DDR4 timings explained: tRRD & tFAW // THE MOST IMPORTANT MEMORY TIMINGS

[@flamoclanlead]

This is invaluable info for us normal people trying to understand ram. Thank you for making these.

[@Darkness-ud5wk]

I understand ram. It makes your computer faster!

[@alaskanassassin7332]

Very good buildzoid. I appreciate the breakdown of jedec to normie speak for people like me. Much luv

[@FcoEnriquePerez]

Trust me, not so normie, you wouldn't even understand that if you were a "normie" lol

[@0vers33r1]

I agree 👍

[@Gekko12482]

@@FcoEnriquePerez on this channel we have a different bar for "normie" lol

[@bact3546]

@@Gekko12482 lmao yeah this channel is legendary

[@Mozburg]

These are the videos I listen to in the background while gaming, then watch with a notepad taking notes, then watch again while actually tuning my ram, thanks a ton dude

[@ghoulbuster1]

Don't give up yet! This is important information that you can't find anywhere else!

[@OCDyno]

Glad to see tRRD/tFAW get the attention they deserve

[@branchprediction9923]

What happens if tFAW is shorter than what tRRD_sg and tRRD_dg take to complete 4 times. Like both tRRD are 8 but tFAW is 26 for example? Does tFAW just get corrected to 32? And what if tRRD_dg is 5 tFAW is 20 but tRRD_sg is 8?

[@exoticspeedefy7916]

@@branchprediction9923 How to I set tRRD-sg and tRRD-dg? I don't see them in BIOS so what are these timings referring to?

[@branchprediction9923]

@@exoticspeedefy7916 they could be named DRAM RAS TO RAS DELAY S and L

[@exoticspeedefy7916]

@@branchprediction9923 I don't see that anywhere.. On a Gigabyte Aorus Master x570

[@branchprediction9923]

@@exoticspeedefy7916 it could also be named tRRDS and tRRDL

[@quitehandsomedude6412]

Talks about not enjoying making video about memory timings. Still makes tons of super long videos about memory timings. Absolute MADLAD.

[@KiriakosVids]

Regardless if the series continues or not, i really appriciate you making all theese episodes. I literally had some pen and paper near me keeping notes.

[@mroutcast8515]

Too bad views dropped. For me this is best video series on your channel. Could please do one more video on all those non-timing settings related to RAM? BGS, GDM, ProcODT, RttPark, all sorts of DrvStr, etc.. because NOBODY ever speaks about these. Just short video on what are those and how do those play into RAM functioning.

[@ChumlyFernando]

This would be amazing! I have no idea what these do and I've been struggling to find videos with this level of translation to normie 😂

[@DevinDTV]

please this

[@flamespirit1]

I upvote this comment

[@Kazyek]

I want this too! I have no idea how ProcODT / RttPark / DrvStr works and I think it might be affecting my impossibility to have a working non-GDM T1 setup

[@blubaustin1]

@@Kazyek there is a guide online which explains these pretty well. The idea with procodt though, is you change it when X speed doesn't boot. The drive strength is there for slight stability tuning.

[@guiorgy]

The quality of the presentation and explanation did improve. Thank you for the effort!

[@ActuallyHardcoreOverclocking]

first

[@Thomas_Aslaksen]

🤣

[@Phizev]

I know how to read timing diagrams, I've spent chunks of time trying to understand memory timings for devices from WiFi routers to desktops. Your memory timing videos are still where I've learnt the most.

[@mohammad-zabih]

[@Itallcostsmoney]

I am so thankful and happy that you are continuing this series! Thank you! Truly awesome content.

[@Itallcostsmoney]

@buildzoid Are you going to be doing any live streams in the near future?

[@ApexLodestar]

I know the views haven't been great with these very educational videos but I greatly appreciate it considering I'm still new and constantly learning. Thank you, Buildzoid.

[@belsteed6050]

Wow great video buildzoid !!!!!! Quality content and much effort into it, congrats

[@EnvAdam]

some of this goes way over my head but this is the type of stuff Ive had trouble learning before, very useful video.

[@KG_BM]

These videos actually helped me to understand mem ocs better. Theres a reddit post explaining each timing but its different to see your spreadsheet for it. You might not consider these videos successful, but the people who want to seek such info will find this resource and be glad it exists.

[@pilerks1]

Thank you so much for these videos, they made me aware of documentation and timings that I never bothered to look at. Please keep making as many videos that are related to this topic, it gets so overlooked and no one really goes into the depth that I know of for free.

[@TheXlen]

Definitely want to see more timings covered, I hope you get to continue the series, either case been pretty educational

[@EdKiefer1]

This is great vid, this info is not explained anywhere else good, keep up great work.

[@halrichard1969]

I can imagine your efforts here will be rewarded as time goes on. Thank You Mr Zoid.🙂

[@luke.m]

Was waiting for you to get past the primary timing explanations. Specifically in context to how they all interact! These videos should be evergreen content, keep going with the series!

[@mdtauk]

Would be nice to see you filling out timings when overclocking in the bios, and talking about where the lower limits are, and how you can reach a sweet spot. Plus Ryzen considerations with the fabric limitations

[@soaphelps]

Thank you BZ these videos have been a super help for me.

[@flint723]

Really enjoyed this one. I mean I knew tRRD and tFAW were crucial for performance (from benching) but now I know why.

[@cacman440]

I've been scared of touching tertiary timings since I have no idea what I'm doing, but I lowered the different bank group timings, tfaw, and rounded down to the next even number for the other tertiary timings and my 5900x is SO much faster. Thanks!~

[@TAP7a]

How have you found it to be better? Benchmark scores, game frames, subjective responsiveness? I’m really interested to see the scale of the impact on the famously memory-sensitive 5000 series

[@cacman440]

@@TAP7a Opening file explorer and navigating through windows, switching browser tabs is near instant. FPS gains in games are within margin of error, so I'm not sure it helps there. Although focus on getting the highest 1:1 ratio you can with infinity fabric/ram first, since that's more important.

[@ghoulbuster1]

@@cacman440 fps doesn't change but you will get less stuttering and games will load faster in general.

[@toonnut1]

Thanks for your efforts on this subject. I appreciate your videos. Thanks very much!

[@awnordma]

These are great, I'm sure long term view counts will be good since you're pretty much the only one doing content on this.

[@rulekop]

As a fledgling system architect this is a legit awsome source for me to understand DRAM performance considerations. Well done!

[@VenerabIe]

I am leaving 10 minutes into the video, because I'm at a stage in learning where this goes over my head. But you've more than earned my subscription in that time. I plan to follow this channel and learn as much as I can. You do an awesome job elaborating and going through things methodically, and you connect ideas so the whole picture makes sense. Once I do some foundational learning, I'm looking forward to returning here (and checking out your other videos) to learn more. Cheers, mate. Really glad you're making these kinds of videos.

[@ErraticPT]

Appreciate all the timing videos, thanks.

[@CompatibilityMadness]

Superb work ! Now, we only need part 5 for write timings... (tWR, tWTR, tCWL)

[@mylittlepwny3447]

I appreciate the effort you put into this series.

[@corneliusjung8617]

Lets keep this series going!

[@shaneeslick]

G'day Buildzoid, Your intro for this one hit the nail on the head exactly why those of us who have watched all the videos & why we find so important "You are Converting JDEC to Normie Speak for us", I think many of have said we are finding it "Dense" because we have so many "😲WOW! so that is how it works" or "🤔When you explain it like that, Yeah Now I get it" moments, especially as it helps me understand things I don't know in your RAM OC videos. My thoughts on your preparation for each of these videos would be like a School Teacher/College Professor having to create the curriculum for their Class, They know & Understand the subject extremely well, are just Breking Down the Information & converting that to Normie Speck for those taking the Classes, "It's not that hard so why don't other people just get it?" like those Teachers/Professors you can read & easily understand the Spec Sheets but for us it takes your help & we greatlyy Appreciate that, I have been Learning about & building PCs since the late '80s, yet this is the first channel where I love sitting & listening to learn what is normally very boring because it is over my head, I find your videos engaging because you do make the topics understandable & I like your sense of humour so also get to have a laugh, I understand these are not your favourite videos to prepare for & make but the Complete series would be great, because I think I am not alone in saying even if a video for the rest of the Timings is made once in 4-8 weeks with a Playlist on the Homepage so we can go back when we need to would be Awesome.

[@strat0caster124]

Imma go home and check my tFAW and tRRD first thing after work today. Huge enlightenment! I love being in this EE world

[@KZ-ko4vm]

Big thumbs up for explaining in a very clear way how it works and what is important.

[@Torbjorn.Lindgren]

AIDA's persistent failure to address this is especially egregious given how easy for them to add another metric which would capture this. The same applies to other benchmarking applications that have the same issue. The easiest metric to add would probably be "Random Memory Bandwidth". It doesn't actually need to be "proper" (secure) random, there's many simple patterns which would be good enough to get a stable and usable number, even if it perhaps wasn't the actual worst case - stable and repeatable is far more important (say fixed seed LFSR)! AIDA tends to like having four numbers so they could show the 100% read, 100% write and then the last two would be random read/write selection from random location, and the average access latency from that test. Many other pieces of hardware have similar bandwidth issues, try find a disk or SSD benchmark that doesn't try both sequential and random bandwidth for example, there's plenty of precedent for doing this in many other areas because as you note it is very important.

[@Elessar_Telcontar]

Thank you for making this series! I have learned quite a lot! I appreciate the time and effort you have put into this series.

[@Mefedroniarz]

I’m very thankful that even those videos dont make huge numbers - you still do them ❤️

[@Trisks]

this is actually fun to watch

[@casper18795]

Really liked this video. The density and pacing is perfect.

[@skachor]

I found this thanks to a comment you made in another of your videos, and here you mentioned jdec docs. Nobody else has, in so little time, given me a path to follow as well as explained in detail the concepts on show. Some people say info dense, I say focused and on topic. Thanks so much for all your work over the years

[@simoSLJ89]

These are the type of videos that people watch when they need, so I think/hope that they will do better in future. Since they can be useful for DDR5 too, maybe changing the titles could help? Otherwise most people won't watch them, since DDR4 is almost EOL.

[@dimwillow7113]

yep gonna watch this when i oc ddr5 with zen 4.. will be nice

[@KG_BM]

I wish i could upvote this 100times

[@n0rie9a]

nah, i watch this as ASMR

[@Shirotoko]

It will all apply to ddr5?

[@HexerPsy]

This is worth gold for someone like me. THANK YOU for making this video! And I secretly hope for more :P The scary part about messing with your DRAM timings, is that you dont know what you are doing - so when it wont boot.... there is a lot of worry. You also dont know what to do if the system seems stable, but memory errors pop up on stress tests. Which of the many timings should you even start with... Previously i used XMP and thats it; some years back I tried DRAM calculator for Ryzen to get more out of it. This video series made me tinker again :D I know you dont like that tool, because the timings are too loose. DRAM calc suggested tRRDS 6, tRRDL 6 and tFAW 24 going from 3600 to 3733; tweaked it down to 5 5 and 22 (sadly it wont do 20 or lower tFAW, neither will it do 4 4). There probably is a bit more tweaking to be done. Thanks again!

[@Gindi4711]

My progress so far: Part1: Boring, I already know this Part2: Already know this. Well maybe I have not thought about it from that angle Part3: Actually useful Part4: Wow that explains a lot of things... Would be a shame if there is no part5

[@ELHV]

It took me a few sittings and a long list of hand written notes to get all of this into my tiny brain. Now I finally have a reasonable understanding of how memory accesses and timings work and why my "fast" G.Skill kit with dreadful tFAW limitations is getting absolutely demolished by a "slow" Crucial kit that can handle tFAW=16. Thank you for putting up with us and creating the series!

[@hudsonbrown31]

You are doing gods work, thank you

[@HoshPak]

I'm so glad I found this video series. When I recently bought my new RAM kit and XMP wouldn't work out of the box I was so annoyed I immediately wanted to return it. Now, I've followed pretty much every advice on here and this supposedly bad kit turned out to overclock like a darling. Thank you for this treasure of a video!

[@sunlbx]

same! 3000 kit hardly worked at 3333, now 3640 super stable!

[@jannegrey]

Please continue. I've had part 3 on mute for at least 72 hours, looping. It really helps.

[@PooPooPerson]

Helps what exactly...?

[@jannegrey]

@@PooPooPerson The number of times it is viewed? I seriously want to see a video on Secondary/Tertiary timings.

[@simoSLJ89]

@@jannegrey Just FYI, YT stops counting after some views. At the beginning, the algorithm considers more views of the same account, to value useful videos, but then the algorithm stops counting, to avoid spam views. So I think that the 3 days loop was at least 70% unuseful.

[@jannegrey]

@@simoSLJ89 Yeah. It is certainly useful for 2 hours - tested i myself. But beyond that, IDK.

[@LA-MJ]

Engagement comment is engaging

[@sandornyemcsok4168]

I would like to share my contsructive critics. I am very good at Excel and I help my colleagues from time to time by building solutions for them. It is not my duty, I do it as fun because I like it. When taking on such project, I usually have to listen to somebody else explaining the task, the procedure which I have never done, sometimes do not even have any preliminary knowledge about it. I have to be able to understand it, set up the flow / algorithm, additionally identify underlying rules, principles that may not have been stated explicitly. Thus I even ask questions to clarify those hidden rules. I also have a solid PC hardware knowledge (I build/continously upgrade my own PC) although I am not an overclocker at all. So after the 2nd video I decided to build an Excel template to model the timings automatically. During this I have encountered quickly the fundamental educational / pedagogical problem with the series: you jumped into concrete examples (regardless how simple they were) to early and left a big gap between the fundations you laid down and the examples presented. You should have spent more time with laying down the fundations, setting up some declarations (for example 1 read command always triggers 4 columns to be read, the appearance of the data on the data bus does not impact the ability of the MC to send commands, etc.), explaining the timings (both primary and secondary) in an isolated manner (only to show what they mean, without going into anything else, like what other timings do they depend on, how do they influence the overall performance when they are combined in the real world scenario) - and personally I would have used a simple diagram because visual presentation usually helps a lot -, establish logical relationships between commands/events in a generic way (again in isolation, only showing small cut-outs to demonstrate them). It takes additional time, but it is an investment for the future. Because when you set up your lecture this way it makes the later parts much easier to follow and understand. When you explain the more and more complex real life scenarios, the building bricks are readily available, the listeners can/will remember of them ("Oh yeah, I can remember of this piece, that piece from the initial session! So this is how they come together into the full picture. Cool, clear, I understand it."). Unfortunately you have not done this that's why the 2nd and 3rd parts are so painful, they left too much in dark. It was only this 4th part that finally started to tell some information that would have been necessary for the previous parts as well. I hope so you can take this feedback positively and learn from it. Thanks for the videos anyway, the 4 part sequence is certainly informative.

[@TheFlyMan3829]

Hey buildzoid. I know the views for these vids arent high but theres something you may not be considering for the long term view count and that's searchability. Right now KZbin is playing favorites and all the top dogs and sucking all the views from smaller creators. The reason for this is because top names are more searchable in the short term, however in the long term, content with high watchtime/searchability like informative content such as this performs better over longer periods of time (say like 6-9 months). So in a sense these are "sleeper videos." Long term the high watchtime of these videos will end up pushing these videos and other's you have done out to more people. SO the instant gratification may be low but this content is super valuable to both us and (most likely) you as well. Hopefully you can take some solace in that despite how unfun these videos are to make. Keep it up man. Love your work as always.

[@B3n85_]

Dear Mr. Zoid. Thank you for making these videos, I know you hated making them but they defiantly helped someone like me to understand what I should actually be trying to accomplish tinkering with timings. I'm very new to this and it helped a lot. I actually increased my Cinebench score 545 points and gained 25 fps in WZ2 (although it crashed). This is the type of information I would pay for if it was what you had on your discord for ex. I had several questions during the last few videos.

[@fracturedlife1393]

absolute legend

[@m4nu507]

I would love a video regarding the rest of the timings, like is there a relationship between the RDRD_sg dg and the WR sg and dgs, do they need to match with the RD sg and dg? Also the RD and WR_dr timings which are very important for dual rank memory. You could probably do one last video to quickly go over these timings and some obscure settings that you rarely find information or very confusing at least like tCKE, tXP, OREF_RI, PPD, tREFI9x, DllBwen, etc.

[@lunabba3424]

Thank you for doing this. These may have not been fun to make, but the passion shows.

[@MJM750]

Thank you for this video. I am sorry you don't enjoy making these; however, I really, really, really appreciate you making these videos.

[@cubeislife1675]

lmao i didn't know anything about motherboard trace topology and i decided to quickly look into it and the first thing that pops up on youtube for me are your videos

[@zydra_ow]

Please continue this series!!

[@jazzistation4802]

YT algorithm just didn't let me know you posted videos and now i have to go back and retrieve part 2 & 3. just so you know

[@Aviation_16]

This series is awesome. And it does make sense. And in practical speak. Just… good. Thank you for doing this whole series.

[@theoc1009]

This video is absolutely insane tFAW 40 from XMP to 16, i gained 6,5% multi core performance on CPU!! WITH 1 FUC*ING SECONDARY TIMING WTF

[@gtijason7853]

Don't worry about the views, probably just YT age restricting and/or being petty for no apparent reason

[@LCTRgames]

These sorts of videos might not have the impact of others, but the long-tail on views for people searching for this information over the years to come must be a lot better. Kinda public service broadcasting for nerds :)

[@spudz7405]

I will re-watch it when I go to do memory overclocking

[@sondre113]

Thank you SO much for these lessons, we are blessed!

[@Hardwayistheonlyway]

What I really want is a gui in bios for memory timing adjustments so you can visualize the timings against each other. Kind of like automation on music production software. Would make tuning memory a dream

[@ActuallyHardcoreOverclocking]

most timings do not affect each other since they all describe relatively specific operations in the memory.

[@Hardwayistheonlyway]

​@@ActuallyHardcoreOverclocking At the very least some values do have to fit within other values. The size of those values could mean you don't finish an opperation before the 'window' closes so yeah a visual representation of that would be super benificial imo. If not for the seasoned overclockers at the very least for noobs trying to grasp timings (probably would also mean you didn't have to make as many memory timing guides too if people could actually visualise it) *Clearly If a value has no dependancies on other values then yeah a visualisation would be pointless. Does that invalidate the need for it where it is relevant though? Imo clearly not otherwise timing diagrams wouldn't exist.

[@PillowOfEvil]

Thank you for this video. My insomnia has been cured. I juat gotta watch this for like 15 minutes or so. Not saying its boring or monotone or anything. Just sounds like a nice lullaby.

[@jeremymoon9088]

Easy, Yung BZ, we're watching it. If u make it, we're watching it!

[@SinX897]

Thanks buildzoid for this informative video. Helped me overclocking my hynix cjr kit.

[@superangrysnowman5869]

I personally appreciate these videos. Thank you for doing them. If you are focused on views ,.. do a video listing the complete ram timings for both single and dual channel B die,. you won't need much explanation. Do it geared towards gamers , high fps and label it in the title. Put Warzone in the tile and every Warzone player who wants extra FPS that they can't get because the game engine is so bad will watch. I personally would like to know more about how ram timings effect FPS in Games. I do Audio and Video production as well and would like to know how Ram timings would effect renders or audio VST plugin usage etc.

[@usrname1y]

Best vide I've seen in a long while and I watch stuff like this all the time. May you live a long and blessed life our lord and RAM timing savior 😁

[@Lokotraktor]

Great video. Easy to understand, important topic and the colours made a world of difference. Thanks!

[@ORy652]

I cannot thank you enough for these explaining videos, learned a TON!

[@cainzjussYT]

Thank you again. Even tho this may seem basic stuff to you... to us normies your explanations are a gold mine of knowledge. i have tFAW on my XMP at 44 I will try to change it and pray it works eventho i have a GB motherboard. The bios on it is a hit and miss with every version.. in terms of things applying and auto reseting.

[@andersjjensen]

Thanks for doing what you dont like. Now I know where to focus my efforts at least :)

[@Pongomuffin]

This is excellent! Thank you!

[@extremegf]

Awesome explanation! Best on YT.

[@PainIs4ThaWeak85]

Thank you again, Buildzoid!

[@baboon81choc]

Thank you for this much appreciated

[@christopherjackson2157]

Best one yet. I think people find it hard not because it's complicated but because it's really alien to them at first. Once people have wrapped their heads around how the system works it's just math. And not even complex math.

[@twisted-t]

Thank you so much! I was really upset when I got Hynix memory and it was worse than the Micron sticks I had and couldn't get the same tRP but I got them to the recommended subtimings and the improvement was pretty noticeable!

[@kclazygaming8543]

Thanks for this it was awesome.

[@RandoTark]

Man, this is some awesome practical visiualization and explanation of some extremely complex stuff. Really appreciate it, I sure hope you do find the motivation someday to cover the write related timings, like the read stuff so far... would be awesome! Im wondering as well if you could cover which timings best help combat heat(especially on sammy bdie) on RAM (i know the refresh related ones do, not sure which ones beyond that do). I have some lower binned sammy bdie dual rank DDR4 and cant really get its primaries tight, but sure as heck do have the timings covered in this vid tight ... and i was amazed at how much more performance was achieved. I suspect I cant get primaries tight cuz I dont wanna run them higher than 1.42v (and I prefer to keep my IMC voltages more conservative as well) , cuz in prime95 + furmark .. the RAM temps manage to it 55c ... =/ so I can tighten primaries a tad but that means unstability due to lower temp tolerance , likewise if I try to make my trfc more aggressive, it lowers temp tolerance. Interestingly, on my kit it seems tREFI has less of an impact on temp tolerance than tRFC.

[@NVMDSTEvil]

This video pushed me to do some more tweaking on my 8x8 4Gb Samsung E-Die setup. Was at tRRD 6-8, now 4-4. tCKE 11 went to 4, tRC 55. Need to see what I can do to get the memory frequency higher now though, only at 2800mhz. 8x8 1T/GD0 maybe just a bit hard to push over 3000.

[@zoson]

Thanks a bunch for these videos. They were really helpful! How much of this series is applicable to DDR5?

[@79aimless]

Thank you Buildziod, us normies need people like you to even remotely understand this WTF stuff

[@VladKov36]

hehe can we bribe you for more of this series

[@ActuallyHardcoreOverclocking]

links in the description

[@ПётрБ-с2ц]

13:15 what's sitting between DQ and IMC? it can't be "delayed" by topology if there's nothing to buffer the signal

[@ActuallyHardcoreOverclocking]

There's some picoseconds between the output pin of the dram chip going high and the corresponding pin on the memory controller going high. So technically the ns I have in the spreadsheet is very slightly faster than what happens irl.

[@ПётрБ-с2ц]

@@ActuallyHardcoreOverclocking ah ok so it's fraction of memory cycle/2 slower

[@ActuallyHardcoreOverclocking]

@@ПётрБ-с2ц yeah it's not much but kinda note worthy when each data trasfer only lasts a fraction of a nano second.

[@ПётрБ-с2ц]

@@ActuallyHardcoreOverclocking actually my question was cause by you saying at 12:40 that "first burst is probably at the memory controller by the time ..." I kind of understand that you were talking about the CPU cores probably

[@ActuallyHardcoreOverclocking]

@@ПётрБ-с2ц yeah I didn't really think that through as I was saying it.

[@marcustaylor670]

I've found when overclocking a rig for gaming with 24/7 stability getting the memory and cache tuned correctly / better makes a massive difference. Wanging the CPU clock up seems to have diminishing returns as you go faster due to the heat and the potential of bottlenecks in the memory. The 'tech' channels can't be bothered to dive into the settings that matter and just want to throw hardware at any problems. I'm not the best at benching but I can get hardware to run games 24/7 much better than most people, I learned my craft on the AMD FX platform which was actually really good value for money when it got cheap. I'll definitely be playing with these settings, I'm tempted to dig out my old hardware and see what effects they have, thanks.

[@AdamKafei]

Thank you for covering this, I noticed I gained a massive amount of performance dropping tFAW from 32 to 24, now I know why. I'm also going to see if I can get it to do 16 at the possible cost of a little tRFC.

[@PhoeniXfromNL]

Thanks a lot for making these videos for us normies interested in learning how to overclock better

[@perdo7195]

Love tou man , I've learned a lot from your videos

[@benjaminchung991]

From a software developer's perspective, does it make sense to try and arrange memory access patterns (or more aptly cache misses) to exploit memory timing behavior, or should that be handled by the memory controller transparently to the "user?"

[@christopherjackson2157]

From software perspective you'd generally focus on minimizing cache misses and let the hardware do it's thing. In general. There are exceptions to everything. Aida for example has been written with the memory in mind because the programmers wanted strictly to measure the memory bandwidth and not the system performance under any sort of realistic workload. They wanted to control for (among other things) the cache across different architectures and cpus. Sequential reads larger than the total cache force the CPU to always be retrieving data from main memory. No clear example of software optimized for specific memory timings comes to me off the top of my head. If anyone knows of one I'd be really interested in hearing about it.

[@jallen418]

Thank you so much for this!

[@bentomo]

Last video or not I will continue to microdose information from other hour long rambling videos!

[@fumo1000]

I was wondering why my trrds,trrdl,tfaw wouldn't improve from 4,4,24..... every time i set tfaw less than 24 i get immediate instability. Thanks for the explanation... I'll try to set tfaw to 16 first with loose trrd and then tweak them

thx for explaining this to all

[@theoc1009]

Does setting the CAS higher (or less important timing) makes the memory more stable and let us to set the tFAW or other more important timings lower?

[@ravtastic9802]

given its astronomical affect on bandwidth output, i imagine AMDs setting for RDRD_dg is RDRDSCL? or is that RDRD_sg and RDRDSC is for dg, and SCL is for sg? but given that some chips can do SCL 2-3 ... what? edit: seems RDRDSCL is RDRD_sg...

[@mroutcast8515]

XMP - are almost always super conservative. Even "casual" (if you can even call tampering with RAM settings casual) manual tune will be far better than your average kit's XMP at the same memory speed. That's why these videos are super cool for all of us tempering with RAM for manual tune.

[@ActuallyHardcoreOverclocking]

12/12/54 is really really bad even by xmp standards.

[@GeneralSouthParkFan]

@@ActuallyHardcoreOverclocking Most DDR5 XMPs use tRRDS 8 & tFAW 32 from what I've seen, so at least they aren't as garbage in that regard compared to DDR4 XMPs

[@flamespirit1]

I liked this video. Make more videos like this