDR5-6000 CL28 ... I didn't however look at the price and the impact on value.
The neat thing about a good kit of Hynix A-Die 6000 memory is that it can be tuned to run at 8000 or higher by manually setting the timings and voltage.
As long as the motherboard is capable of running 8000; both Hynix 16gbit A die (H16A) and Hynix 24gbit M die (H24M) are easily capable of running at 8000mts, also most of the performance improvement will come from tuning the secondary/tertiary timings especially tREFI.
Guess i should look into getting a Expo 6000M/T memory kit for my AM5 Ryzen 7700X chip in the future. Right now running 5600M/T RGB Ram from Teamgroup.
There's another reason why with Zen 4, AMD recommends 6000 kits. It's called the Infinity Fabric link between the CCD and IOD chiplets ALONG with having to connect through a multiplexer on the IOD along with conversions on each die from parallel to serial before data goes onto the Infinity Fabric. As far as I know about the BEST that can be clocked is 3GHz unless you won the silicon lottery. From testing I've seen in the past with mem kits, AFTER that AGESA came out that improved their mem controller, faster kits than 6000 MT/s didn't offer any improvement, which makes sense if you can't get the data from the mem controller on the IOD to the cores on the CCD faster than 3GHz, which is saying that the mem controller, with a clock speed of 3GHz to clock 6000 kits, in sync with the IF transfer speed of 3GHz is about as good as it gets for Zen 4, so even though they made the mem controller more stable with faster kits it doesn't do anything to improve performance since the limiting factor is the Infinity Fabric. Since I haven't really had an interest in either Zen 4 or 5 until the 9800X3D came out, it looks like the IF can now clock a little faster, around 3.2GHz so mem kits at 6400 is about as good as it gets and then mem clock and IF clock is the same speed. My expectation is faster kits WILL benefit Zen 6 which is what I'm waiting for. AMD is SUPPOSED to move to direct connects between die, getting rid of IF so the transfer speed between die will increase and therefore the faster kits will benefit Zen 6, but even with that I'm guessing 6800 - 7200 will be the sweet spot because you STILL have to transfer die from one die to another and that's slower than on-die transfers. I'd be really happy if the sweet spot is 7200 - 7800. AMD then might be able to increase core counts along with maybe moving to a mix of p and e cores, with AMD's e cores being their C cores which would make their CPUs even more efficient. Also hoping that CPU-chipset interface will be improved to 16GB/s for Zen 6 so that newer Xx70 MBs will have much better interfaces and top speeds for components off the chipset. Intel moved that link to 16GB/s back with 12th gen so in that way Intel is better for people that need the interfaces on a high quality MB. I'm guessing Zen 6 will give another 10% over Zen 5 AND the 8 core X3D part will be even smoother at frame time than the 9800X3D which is already excellent, as in a class by itself.
I discuss the infinity fabric in the video ... the IF clock speed, or frequency, is denoted by FCLK. It's tough to get anything above about 2133 to function on most systems, but to sync UCLK with FCLK for a DDR5-8000 kit in 2:1 mode you only need to run it at 2000. For a 6400 kit in 1:1 mode you should try to run an FCLK of 2133, which is challenging on most systems.
While I do not have an AMD system, I found an enourmous improvement in MSFS 2020 by going from the 4800mhz defualt ram to 6000 CL- 30 Ram but the difference in frame rate was not that much. The real difference was that the sim went from having bad stutters to being almost completely smooth, and the 1% low frame rate more than tripled. In games that are GPU limited, faster memory may not help, but in games that are CPU limited, it can make a big difference to the smoothness of the sim at 4k. Putting in faster memory made far more difference in the smoothness of the sim than all of the graphics tweaks I had been trying. For people running MSFS 2020 or 2024, I highly recommend running an overclocked memory.
We need quad channel memories in our PCs, especially in our laptops. This is now more efficient, caching files, NPU memory access speed to faster operations(fast means less delay, not utilizing NPU or GPUs longer)...
I’m going with the Gskill 2x32gb 6000mhz cl30 ram. I’d rather have 64gb for future proofing. And cl28 vs cl30 doesn’t seem like that big of a difference
So now I'm wondering if the Ry 7 9700 Blackbird Tech tuned can withtand the new power efficiency champion (according to GN's Tech Jesus) but the Ry 5 7600x3D is US's Microcenter Only, which can probably get away with even cheaper DDR5-DRAM
I would add that those timings are actually very loose timings vs us overclockers run on our tuned profiles.6400 cl 28 36 36 on A die is actually possible with some effort. Ram latency is pretty much in the same realm for all the kits in the video.
That's a good add ... latency didn't have a meaningful impact from my testing, so I decided not to include it. I only tested 6400 CL30 and 6000 CL28, so perhaps 6400 CL28 would have showed a larger impact.
Thanks for the video! It would be awesome to see a comparison between 2 and 4 sticks of DDR5 on AM5 (2x1R, 2x2R, 4x1R, 4x2R). Is there really a performance drop, and if so, how much? Is it even worth getting 4 sticks? Quick tip: maybe adjust your teleprompter (or whatever you’re using). It’s a bit noticeable that you’re reading - your eyes are moving, and it’s not quite focused on the camera.
I wouldn't recommend 4 dimms ... it's still really tough to get 4 dimms of DDR5 memory stable at EXPO or XMP on any platform. Still trying to figure out the best approach when it comes to producing the content ... I think it has been getting better but still some work to do ... hopefully I can use AI to make some small adjustments 😉
Another detail is that the fastest kits are also in smaller capacities. I was kind of limited by 32GB on AM4 so for AM5 I right away got 2×32GB 6000, least amount of fiddling, set EXPO, check values, okay, good to go.
Very informative video. I've always had Intel CPU, so tuning AMD Expo memory is completely new thing to me. But AMD released so good gaming CPU that maybe it's time to change from Intel to AMD soon. I have overclocked and delidded 14900KS, but I have a hunch that new AMD cpu could be more stable in everything, because it's so much cooler CPU and draws a lot less power. I subbed :)
Bro Neat explanations, point representaions, reasons for what and what not to do and why not to do. Nicely done. One recommendation if possible: The thumbnail of games during their fps showcase if possible can you make them slightly bigger. One question if 8000 is reallt this unstable why is AMD even recommending these speeds ? will there come a time when these speeds be beneficial ? both in gaming and productivity. Cause even now now the gain is so low.
Good suggestion ... probably time to change my chart template anyway ... will see if I can incorporate larger images. Wrt DDR5-8000, it's actually not that bad when it comes to stability but the CPU architecture is what limits the effectiveness. So my guess is that it will only change when AM6 is introduced.
I went beyond expo at 8000hz running at 36, 45,45,45. For my 9800x3d, I'm in love!! I also have it paired with a 4090, using a 480fps oled monitor at 1440p. It looks amazing, and I think I'm the bottle-neck now in terms of speed.
I am currently testing the 9800X3D and I ran DDR5-8000 at 36-48-48-98 ... it's an amazing CPU. One thing to be careful about is to check your memory bandwidth in a tool like AIDA64 ... just to make sure you didn't try to push it too far.
@@maxbr6865 1.43, but there's been a bios update though that improved my expo numbers slightly further than mine with less power. They did tweak some settings. I wouldn't have thought to use. It feels slower than mine, but more stable if that makes sense. I never had any crashes or anything on both timings. I might even keep the settings, and give it more juice to see how it feels.
If you still have the gear to do the job, can you revisit, with the 8700g ryzen 7, monolithic die perhaps more focus on the higher speed memory, and a seemingly - high from factory - memory controller.
@@blackbirdpctech id be super interested, the combintion of motherboard speed increase with ddr5 between the x670 series and the x870 series may also be a strong component for a ram monolithic die build. theres reasons to believe the 8700g can become faster then the 7700x, whilst also being cheaper then it, and coming with igpu. which makes it excellent value (perhaps not with its corresponding "speed up" parts though) Cheers
Big thanks for another great video! Now I will just stop worrying about ram and get whatever is the best value at the Black Friday sales lol. Looking forward to the 9800x3d video. I have a feeling that 9800x3d won’t add much value for gaming at 4K than 9700x.
I got a 9800X3D today and 8000MHz RAM on a X870E motherboard. I do play in 4K and guess what? You are right ;) But the 1% lows and the overall snappiness from the OS and day- to-day use coming from a 5800X3D? Wow that is night and day!
@@allenzhang373 I think the RTX 5090 will offer a significant increase in performance however I think it will retail for $1,999 US. They have no competition at the ultra high end, so my guess is that Nvidia will price it high.
@@blackbirdpctech Yes. I know from AM4 platform (5600G) memory has a good effect. I wonder if, compared to AM4, the higher bandwidth of AM5 and 2×32bit of the memory bus is a bigger benefit as often repeated. Just seeing the AM5 APU numbers in itself woudl be interesting. I expect a bigger boost than seen here, but not 20% or higher as some claim. I would test but no 8700G and no money to drop on a bunch of kits. I did go for 4400 on AM4 as an upgrade from 3000 and frames showed a good (not massive) increase. A comparison with AM4 would be a nice bonus.
One of my first videos was upgrading my media center PC from a 7700X to a 7800X3D ... my conclusion was that it wasn't worth it when playing on a 4K 120Hz TV.
In short, I got a 9800X3D with 8000MHz RAM so you didn't have to. And it didn't improve my games in any meaningful way, but it does work straight out of the box on my ASUS X870E Hero motherboard so I guess the benefit is that it is "just as good" as a tuned in 6400MHz kit with just using EXPO. Worth the cost? No today no :D
A 9700X should benefit more from fast memory, so it's good to hear that the 9800X3D improvement wasn't meaningful. I might try running an 8400 XMP kit to really push it 😉
The more interesting test is a tuned 14900ks with ddr5 8000+ at CL34 (tuned timings), 5000+ Ring bus vs. 9800x3d, 8000+ at Cl34 (tuned timings), 2133+ FCLK. I have a feeling they would be very close in the 1% lows, but the AMD averages would be higher.
I was planning to run some tests comparing my 14900KS against the 9800X3D ... the challenge for Raptor Lake is whether to constrain it to stay within the Intel Default limits. If I test outside of the limits then I am pretty sure I will instantly get 100 negative messages ... but it will be fun 😉
@@HotDogHunterX that sounds good however it would not be possible to run an FCLK of 2400MHz, so you would lose synchronization when running the system in 2:1 mode ... would be interesting to test and see if I could get something to work. I'm pretty sure any X870E can run CUDIMMs, not just MSI.
To be honest lower latency can outperform higher speed depending on cpu. This may change with CU dimm in the future with on memory controller that boost clocks. As latency gets better on higher clocks it will change and as cpu get faster chipset interfaces.
You said EFF CLK. What did CLK do to you? HA! I'm guessing AMD knows that there is performance to be gained in a better memory control on their next gen MB platform. Informative video! Good job!
you make many good videos that usually answer all my questions lol, thanks! for a 9900X, what kind of combo do you suggest for that processor in terms of memory, water cooler and motherboard?
There are many different options, so it really depends on your budget. I will do a post tomorrow in advance of Black Friday with some of my favorite components.
I still have a question I’m unsure about: is being able to run 1:1 at 6600MHz essentially just down to the silicon lottery? That’s the impression I’m getting. Originally, I planned to go with 6400MHz CL32, but then I came across a slightly faster 6600MHz CL32 option. I thought it might offer better FPS, but I understand it’s less stable. I want high memory capacity because I’m building my first PC for hybrid tasks-gaming and productivity (Photoshop, Blender, Unity, etc.). With that in mind, does memory density affect FPS in gaming? I was considering 2x48GB instead of 2x32GB to have more memory for multitasking, but would going with higher density negatively impact FPS instead of providing benefits? My reasons for opting for higher MHz and more memory are: first, I assumed 6600MHz CL32 would be faster than 6400MHz CL32, albeit less stable; second, I want high memory capacity for multitasking across gaming and productivity. So, overall, my questions are: Should I go for the RAM with higher MHz despite potential stability issues? Should I choose 2x32GB or 2x48GB sticks? Lastly, I’m really glad I found this video! I’ve been searching for a comparison between 6400MHz and 6600MHz and a breakdown of the trade-offs, but couldn’t find anything-then your video popped up and was super helpful. Thank you for making it! (also sorry for the wall of text lol)
I wouldn't recommend buying DDR5-6600 for AMD AM5 Ryzen CPUs ... it's not just less stable, you will very likely be unable to boot your system. For my 9700X, which is a golden sample, I was able to get it to boot in to Windows once ... the other 10 times I tried I couldn't get past Bios. So stick with 6400 or 6000. Higher capacity is fine as long as you use only two dimms ... you will not see any meaningful difference in performance.
Considering what we get for generational gains, (or losses, depending which company you go with 😂) a 2-4% uplift in the more meaningful 1% lows seems pretty worthwhile for $10-20. What percentage is that of the total cost of the system? I picked up a GSkill 6400, CL32 64gb kit and an Asrock X870E Nova motherboard for my new build. Now I'm just waiting for my 9800X3D to ship and I'll be in gaming bliss. Knock on wood, assuming everything just works.
@blackbirdpctech Lucky! I bought mine from Amazon, but they still haven't gotten back to me with a shipping date. When I go get it, I'm going to wait a bit for more info on weather or not AMD is really serious about going after warranties just for activating PBO and doing some common sense tweaks.
@@zackw4941 I got lucky and snatched one on Newegg about 20min before they were supposed to go on sale at 9am US EST. I also got one on Amazon but it looks like it will take a few weeks.
I was planning to include it in my benchmark suite for the 9800X3D however when I went to test it I couldn't get into the menu, let alone play/test the game ... will wait a week for them to iron out some of the issues and try again.
How do you think a 6000 cl28 kit would compare? There’s a trident royal neo kit floating around with those specs. It seems like 6400 clearly outperforms 6000 cl30, so I’d be curious to know if that would remain true against tighter timings.
The small decrease in CAS latency won't have a meaningful impact on the results however that's what I would buy, either the 6000 CL28 kit or 6400 CL30 kit ... I list both of them in my description as top picks.
I tried to copy your 9700x video, got the 6400 kit. I tried the CO at minus 30 and the ram at the cl30 timings but one of them made it unstable (post fine but windows wouldn't load). Ended up just sticking with EXPO and using CO at minus 10. I couldn't be bothered continued testing to work out which one it was. Any guesses which would be more likely? All the other tweaks worked fine with karhu, prime and cinebench. Thanks for another great video.
I would recommend sticking with EXPO and re-trying a -30 CO ... the boost in performance will be greater with the undervolt. One tweak I added for the 9900X and 9950X videos was changing your tREF to 65535 ... that's something you should do also ... will help to significantly reduce your system latency.
Great video, man! 👍 I have already watched a bunch of DDR5 for AM5 tests and there have been similar results. Now think what about memory timing settings? I mean those ones set in the XMP and EXPO profiles for the particular RAM set. There are many DDR5 RAM sets on the market which built with the same memory chips (like Hynix A-Die for example), have the same speed but have different XMP timing settings and different price. So is it worth to pay some extra cash for tighter XMP settings or could I set the same timing profile manually with the same speed for any memory set built with the same A-die chips ?
If it has the same memory chip then it will typically overclock the same ... the RAM cooler design helps a little, but more for when you really stress your system or are using it for longer periods of time, and you can add a fan to resolve this anyway. So yes, it's best to buy the cheapest DDR5-6000 CL30 kit that you can find and simply dial-in the tighter timings from a more expensive kit.
This was very informative. A question though, it seems that 2 x 16GB is being pushed as standard for gaming. Is the benefit of having 2x 32GB of any significance for gaming purposes?
For gaming it will only have an impact if you start to run out of memory capacity. It does place a higher load on the memory controller and it will obviously help in professional workloads like rendering, but for current games 32GB is typically sufficient. You can also buy 2x24GB kits for DDR5, which is a great middle ground.
That was one of your best videos so far imho. Superb information!! :D Can't wait your review on the 9800x3d sir
Thank you, I really appreciate the feedback!
@@blackbirdpctech Informative video; and while I feel the fps/$ metric on the cost of the product is important, the overall cost of the system is as well. If I were building a $2000+ system like what is shown in this video, and I were down to choosing my RAM, it's no longer a 95% price increase for ~4% performance, but more like a 5% increase for 4% performance.
It probably still isn't the most efficient use of your money of course. Generally, x3d does more for games VS RAM speed, and core count does more for production. This applies even more so to people who might want more than 32 GB of RAM (of course that is only for crazy people like me who like to have 50 chrome tabs open while they run a video encoder while playing a game and watching a RUclips stream).
Much of this comes from the fact that RAM has long since hit a point of diminishing return on raw throughput, and now latency is king (it takes about a 10% or more boost in throughput to match the performance gained from a 2% reduction to total memory latency) and higher speed memory kits don't generally improve much in the relative latency department (CL30 6000 is the same 10ns first word latency as CL40 8000).
And now I've spent way too much time rambling off topic to write a nearly 250 word essay/lecture that will be read by like 3 people.
@@credge nice comment and interesting take on fps/$ ... my only concern with using total system cost is that you can usually achieve the same performance with vastly different cost builds, so I don't want to mislead people in to thinking something is better or worse than what it is simply because of the components I chose. For example, if I selected an MSI X870e Godlike the total system cost would be high, versus an MSI X670e Tomahawk which would be much lower ... the performance of the system wouldn't change, but the price would.
I need to look at latency again because the last time I looked at it the gaming performance didn't really change.
@@blackbirdpctech I don't mean to imply that total system cost should trump RAM cost when buying RAM, because it shouldn't, and you can't be expected to include it on your video anyway, since every viewer has a different setup. I didn't really mean it as a criticism against your video, but more something that people should consider for themselves on a case-by-case basis. Because if I'm building an already top-tier system, then do I really want to lose 5% performance to save $100 on my $2,500 system? Probably not. But if it's an $800 system, then that $100 should go towards something else, like a better CPU or GPU. But the 6400 kit on a $1,200 system? Now it's 1% cost for ~3% gain (overall, but also only at 1080p).
Do I feel the conclusion that you should not buy the more expensive memory based on the fps/$ metric is a bit binary? Yeah, but I would also agree that the 8000 kit would not be worth it for the vast majority of setups; however, the 6400 kit on the other hand would have it's place in some builds (being only $12 more expensive instead of $107. But this assumes you're building a whole new computer, I would never recommend paying full price to replace an existing 6000 kit without a really good reason.
I do have to admit that my latency assertion could be behind the times as the time I discovered the latency thing was when looking at DDR4 benchmarks: where a 3800 kit lost to a 3200 kit, but a 3600 kit beat them both, but only beat the 3200 by a small amount; and then when I divided out the access time, I found that the 3800 kit had worse access time, while the 3600 and 3200 were very close on access time (tiny bit in favor of the 3200) but the 3600 got a small edge in overall performance because it had a close enough access time but significantly higher throughput, but the 3800 kit could not compensate for it's inferior access time with even more throughput. It was on that video that I did some calculations and came up with a rough estimate that, on average, it took a 10% bandwidth increase to match the performance of a 2% reduction in latency on DDR4 (it did vary from game to game, with some games actually preferring the 3800 kit). But again, this is DDR4 memory, not DDR5.
However, on your DDR5 kits: If you calculate for total latency in time (rather than clock cycles) on the four latency numbers on the 3 kits (I'll label them #1-#2-#3-#4). 8000 is better for #1, all 3 kits are functionally tied for #4; but for #2 and #3 the 6400 and 8000 kits are basically tied but both better than the 6000 kit.
And while the 8000 kit wins overall, the 6400 has closer performance to the 8000 than it does to the 6000 kit, even when you include the extreme statistical outlier of MS Flight Sim, this makes me think that total latency is doing more work than the actual throughput of the ram, though the throughput is still relevant.
And of course none of my observations take into account the MCLK:UCLK:FCLK relationship. Like how much do you lose when you go from 1:1 to 2:1 MCLK:UCLK; how much do you gain going from syncing up the UCLK and FCLK? It could be that I am completely wrong and that those are making more difference than anything else (though the benchmarks do still support the idea of the 6400 kit having a place in some mid-range or higher builds).
Games do the same thing with load times as well (sequential read speeds help, but random reads are more important), where a 7 GB/s PCI-E 4.0 SSD does generally do better than a 3.5 GB/s 3.0 SSD, but a 2 GB/s SSD with shorter seek time (Optane) will usually beat them both (but they're still all SSDs, so you have to be tracking the time very accurately for it to even be measurable.
And naturally none of this applies to production workloads or synthetic benchmarks, which give roughly zero shits about access time.
@@credge I didn't think it was a bad idea, I simply have to come up with a way to show it in my videos because it's a great point ... perhaps I could expand my recommendation for a range of system prices if building new.
First time builder here. This cleared a few things up for me . Thanks for sharing 👍
Glad I could help … good luck with your build!
What a fantastic video! I pulled the trigger on the 9700X (not waiting for 9800X3D) and it smokes. I put in 2x16 KLEVV Bolt V 6000 RAM and it's running expo timings without a hitch. 30-36-36-76 Getting great benchmarks in 3DMark - well above average for my hardware combo. I might have some good silicon.
This video has convinced me not to chase faster RAM - I'll be returning the KLEVV 6400 RAM that I never opened. Great price on both sets, but the performance diff will be negligible with my mid-tier GPU, not worth the extra heat/watts and potential instability.
Thanks for all the practical and scientific advice - this was a lot of work to put together - very much appreciated!
Great choice on buying the 9700X, it’s a fantastic CPU.
5:33 WOAH! That's The Talos Principle II!!! I love that game so much.
7:21 I cannot believe you included The Talos Principle II in the benchmark. You have just earned my subscription to your channel.
Great game … welcome to the Blackbird PC Tech community!
This is an excellent video to begin with. A clear explanation that does not insult our intelligence or gloss over details. Seriously really well done. Instant sub.
Also, I bought a Team Group 2x24 8000 CL 38 EXPO kit because I had a hunch and just wanted to try it out. When my 9800X3D arrives I'll find out if X3D negates the advantage. The best thing is, MS Flight Sim is the main reason I'm doing this upgrade, so it's good to see it benefits a little bit from my choice!
One thing that's missing: are you just applying EXPO? I keep hearing that EXPO is "loose" and one should tighten one's timings. More information on how to do this would be useful.
looser means timings are set to high which is heavy on latency
its better then to lower the timimg of the CAS and RAS...
The actual first 4 timers on the ram
which makes a big difference if changed
lets say ram has CAS of 40 on DDR5....then bring it down to 38,,and luckil if you get 36
but you need good binning modules
DDR4 is popular for speeds of 3000-4000mhz
with usual CAS timing of 12-18
high speeds and low timing in any module is a winner
You can take a look at my previous reviews on the 9700X, 9900X, 9950X and the performance prediction for the 9800X3D to see step-by-step tweak guides. I enabled EXPO however I also tightened the timings and increased the tREF to 65535.
A little bit? The uplift to those 1% lows was massive! You're going to be glad you did that.
@@dkindig I suspect the effect will be smaller on the 9800X3D because of the huge benefit of the 3D cache. But I am looking forward to trying it out!
@@dkindig anything above ~3% is meaningful, anything below is not ... I wouldn't call the uplift massive.
After days spending to get right DIMM this video was what I needed. Amazing content.
Happy to hear and I appreciate the feedback!
Damn this youtuber is a God send. Videos i havent seen before. Other youtubers probably copies your content.
Glad you liked it!
I'm just going to get 6000Mhz mem and run it 1:1 with the fclk running 2000Mhz. I think that's the sweet spot of performance, stability, price and no headache.
It’s definitely the no headache solution
Great great vid. Thank you, sir!
Glad you liked it!
wow best video on EXPO ram yet.. good job
Thank you, really appreciate the feedback and glad you liked it!
simple ,clear explanation and presentation,cheers brother
Thank you, really appreciate the feedback!
@@blackbirdpctech 9800x3d x870 aorus master does 8000 8200 q100% guaranteed
I have that motherboard and I was testing a 9800X3D on it until I discovered that the performance was lower than on my X670e … I’m in contact with gigabyte to try to understand what the issue is … I would avoid using this board with the current bios. And no, there is no guarantee that 8000 will run stable on any AM5 board, though I’ve had good luck with all of my Aorus boards.
Thanks for another good vid
It’s always a tough call, I really wanted to test 3 kits and I didn’t want the video to be too long … will test that in my 9800X3D review and tweak guide video.
@@blackbirdpctech Awesome, I'm looking forward to that
OMG! ~8:15 Someone who finally gets it mostly right. That's rare.
But for DDR5-6000 you can also go for FCLK 2166+. You will get a latency penalty for leaving the 3:2 ratio but the FCLK increase will lower latency more than the penatly is. But that's more an exception but a rule. Going with small ratios is definetly beneficial.
Yes, pushing FCLK as high as possible is always beneficial ... but you have to be prepared to do a CMOS reset when things go wrong, so I wanted to avoid that.
@@dasharius3021 This guy buildzoids
Absolutely, buildzoid is amazing … one of the most knowledgeable people on RUclips … I definitely respect him.
@@blackbirdpctechYes, but he was saying to go for the fastest UCLCK as that is the bottleneck. Since Zen4 the IF is double speed half width links, it's decoupled not synchronised. If you think about it, you'll realise it must be because DDRAM transfers in bursts of cachelines then has (long) delays, while IF has to transfer i/o constantly between the CCD & IOD and there's PCIE too, not just memory. Probably the cache protocol messaging goes over IF too.
@@blackbirdpctech Generally yes the CMOS Reset problem. But you already had FCLK of 2133 @ DDR5-6400. Since lower SOC Voltage makes FCLK more stable I guess 2166 FCLK @ DDR5-6000 should work quite easily. If not. 2133 definetly would work since it already worked @6400. And since @6400 you need much higher SOC it makes FCLK more unstable.
In essence: From the data you provided I guess DDR5-6000 with FCLK2166 @ lower SOC Voltage is not a long stretch :)
But just my 2 cents. Have a nice day.
My system is running with a 7700X and a XPG 32 Gb 6400mts Expo kit. It is running very stable.
That's great to hear ... I recommend checking to see if it is running in 1:1 mode.
Sure, I will check it.
Ive been searching all day for this info. And you explained it simple enough that even my novice brain could understand. Well done.
Glad it helped!
@@blackbirdpctech since you're here. Any reason to go with more than 32gb?
@@TheTiltdJ if you are going for DDR5-6000 CL30 RAM, and you can afford the additional cost, then there is really no reason not to get a 64GB kit ... newer games like MSFS 2024 are supposed to want more than 32GB ... but I've never seen the memory demand in games get close to 32GB. Going dual rank does place a higher load on the memory controller, so if you are shooting for 6400 or 8000, then stick with 32GB or 48GB kits ... it will be tough to get 64GB stable at those speeds.
Best video I’ve seen on PC building. Subscribed.
Welcome to the Blackbird PC Tech community!
I got my Aorus X670 Elite AX mb with a G.Skill Trident Z5 NEO 6000MHz, CL36 kit when AM5 first came out. Paired with a 7600x. I read so many people on reddit having boot failures when trying to overclock their ram, so nervously I just let my RAM sit at stock settings at 4800. I've since updated my bios a few weeks back and going to give this another shot. Planning on upgrading my 7600x to a 9700x or 9800x3d in the next few weeks so I suppose I'll just wait til then.
There were definitely some issues when AM5 was first introduced but they’ve been resolved over time with multiple bios/agesa updates.
@@blackbirdpctech Glad to hear it. Subscribed!
Welcome to Blackbird PC Tech!
I built my 1440p gaming system the week the 7800X3D launched.
Had a quick look around at price/performance and a few videos here and there and decided 6000 CL30 was the way to go.
Glad to see an in depth video that shows I chose well. 😀
I would say that buying a 7800X3D at launch and using a kit of DDR5-6000 CL30 with it was a very smart purchase.
Great cpu choice! I also got a 9700x and this thing is a beast. Never got all the bad day 1 reviews! Got mine fined tuned now and my cinabench 23 and 24 scores are better than the oc 9800x3d reviews ive seen. Plus i game a 4k so the cache wouldn't do alot for me.
I agree, it's a fantastic chip ... did you watch my 9700X vs 7800X3D review? I include a comprehensive tweak guide in that video ... might help you find something you missed (such as setting the tREF = 65535).
Ok with this you've earned a sub
Welcome to the Blackbird PC Tech community!
What an amazing video! Thank you so much!
You are very welcome!
I bought the 9800X3D and paired it with the X870E Crosshair Hero, using a 32GB (2x16GB) 8000MT/s RAM kit with timings of 38-48-48-84. The setup is flawless, with no crashes and stable performance while gaming. The MCLK is set to 4000, FCLK to 2000, and UCLK to 2000. I also tested a 32GB (2x16GB) 6000MT/s RAM kit with timings of 28-36-36-96. While the 6000MT/s kit showed higher read, write, and copy speeds, it had higher latency compared to the 8000MT/s kit, as tested using AIDA64.
That matches my experience too ... nice setup ... my 9800X3D review and optimization guide will be released late this week 😉
@@blackbirdpctech Great! I look forward to your review. I'm thinking of going 2100 on the FCLK but I will have to test to see if worth it. Thanks for replying 👍
@@boestube5805 I don't think you will find much value doing that ... you could however try pushing your RAM to 8400 and then it would make sense.
Amazing video! I have a 7800x3d with DDR5 6000/CL30 game at 1440p and it’s clearly the right combo based on performance to price.
Glad you liked it, nice combo!
3d cache doesn't care about ram even a 6000c38 kit runs the same.
@@jeroentjeuh3780 that's simply not true. Due to the extra L3 cache the dependence on RAM is less, but you will still see a benefit to running higher speed memory. Whether it's worth it is another question.
@@blackbirdpctech and only on extra low resolution/quality settings you'll notice it. So in real world, nobody is buying an X3D CPU + 4090 to game at 1080p medium or low quality settings. This makes the case for high speed memory kits even worse, even if you ignore the price of the memory.
@@dawn-moon that’s also not quite accurate. If you use features like ray tracing or path tracing, your cpu will be heavily loaded at 4k/ultra, so you will indeed notice a difference. You can try games like cyberpunk, F1, black myth: wukong, etc. to see for yourself.
As for increasing the memory frequency, it does not bring linear performance improvement. I discovered this when I watched tomshardware’s test of DDR4 memory in January 2017（i7-7700k）. The increase in gaming performance from memory frequency is minimal. Although there are some differences between the hardware conditions back then and now, the data you provided is generally in line with my predictions. Thank you brother
You are very welcome.
Each platform and CPU architecture react different to this. You cannot just say "the increase in gaming performance from memory frequency is minimal" just because it was like that in 2017 on Intel, even though so far yes, it would be true in most cases.
@yuan.pingchen3056 if your GPU has more headroom then faster ram will increase fps slightly
The thing is this video is comparing what is already considered the sweet spot to "What if you went higher than that?" Of course it's not going to show much, because they've already put in the optimum scenario for most users. If they used 4800 MHz DDR5 you're very well gonna see a difference in performance. The actual point of going past the sweet spot is reducing memory latencies to increase the lowest FPS possible, this helps reduce stutter in games when the difference between the highest FPS and lowest FPS is smaller.
Clear and well explained concepts. Subscribed.
Welcome to the Blackbird PC Tech community!
One of the best videos out there, covered more in 13 minutes in one video than I could find in many searches and watching other videos
That's great to hear, appreciate the feedback!
As a noob, great and simple explanations about ram settings. Thank you !
Happy to hear it helped!
indeed. Videos like this make me spare my money for my future configuration.
It’s always way more rewarding when it takes effort 😉
Thank you for the clear answer and solid testing.
You are very welcome!
That's just as simple as we need.
Keep up!
Glad you liked it!
Thank you. Moving up from a 5600X to a 9800X3D and have been having to get up to date on all this.
Glad it helped and enjoy your new cpu, will be a significant upgrade over a 5600X!
Great review! Perfect timing (pun intended). Thank you!
Thanks, love the pun!
Excellent informative video. Thank you
You are very welcome!
Really useful information. Thank you.
Glad you found it helpful.
Awesome video, thanks for the info!!!
Glad you liked it!
Really good video, mate. If someone doesn't know what they are doing OC'ing RAM on AM5, this should be their go to
Thanks, I appreciate that!
Hello , I don`t ussualy comment , but Great Job on the video , you got my Subscribe Sir
Thanks and welcome to the Blackbird PC Tech community!
Great video. I'd warn people though that even with the new 9000 series chips and 870 boards you're not guaranteed to hit 6400 with 2133 FCLK. I have a new chip that when run with these settings will throw errors within a few hours in Testmem5 Extreme1anta777. Given what we're seeing here, it's good to know that it's hardly worth the trouble going above 6000 as you said.
Exactly … it’s worth trying but I wouldn’t be upset if it doesn’t work
Very nicely done. :)
Glad you liked it!
You could redo the tests with a 7900x, 7950x, 9900x or 9950x. They have more cores and more memory lanes, much more bandwidth at high clocks. The results would change but not for gaming (those would be the same). If you notice Aida's results there is no improvement in read/write/copy bandwidth at higher clocks due to the 7600x, 7700x, 9600x or 9700x limitations.
That's interesting ... I might wait until the 9950X3D comes out to redo the tests ...
@@blackbirdpctech That would be awesome!
Damn this video was amazing. Subscribed.
Welcome to the Blackbird PC Tech community!
@blackbirdpctech one question I still have is if I can use an ddr5 kit that's xmp not expo if it's on the motherboards qvl? (Looking at the aorus x870 elite ice with a 9800x3d, ram kit CMH32GX5M2B6000C30W)
Great question and yes, shouldn’t be an issue even if it’s not on the QVL. The 8400 kit I used in the video was an XMP kit that ran fine.
same 100% had to
By far best video on Ryzen memory. I will buy 6000 CL30 for my 9800X3D.
Thanks, glad you liked it!
I have the same cpu too..is 2x16 better than 2x32?
@@JosephKarthic from what i’ve seen online, no. just make sure you get cl30 and you’ll be fine
@@itsmekrazyy Yes, I bought the gskil neo 6000 c3036 2x32gb kit...havent setup it yet...
@@JosephKarthic oh cool, hope you enjoy it
You absolutely nailed it 👌 Real knowledge here unlike others these days. 👏 Make a video on ram timing / tuning next 😊
I have wondered how best to tackle RAM timings while making it interesting ... it's an incredibly boring topic with a lot of details ... I would likely have to break it up in to multiple videos ... let me know if you have any ideas because I'm worried that I spend a lot of time on it and 4 people watch it 😉
You can do 1 video explaining what the timing parameters are and their effects on systems. Then another on tuning on Ryzen systems another on Aerolake. I agree it's a daunting task however I'm sure you will make this enlighting and entertaining at the same time for us 😄
Thanks for the info, this is gold. By the way, the best game to test your PC stability is Starship Troopers Extermination on Horde mode. Reportedly, most PC gamers crashed in mid-game and the last game during the heavy scene.
Thanks, will check that out.
Great video, succinct and relevant. You've got yourself a new subscriber!
Welcome to the Blackbird PC Tech community!
I just bought a 9800x3d and I've Been searching the internet for days in order to find the best ram and this is hands down the best explanation, very clear and easy to digest.
Glad it helped!
but does this test also apply to the x3d chips?
@@JamesC1981 yes, the trends are the same however the magnitudes will be lower since X3D chips have extra L3 cache, so their dependence on RAM is reduced. That said, the recommendation from the video holds true for any 7000 or 9000 Ryzen chip.
@@blackbirdpctech excellent good to know
@@JamesC1981 I will also include the impact of DDR5 speed scaling on the 9800X3D in my upcoming review
Great video and very easy to understand.
Glad you liked it!
Thanks for the effort that you put into these videos to help inform us (me being a gamer and just now getting around to my first ever pc build with little knowledge on the finer details of the components), you helped me decide between a 7800x3d and 9700x. I ultimately chose the 9700x and I am very happy with it thus far. I love your videos and hope more people like me will come across your channel. We need more tech tubers like you!
Awesome and congrats on your first build, the 9700X is a killer chip!
honestly any gamer would recommend the X3D chips
best gaming cpus in the world
and if you chose the 7800x3d you wouild have had the second fastest cpu in the world now
as the 9800x3d {close in price} beats it with newer tech and faster cores
and its a unlocked cpu
besides for a normal cpu u need fast memory
for x3d chips there is some forgiveness
as memory data is stored on cpu cache
which gives more fps in all games
You should take a look at my 9700X review and not listen to the flawed day one reviews.
@starofeden9277 i never asked for your input nor do I care what "every gamer" has to say, hence the reason i said we need more texh tubers like this guy. None of the other tech tubers with hundreds of thousands of subscribers have went in depth with the explanation of comparison between the 2 cpus like this guy has. If you wanna spend +$100 on the other cpu go ahead lol, I'll use it for games. I will also choose to listen to the rocket scientist instead of these other guys who rush out there reviews so that there subscriber base doesn't get upset with them. Also the 9700x bundle from microcenter was cheaper and the fps between the two was too close to justify purchasing the crazy expensive 7800x3d. Not to mention the 9700x will also be better for productivity from what I have seen.
Running a 7800X3D (UV) with G-Skill 48GB DDR5 6400 CL32. Hammer Expo in and runs very good. Very good video, thanks for sharing.
Nice setup!
First time on AMD platform and I've watched so many explainers on this fclk, uclk stuff thinking I made a mistake and should go back to intel. This one finally clicked it for me
Happy to hear that … I try to make videos that I would watch and I was genuinely always confused by this, so I decided to make a video on it to make it as easy as I could.
Wow! That's a quality vid! Thank you 🙏
You are very welcome, glad you liked it!
i dont know how much it'll matter, but if there could be a 9800x3d test version, would be a nice watch
nice video, currently the only true need for fast tuned ram is Star Citizen, it benefits massively because of how broken the engine is and its massive cpu heaviness.
Microsoft Flight Simulator scales well with RAM too ... another not well optimized game engine
would've loved to see 6000mhz with CL28 to see what kind of difference reducing latency makes compared to increasing hertz
I tested 6000 CL28 and 6400 CL30 and the difference wasn't meaningful in the games I tested. That said, my top choice would be DDR5-6000 CL28 ... I didn't however look at the price and the impact on value.
@blackbirdpctech thank you
I bought the Corsair Dominator Titanium 6600 CL32 2x32GB kit will see how it works with my 9800X3D coming tomorrow. I'm using Asus Rog Strix
X870-A MB
It will work however there is a very very low chance that it will run in 1:1 mode at 6600 ... I would recommend trying 6400 CL30 if 6600 doesn't work.
The neat thing about a good kit of Hynix A-Die 6000 memory is that it can be tuned to run at 8000 or higher by manually setting the timings and voltage.
That's an interesting point ... would be interesting to see if you can do that reliably across multiple DDR5-6000 kits.
@@HotDogHunterX But the Infinity Fabric clock timings go out of wack.
As long as the motherboard is capable of running 8000; both Hynix 16gbit A die (H16A) and Hynix 24gbit M die (H24M) are easily capable of running at 8000mts, also most of the performance improvement will come from tuning the secondary/tertiary timings especially tREFI.
@ivantsipr The CPU doesn't support those speeds you throw the Infinity Fabric clock out.
ABOVE 6400 MT/s your just wasting money for little to no improvement in real world performance.
Guess i should look into getting a Expo 6000M/T memory kit for my AM5 Ryzen 7700X chip in the future. Right now running 5600M/T RGB Ram from Teamgroup.
Probably won't be a big boost in performance but still a good idea if you can get it for a great price.
Great explanation on this 👏
Thanks, glad you liked it!
1:1 ratio is perfect!
Excellent video!
Glad you liked it!
This video was very informative. I will stick with 6000mhz cl30 ram then. Thanks for saving me a lot of money!!
You’re welcome … glad it helped.
Another amazing video 😎👌
Thanks, glad you liked it!
Nice video my man
Glad you liked it!
Still think you have some of the best content on youtube
I really appreciate that 👍
There's another reason why with Zen 4, AMD recommends 6000 kits. It's called the Infinity Fabric link between the CCD and IOD chiplets ALONG with having to connect through a multiplexer on the IOD along with conversions on each die from parallel to serial before data goes onto the Infinity Fabric. As far as I know about the BEST that can be clocked is 3GHz unless you won the silicon lottery.
From testing I've seen in the past with mem kits, AFTER that AGESA came out that improved their mem controller, faster kits than 6000 MT/s didn't offer any improvement, which makes sense if you can't get the data from the mem controller on the IOD to the cores on the CCD faster than 3GHz, which is saying that the mem controller, with a clock speed of 3GHz to clock 6000 kits, in sync with the IF transfer speed of 3GHz is about as good as it gets for Zen 4, so even though they made the mem controller more stable with faster kits it doesn't do anything to improve performance since the limiting factor is the Infinity Fabric.
Since I haven't really had an interest in either Zen 4 or 5 until the 9800X3D came out, it looks like the IF can now clock a little faster, around 3.2GHz so mem kits at 6400 is about as good as it gets and then mem clock and IF clock is the same speed.
My expectation is faster kits WILL benefit Zen 6 which is what I'm waiting for. AMD is SUPPOSED to move to direct connects between die, getting rid of IF so the transfer speed between die will increase and therefore the faster kits will benefit Zen 6, but even with that I'm guessing 6800 - 7200 will be the sweet spot because you STILL have to transfer die from one die to another and that's slower than on-die transfers. I'd be really happy if the sweet spot is 7200 - 7800. AMD then might be able to increase core counts along with maybe moving to a mix of p and e cores, with AMD's e cores being their C cores which would make their CPUs even more efficient.
Also hoping that CPU-chipset interface will be improved to 16GB/s for Zen 6 so that newer Xx70 MBs will have much better interfaces and top speeds for components off the chipset. Intel moved that link to 16GB/s back with 12th gen so in that way Intel is better for people that need the interfaces on a high quality MB.
I'm guessing Zen 6 will give another 10% over Zen 5 AND the 8 core X3D part will be even smoother at frame time than the 9800X3D which is already excellent, as in a class by itself.
I discuss the infinity fabric in the video ... the IF clock speed, or frequency, is denoted by FCLK. It's tough to get anything above about 2133 to function on most systems, but to sync UCLK with FCLK for a DDR5-8000 kit in 2:1 mode you only need to run it at 2000. For a 6400 kit in 1:1 mode you should try to run an FCLK of 2133, which is challenging on most systems.
While I do not have an AMD system, I found an enourmous improvement in MSFS 2020 by going from the 4800mhz defualt ram to 6000 CL- 30 Ram but the difference in frame rate was not that much. The real difference was that the sim went from having bad stutters to being almost completely smooth, and the 1% low frame rate more than tripled. In games that are GPU limited, faster memory may not help, but in games that are CPU limited, it can make a big difference to the smoothness of the sim at 4k. Putting in faster memory made far more difference in the smoothness of the sim than all of the graphics tweaks I had been trying. For people running MSFS 2020 or 2024, I highly recommend running an overclocked memory.
Great comment!
We need quad channel memories in our PCs, especially in our laptops. This is now more efficient, caching files, NPU memory access speed to faster operations(fast means less delay, not utilizing NPU or GPUs longer)...
Got the G Skill Ultra Low Latency cl28 6000MHz 2x16gb with the Ryzen 7 9800X3D and an Asus x870-E.
That is my top pick for RAM ... nice combo!
Good choice1 I went with the G.skill cl38 8000Mhz kit and I beleive there's no performance difference at all LOL!
@@laggmonstret lol
I’m going with the Gskill 2x32gb 6000mhz cl30 ram. I’d rather have 64gb for future proofing. And cl28 vs cl30 doesn’t seem like that big of a difference
So now I'm wondering if the Ry 7 9700 Blackbird Tech tuned can withtand the new power efficiency champion (according to GN's Tech Jesus) but the Ry 5 7600x3D is US's Microcenter Only, which can probably get away with even cheaper DDR5-DRAM
Yes I saw that video ... I have a Microcenter about an hour away, so I might head down there over the weekend to pick one up.
@blackbirdpctech Now that'll be very interesting, a cache vs cores & SMT battle.
Great video! gave u sub bro
Thanks and welcome to the Blackbird PC Tech community!
Awesome!! Thank you.
It always helps to look at the data 😉
Excellent, thank you!
Glad it was helpful!
yeah, I'll probably stick with the CL28 I got a good price and see how high it'll run. thanks!
That would be my top choice, DDR5-6000 CL28.
wow this was an excellent video thank you! This is definitely going to save me (and likely others) a lot of cash!
Thanks, glad you found it helpful!
I would add that those timings are actually very loose timings vs us overclockers run on our tuned profiles.6400 cl 28 36 36 on A die is actually possible with some effort. Ram latency is pretty much in the same realm for all the kits in the video.
That's a good add ... latency didn't have a meaningful impact from my testing, so I decided not to include it. I only tested 6400 CL30 and 6000 CL28, so perhaps 6400 CL28 would have showed a larger impact.
@@impuls60 and you can tune secondary timings and get even better performance
You can tune secondary timings on all of the RAM kits, so it negates the relative benefit.
Thanks for the video! It would be awesome to see a comparison between 2 and 4 sticks of DDR5 on AM5 (2x1R, 2x2R, 4x1R, 4x2R). Is there really a performance drop, and if so, how much? Is it even worth getting 4 sticks? Quick tip: maybe adjust your teleprompter (or whatever you’re using). It’s a bit noticeable that you’re reading - your eyes are moving, and it’s not quite focused on the camera.
I wouldn't recommend 4 dimms ... it's still really tough to get 4 dimms of DDR5 memory stable at EXPO or XMP on any platform. Still trying to figure out the best approach when it comes to producing the content ... I think it has been getting better but still some work to do ... hopefully I can use AI to make some small adjustments 😉
Another detail is that the fastest kits are also in smaller capacities.
I was kind of limited by 32GB on AM4 so for AM5 I right away got 2×32GB 6000, least amount of fiddling, set EXPO, check values, okay, good to go.
Yeah, the only good news is that you can get fast 2x24GB kits, which is a somewhat middle ground.
Very informative video. I've always had Intel CPU, so tuning AMD Expo memory is completely new thing to me. But AMD released so good gaming CPU that maybe it's time to change from Intel to AMD soon. I have overclocked and delidded 14900KS, but I have a hunch that new AMD cpu could be more stable in everything, because it's so much cooler CPU and draws a lot less power. I subbed :)
Thank you and welcome to the Blackbird PC Tech community!
Bro Neat explanations, point representaions, reasons for what and what not to do and why not to do. Nicely done.
One recommendation if possible: The thumbnail of games during their fps showcase if possible can you make them slightly bigger.
One question if 8000 is reallt this unstable why is AMD even recommending these speeds ? will there come a time when these speeds be beneficial ? both in gaming and productivity. Cause even now now the gain is so low.
Good suggestion ... probably time to change my chart template anyway ... will see if I can incorporate larger images. Wrt DDR5-8000, it's actually not that bad when it comes to stability but the CPU architecture is what limits the effectiveness. So my guess is that it will only change when AM6 is introduced.
nice video!
Thanks, glad you liked it!
I went beyond expo at 8000hz running at 36, 45,45,45. For my 9800x3d, I'm in love!! I also have it paired with a 4090, using a 480fps oled monitor at 1440p. It looks amazing, and I think I'm the bottle-neck now in terms of speed.
I am currently testing the 9800X3D and I ran DDR5-8000 at 36-48-48-98 ... it's an amazing CPU. One thing to be careful about is to check your memory bandwidth in a tool like AIDA64 ... just to make sure you didn't try to push it too far.
the bottleneck is the 4090
There will always be a bottleneck, simply saying that isn’t helpful because it’s the most powerful gaming card that you can purchase.
Hi, how much did you put the volt?
@@maxbr6865 1.43, but there's been a bios update though that improved my expo numbers slightly further than mine with less power. They did tweak some settings. I wouldn't have thought to use. It feels slower than mine, but more stable if that makes sense. I never had any crashes or anything on both timings. I might even keep the settings, and give it more juice to see how it feels.
If you still have the gear to do the job, can you revisit, with the 8700g ryzen 7, monolithic die perhaps more focus on the higher speed memory, and a seemingly - high from factory - memory controller.
I don’t currently have an 8700g however I can get one … will be interesting to test.
@@blackbirdpctech id be super interested, the combintion of motherboard speed increase with ddr5 between the x670 series and the x870 series may also be a strong component for a ram monolithic die build. theres reasons to believe the 8700g can become faster then the 7700x, whilst also being cheaper then it, and coming with igpu. which makes it excellent value (perhaps not with its corresponding "speed up" parts though)
Cheers
I didn’t realize that it was a monolithic design …
Big thanks for another great video! Now I will just stop worrying about ram and get whatever is the best value at the Black Friday sales lol. Looking forward to the 9800x3d video. I have a feeling that 9800x3d won’t add much value for gaming at 4K than 9700x.
Yes, I can confirm that now before testing ... that's one of the reasons why I keep pushing the 9700X as the value king.
I got a 9800X3D today and 8000MHz RAM on a X870E motherboard. I do play in 4K and guess what? You are right ;) But the 1% lows and the overall snappiness from the OS and day- to-day use coming from a 5800X3D? Wow that is night and day!
@@laggmonstret I have a comprehensive tweak guide coming for you soon for the 9800X3D ... it's an awesome chip ... I really like the 9000 series
@@blackbirdpctech do you think rtx 5090 will make much difference? my gut feeling is that it'll still be GPU limited for 3A games at 4k.
@@allenzhang373 I think the RTX 5090 will offer a significant increase in performance however I think it will retail for $1,999 US. They have no competition at the ultra high end, so my guess is that Nvidia will price it high.
This would be oh so interesting with an APU. I think you may see larger diferences there.
Such as an 8700G?
@@blackbirdpctech
Yes.
I know from AM4 platform (5600G) memory has a good effect. I wonder if, compared to AM4, the higher bandwidth of AM5 and 2×32bit of the memory bus is a bigger benefit as often repeated. Just seeing the AM5 APU numbers in itself woudl be interesting. I expect a bigger boost than seen here, but not 20% or higher as some claim.
I would test but no 8700G and no money to drop on a bunch of kits. I did go for 4400 on AM4 as an upgrade from 3000 and frames showed a good (not massive) increase. A comparison with AM4 would be a nice bonus.
I swapped back to the 7700X from 7800X3D. With a 4090 the X3D doesn’t really benefit performance in the games I’m playing
One of my first videos was upgrading my media center PC from a 7700X to a 7800X3D ... my conclusion was that it wasn't worth it when playing on a 4K 120Hz TV.
@@blackbirdpctech Why?
Because the increase in fps at 4k/ultra was relatively small … the gpu is carrying most of the load at these settings.
In short, I got a 9800X3D with 8000MHz RAM so you didn't have to. And it didn't improve my games in any meaningful way, but it does work straight out of the box on my ASUS X870E Hero motherboard so I guess the benefit is that it is "just as good" as a tuned in 6400MHz kit with just using EXPO. Worth the cost? No today no :D
A 9700X should benefit more from fast memory, so it's good to hear that the 9800X3D improvement wasn't meaningful. I might try running an 8400 XMP kit to really push it 😉
8000 RAM It's just better overall, but the higher timing kills it..
X3D doesn't benefit much from memory
Back in my AM4 days I went with DDR 4 4000 CL 14 AM5 I have a nice kit of 6000 MHz cl 30 2x32GB for when I get my hands on a 9950x3D.
@@laggmonstret what did you upgrade from??
The more interesting test is a tuned 14900ks with ddr5 8000+ at CL34 (tuned timings), 5000+ Ring bus vs. 9800x3d, 8000+ at Cl34 (tuned timings), 2133+ FCLK. I have a feeling they would be very close in the 1% lows, but the AMD averages would be higher.
I was planning to run some tests comparing my 14900KS against the 9800X3D ... the challenge for Raptor Lake is whether to constrain it to stay within the Intel Default limits. If I test outside of the limits then I am pretty sure I will instantly get 100 negative messages ... but it will be fun 😉
I’d love to see this comparison too as I currently have the exactly Intel system as you with cl34 @ 8000.
What would also be interesting is using MSI x870e with CKD 9600 dimms in Bypass mode, tune that....
@@HotDogHunterX that sounds good however it would not be possible to run an FCLK of 2400MHz, so you would lose synchronization when running the system in 2:1 mode ... would be interesting to test and see if I could get something to work. I'm pretty sure any X870E can run CUDIMMs, not just MSI.
To be honest lower latency can outperform higher speed depending on cpu. This may change with CU dimm in the future with on memory controller that boost clocks.
As latency gets better on higher clocks it will change and as cpu get faster chipset interfaces.
I’ve tested that multiple times now and speed has always come out on top.
good stuff ty
You are very welcome!
Clear and to the point 👍🫡
Glad you liked it!
You said EFF CLK. What did CLK do to you? HA! I'm guessing AMD knows that there is performance to be gained in a better memory control on their next gen MB platform. Informative video! Good job!
CLK is a major pain and so is FCLK 😉
you make many good videos that usually answer all my questions lol, thanks! for a 9900X, what kind of combo do you suggest for that processor in terms of memory, water cooler and motherboard?
There are many different options, so it really depends on your budget. I will do a post tomorrow in advance of Black Friday with some of my favorite components.
I still have a question I’m unsure about: is being able to run 1:1 at 6600MHz essentially just down to the silicon lottery? That’s the impression I’m getting. Originally, I planned to go with 6400MHz CL32, but then I came across a slightly faster 6600MHz CL32 option. I thought it might offer better FPS, but I understand it’s less stable.
I want high memory capacity because I’m building my first PC for hybrid tasks-gaming and productivity (Photoshop, Blender, Unity, etc.). With that in mind, does memory density affect FPS in gaming? I was considering 2x48GB instead of 2x32GB to have more memory for multitasking, but would going with higher density negatively impact FPS instead of providing benefits?
My reasons for opting for higher MHz and more memory are: first, I assumed 6600MHz CL32 would be faster than 6400MHz CL32, albeit less stable; second, I want high memory capacity for multitasking across gaming and productivity.
So, overall, my questions are:
Should I go for the RAM with higher MHz despite potential stability issues?
Should I choose 2x32GB or 2x48GB sticks?
Lastly, I’m really glad I found this video! I’ve been searching for a comparison between 6400MHz and 6600MHz and a breakdown of the trade-offs, but couldn’t find anything-then your video popped up and was super helpful. Thank you for making it! (also sorry for the wall of text lol)
I wouldn't recommend buying DDR5-6600 for AMD AM5 Ryzen CPUs ... it's not just less stable, you will very likely be unable to boot your system. For my 9700X, which is a golden sample, I was able to get it to boot in to Windows once ... the other 10 times I tried I couldn't get past Bios. So stick with 6400 or 6000.
Higher capacity is fine as long as you use only two dimms ... you will not see any meaningful difference in performance.
@@blackbirdpctech Thank you so much for answering my question! I’m now confident about which RAM stick to buy. Hope you have a great day!
Considering what we get for generational gains, (or losses, depending which company you go with 😂) a 2-4% uplift in the more meaningful 1% lows seems pretty worthwhile for $10-20. What percentage is that of the total cost of the system?
I picked up a GSkill 6400, CL32 64gb kit and an Asrock X870E Nova motherboard for my new build. Now I'm just waiting for my 9800X3D to ship and I'll be in gaming bliss. Knock on wood, assuming everything just works.
That's a good way to look at it ... I just got my 9800X3D today, so I plan to release a comprehensive optimization guide soon.
@blackbirdpctech Lucky! I bought mine from Amazon, but they still haven't gotten back to me with a shipping date.
When I go get it, I'm going to wait a bit for more info on weather or not AMD is really serious about going after warranties just for activating PBO and doing some common sense tweaks.
@@zackw4941 I got lucky and snatched one on Newegg about 20min before they were supposed to go on sale at 9am US EST. I also got one on Amazon but it looks like it will take a few weeks.
I'm running CL30 6600mhz ryzen 9 9950x
Are you sure that it's running in 1:1 mode? If it is then you got very lucky, congrats!
Could you test msfs 2024 with memory scaling on 9800x3s the outlier on msfs is pretty wild
I was planning to include it in my benchmark suite for the 9800X3D however when I went to test it I couldn't get into the menu, let alone play/test the game ... will wait a week for them to iron out some of the issues and try again.
How do you think a 6000 cl28 kit would compare? There’s a trident royal neo kit floating around with those specs. It seems like 6400 clearly outperforms 6000 cl30, so I’d be curious to know if that would remain true against tighter timings.
The small decrease in CAS latency won't have a meaningful impact on the results however that's what I would buy, either the 6000 CL28 kit or 6400 CL30 kit ... I list both of them in my description as top picks.
great video
Glad you liked it!
I tried to copy your 9700x video, got the 6400 kit. I tried the CO at minus 30 and the ram at the cl30 timings but one of them made it unstable (post fine but windows wouldn't load). Ended up just sticking with EXPO and using CO at minus 10. I couldn't be bothered continued testing to work out which one it was. Any guesses which would be more likely? All the other tweaks worked fine with karhu, prime and cinebench. Thanks for another great video.
I would recommend sticking with EXPO and re-trying a -30 CO ... the boost in performance will be greater with the undervolt. One tweak I added for the 9900X and 9950X videos was changing your tREF to 65535 ... that's something you should do also ... will help to significantly reduce your system latency.
Great video, man! 👍 I have already watched a bunch of DDR5 for AM5 tests and there have been similar results. Now think what about memory timing settings? I mean those ones set in the XMP and EXPO profiles for the particular RAM set. There are many DDR5 RAM sets on the market which built with the same memory chips (like Hynix A-Die for example), have the same speed but have different XMP timing settings and different price. So is it worth to pay some extra cash for tighter XMP settings or could I set the same timing profile manually with the same speed for any memory set built with the same A-die chips ?
If it has the same memory chip then it will typically overclock the same ... the RAM cooler design helps a little, but more for when you really stress your system or are using it for longer periods of time, and you can add a fan to resolve this anyway. So yes, it's best to buy the cheapest DDR5-6000 CL30 kit that you can find and simply dial-in the tighter timings from a more expensive kit.
This was very informative. A question though, it seems that 2 x 16GB is being pushed as standard for gaming. Is the benefit of having 2x 32GB of any significance for gaming purposes?
For gaming it will only have an impact if you start to run out of memory capacity. It does place a higher load on the memory controller and it will obviously help in professional workloads like rendering, but for current games 32GB is typically sufficient. You can also buy 2x24GB kits for DDR5, which is a great middle ground.
@@blackbirdpctech Thank you for the speedy and efficient answer, friend.