Reacting to people's RAM timings episode 3
Вставка
- Опубліковано 6 сер 2024
- My Patreon: / buildzoid
Teespring: teespring.com/stores/actually...
Bandcamp: machineforscreams.bandcamp.co...
Twitch: / buildzoid
Twitter: / buildzoid1
Facebook: / actuallyhardcoreovercl...
#Overclocking #RAM #DDR5 #DDR4 - Наука та технологія
These are awesome Buildzoid. Even though you sound annoyed at times doing them, we learn a lot!
Amen to that
Yes, he does sound annoyed at times . . . "CAS means nothing" over and over. But then he puts up some bench marks, CAS 28, vs CAS 40, (if I remember correctly) and the benchmark (primarily relating latency I believe) goes from 9.856 to 8.9XX something something. Ok, a lot of very nuanced technical information there, flashing by in the benchmark dialog boxes. Now, what can I do with that? Buy cheaper memory, get better frame rates? Or is it only a better benchmark number? I appreciate the basic effort, knowledge and information involved, but hearing that "CAS doesn't matter" 10 times, is not useful. MO anyway.
I love it. The more annoyed he is the more I learn.
Thank you for doing these, always learn something new.
This man fucks.
Most interesting setup was the last one. There are only so many B-die + Ryzen setups one can look at before getting bored.
I'll take anything that isn't about the most recent Intel/AMD platforms at this point.
Why not doing a video explaining the relations of the timings and what "algorithm" to use for inputing values at each and test them and also what impedances to try and why to use the specific values? This would make a lot more sense and help a lot more people than to just comment on random timings which ok you elaborate somewhat on your comments but its confusing in that way for the uninitiated to understand what to do for his kit.
That was my CJR, yeah some of those tertiaries are at auto. Was my first foray into Mem OC and CJR guidelines were hard to find. Was scoring around 275 on IBT and it passed HCI at 800%.
Thanks man. It’s pretty cool to see u keep breaking down these numbers. Very informative with ur way of thinking.
36:38 - incredible that you instantly noticed what is a literal mistake in my timings. tRDRDSCL and tWRWRSCL were indeed both meant to be 4! I think somewhere down the line after updating my BIOS and re-entering in my timings, I must have missed one because I checked my BIOS now and tRDRDSCL was left on auto haha. Thanks for the suggestions and excellent information as usual!
Should I post my AliExpress x99 ram timings build?
More X99!
Thanks for another ram timing react vid. It actually helps me understand what i should be looking for and how i should be tuning.
Fixing up my memory timings improved performance by 23% over baseline, these examples are really helpful. Thanks!
I'm the one at 17:50 Its a 3600 Cl16 B-Die kit. TRFC at 300 was giving me random blue screens every now and then. Raising it just a bit fixed that. That's why it's 324 unfortunately.
Makes sense. Very strange that it wouldn't do 300.
@@ActuallyHardcoreOverclocking Perhaps it's my old board and 7700k not being able to handle the kit. Would explain why it refuses Cl14 all together. I'm considering pointing a small fan at the ram kits.
My vipers didn't like 300 either. 310 was stable
Quick update. I increased my fan intake fan speed close to my ram and now looks like 300 is stable so far. I hope it says like that.
LOL, getting straight into the video and immediately just bashing people's "generic b-die timings". This is why I sub.
Thank you these types of videos are pretty helpful.
bz wrt curve optimiser, the only stresstest i found that works to determine stability for idle by using stressed loads was OCCT's small, variable, AVX test cycling and swapping (ie, core 0 singlethread, then all cores except core0, then core 1 singlethread; by checking both cycle and swap options at 10/5s each)
i had a setup that had passed literally an entire day at work doing corecycler on both prime95 all sorts of sizes, ycruncher tests, all of it no errors; yet closing it and youtubing for a bit it just blackscreens. the unstable setup also happily did a couple of days of X265 encoding without crash; and gaming without crash.
OCCT avx showed immediately that cores were spewing errors galore. and only once occt was satisfied under avx singlethread load, did the idle blackscreens stop.
like i was amazed some cores passed corecycler/p95/yc for over 12 hours with no issue and spewed errors instantly from occt. like to the tune of one core having to go from -21 passing to -8.
can reccommend. went from pulling my hair out over multiple full days of testing to fully dialed in and rock stable within an hour.
i dont think the recommended corecycler methodology of using SSE workloads to stretch the boost is viable. (occt also passes SSE with way too low an undervolt) it just lets too much instability pass that avx will kill it. i think better to just pump EDC till it can mostly-sustain max boost on a single core with avx. between cycling between single-thread and all-but-one threads this keeps the cpu hot to ensure pathological voltage requirement.
also as a side note, once i was occt stable on curve optimiser, i could just put the boost override straight up to +200 and no issues whatsoever. i somewhat expected the curve would need backing up a bit but apparently nope.
corecycler may well work if you use avx aswell. but it doesnt keep the cpu as hot as alternating between ST and mt.
I learn something every time I watch one of your videos.Thanks!
I've gotten so much out of these. Thank you
I was *literally* looking for this on Twitter earlier in the day and figured "ah, well, I'll see it soon I'm sure".
And then I completely missed it. You're too fast at getting these out man. 😭
Spent three to four weeks off and on working on my current profile and finally got it where I want it. Guess I gotta wait another month for feedback.
Video idea: what is memory training, how it works and why do we even need it if we are already setting bajillion timing settings in BIOS?
Your reaction video 📹 was to the Moon 🌙 🤣 and back BZ! I like how you critique the postings on Twitter in a way that is very entertaining 😄! Way to go BZ!
I vote RAM timings once a month! :)
36:17 I have the same board and it set crazy VDDG and VDDP voltages. I found that lowering those helped me tremendously. Also, tRFC doesn't calculate tRFC2 and tRFC4 properly, I have to dial them in manually with the formulas or else they are fucked up, as seen in his screenshot
You could do a vote-system for those videos, where you let the community vote on which timings look interesting on twittrer, before you give your 2 cents to each of them. That way we would have less generic ryzenb-die configs and hopefully no 3200 cl16 xmp setting in here.
Thanks and respect to the smart guy at the end with the 5700G @4.9Ghz and 4733 cl18 ram at 1T LOL That's one fully pumped system and exactly what i want to build lately xD
This is excellent information and helps me squeak away from XMP after making the jump from 32GB of Micron E to 64GB and losing everything about my old profile. Thanks.
tnx for roasting my RAM, too, BZ, 35:34. they;re B dies but i think pretty not good ones. the primaries are at they're best, can't go lower with them, and at the secondaries, still some refinements to do, but it;s good to know that this random RAM OC failed is normal for msi boards :))
Bro, how do I always miss these on Twitter? Will you do another?
My guess for some of these oddly specific tRFC numbers are increments of tFAW 16 i.e. 256/272/288/304 etc... Is it possible to find out if tRFC really rounds up the nearest tFAW multiple or is that just one of those RAM overclock gospels?
After watching this, I decided to OC the RAM in my daily system. Bricked my board and corrupted my OS. So that was an expensive lesson.
Lmao, howd u manage that?
You mention the curve optimizer..interesting, I been having this reccuring freeze problem.. if I get crashes only in idle, randomly freezes my windows, I can still move mouse and even use softwares for a bit but the windows bar at the bottom is frozen, then eventually it freezes, have to hard shut down (can play hours without issues, only when alt tabbing it can happen sometimes)
I tried so many things, thought it was my memory OC, but it happened even with ram on auto (default speed).
I tried to put a 4650mhz full core OC last night as I was suspecting it could be related to boost or eco mode, 8hours fine on realbench, I open a browser and instantly get a freeze...
really informative to watch, spent a couple weeks by now playing with my ram but seeing what other people are doing with similar kits makes me want to dive back in and try some things.
though I do find ram testing to be the hardest thing ever >.> used to do what I call "short" tests but what I got "stable" with that led to memory could not be read errors wile playing games, now switched to a "long/normal" test duration and I'm ending up a lot more stable but I can still pass TM5 with the extreme777 preset and games will occasionally still run in to memory could not be read errors, doesn't go away until I back off certain timings...hard to test for when you need to change something then try to go hardcore in to gaming to really test if it worked for about a week :P
but to my surprise my cpu's minimum fps in forza horizon 5 went from ~79 > ~91 so yay progress/results (as I don't know what is up with aida 64 but stock vs tweaked to hell and back = just some bandwidth differences for me latency is not changing at all >.>
try stress testing the RAM while the GPU is under load. When playing games the heat from the GPU can heat up the RAM which can cause it to error at settings that work at lower temperatures.
@@ActuallyHardcoreOverclocking >.> that makes way to much sense tbh.
the ram is practically located above the GPU VRM so when hitting it with a load for a couple hours....yea ic how that might cause the temp's to change to the point of causing a problem (my exhaust is also only on top so I'm for the most part pulling the heat up in to the ram which....probably is not helping :P
oh well maintenance soon where I'll be upgrading the fan's with something more silent that I can run faster hopefully moving enough air to....lessen that problem (my current fan's are...in the grand scheme of fan's not actually that good at being a fan :P....at least not a silent one while still providing decent airflow/pressure), case is a Lian li 011D with a bottom/side intake and top exhaust so to move air around the CPU area basically requires just ramming as much air in to it as I can....should work out I think :P
I had the same issue with curve optimizer passing any hard stress test only to crash when doing nothing. I think Windows default power management is the one messing things up, because once I set all power consumption option to NEVER (example shut down disks after 20 min of inactivity), the issue was gone for most part. I can't be 100% sure because I was testing differnt things and having shut downs related with the CO....but not when idle as far as I can remember. Anyways I think the -30 offset that seems to be the generic number is very optimistic unless you have a platinum sample, great cooling and great overall bios settings. Going for a mix of -5 and -20 depending of the needs of each core seems to work better. Btw my ram sucks, Kingston "fury beast" Hynix AJR, won't recomend for OC or great performance.
I should pay attention to when you do one of these but I never open twitter... I think I have a reasonably interesting setup (:
Straight 20.0Ω for drive strengths and 36.9Ω ProcODT are Auto rules for that MSI mobo
Im on a Crosshair VI Hero and I have the memory set to 1.46v. ZenTimings reads 1.54XXv. Anyone know why? Similar behavior with CPU vCore.
i kind of wonder since you mentioneed that GDM off / 1t / 2t isn't much of a difference if that may be kit type dependant? with my current B-die I fully agree I put a bunch of time in to getting GDM off/stable at 3600Mhz but when I checked the latency it was within margin of error where on my previous kit (hynix CJR) getting GDM off was by far the biggest boon to latency out of all the tweaks I tested knocking off 2ns...i kinda expected the B-die to have the same appreciation but :( though my previous kit was also 2x16 where my current is 4x8 (3900X) so maybe that's why? or it has to do with the setup times I'm using...i honestly have no idea what it does or what I'm actually manipulating by changing it but I do know that 0 was instacrash 63 was almost instacrash and 51 seems pretty stable but might try lower to see if that results in more stability when I try to tackle TRDRD and TWRWR as recent tweaks destabilized my settings there >.>
I think 2T GDM off is better than GDM on with 1T on many duel rank systems. I tested mine and 2T not only had better stability, but lower latency than GDM on 1T. So yeah single rank you DONT want to be running 2T however
Hi @buildzoid and thank you for commenting my settings from ~9:00-11:30min.
you said that my Drivestrength is kind of whack like 20-20-20-20 on my unify-x but it sets it on auto. should I change them all to 24 or so? This maybe helps to stabilize cl14?
thanks in advance :)
I guess MSI must've changed that in a BIOS update at somepoint then. It can't hurt to try all 24s.
Is there a way to send you my timings when I dont have Twitter?:) Thx
Holy shit that was quick BZ.
Hey, Buildzoid.
My r5 5800x can run at 46,5ghz and IF 2133 on 1:1. Is that common or just me?
With my r5 3600 i couldnt get past 1900.
I am playing with timings and would like to show you if you want
Do you do paid ram oc? Tried to find contact details blog cant be found :/
Do you know any ddr5 motherboard could dump RMT results out?
I’m running 16-16-16-30 at 4000/2000FCLK at 1.46 I’m not remembering secondary timings but I didn’t spend much time on it.
Quite a few 5000 series ryzens won't do 1900 bclk. My old 3700x did, but my new 5700x flatly refuses to run it. I was highly disappointed, however other than that the chip murders the 3700x so still happy. Also..I never see these things until after you review them - have some rev-e timings I hope you would like.
My 5800X3D wont do 1900 either. Tho one would assume it being the newest and B2 revision it would do it but it's the same as my old 3800X stuck at 1866.
any way to enable xmp on 128gb ddr5?
I'm thinking I might post mine in the hopes for another episode I set my samsung ddr5 5600mhz cl36 to 6000mhz at command rank 1T with no additional voltage and I'm kinda hoping 6000mhz 1T 1.25v at 5600mhz xmp timings is seen as tight
*Or at least somewhat tight*
ty
the cjr part was actually the best part about it, wrong memory timings is better than the good ones :D
where download mem tweakit?
You complained a lot about tRFC being above 300, so I went and tried to tighten it on my kit but I can't go below 480. Could it be because of my motherboard? I have an MSI B450 Tomahawk Max.
do you have bdie?
If u have micron thats good
I can’t go below 534
@@nickpowell7432 Ah yeah it's hynix, that explains things
👍 I hope I see you collecting submissions for the next video.
Mine are something like
16-20-20-38-74
Last number is probably wrong but its micron crap running at 3200MHz instead of stock 3000MHz
32GB a stick 2 sticks
3266MHz works but higher doesn’t so i use 3200MHz
And my timings i cant touch unless i wanna push them up and make it worse
Mine run at 1.45v as set by the motherboard not by me but they sit at a chilly 40c so it means nothing as hotter is better to a point anyway
Keep missing these posts lmao
So much weird stuff in these screenshoots!
So for all ddr5 potential owners there is an issue with Samsung 6000Mhz CL36 Kits and CL40 from Gskill. MSI carbon and Strix E have issues running it at xmp or even fine tuned. Hope billzoid can expose this issue and call out these manufacturers
Awww man, I wished my setting would come this time. I could really need a qualificated opinion of it.
new AGESA (1207 b2 iirc) fixes windows 11 EDIT - My board (X570 Elite) defaults to 4-5 for SCL's. Gigabyte boards are weird.
"just oof"
Yo bz, just wanted to let you know there's a $500 3090 on ebay right now. Homie said it's unfixable because two shops gave up on it
probably a 500USD brick then. Just because some cards are repairable doesn't mean all are. It's really not that hard to find AMD and Nvidia cards with dead GPU cores.
@@ActuallyHardcoreOverclocking Ah, true, I figured you were its last hope, didn't know if it was worth the gamble of getting stuck with some SMD's you probably couldn't use anyway
seen your vid on the samsung m die too late..........patriot viper ddr4 3600mhz 64gb kit 3200mhz stable
GO AWAY 🤣
33:45 probably broken OS
Is it true that tRFC gets rounded up to multiples of 16?
For example, if I set it to 320 I get 320, but 321 will be 336.
AFAIK no. Also there's not really any way to test it properly unless you had DDR4 that can handle 2 digit TRFC.
B-Die Dual Rank 4000 MHz 17-17-17-30-320 2T in G1 on 12700k while running on iGPU - impressive due to iGPU or w/e? :D
IDK if the iGPU makes a difference to the memory stability.
@@ActuallyHardcoreOverclocking Me neither. Added that comment as a joke.
I'm Carlos Díaz on twitter in that video, I tried 3733 and I get errors on khartu and other tools unfortunately
Ohhhhhh, you didn't get to my post :/ Oh well, you can't cover everybody I guess...
post again the next time :P
@@bananya6020 Yeah, but it takes a lot of luck to check twitter in the exact few hours between him posting and making this video. But will try!
I ran out of the bathroom half nekkid and kicked my kid off the computer mid game so I could post mine and still didn't make it.
@@jayferasaurus lol
@@jayferasaurus 😂
tWR is weird and can show rare errors. You probably not checking stability very well. tWR could be that loose (14 and even 20)
32:24: Windows 11 be like 🥴
O man hun saying 62ns is bad looks at mine o no
Don't take this the wrong way, but didn't you mention in a few of your older videos that setting tertiaries is tedious and not always worth your time? It just seems like you're putting an awful lot of emphasis on them in these videos, but you never really stress the importance of setting them. I would wager that most of these dudes don't set them because it's a hassle to have to worry about them and re-run memory tests.
The PC overclocking/enthusiast space should be a positive environment where we learn from one another. If these videos are meant to have strict requirements then please mention it. I felt like you went really hard on a few individuals for little to no reason.
A lot of submissions are probably just because they want a feature. It's a matter of acquiring good bins and investing time. That isn't really something most individuals will be able to manage.
bz roasts whoever he wants m8, what do you think these videos are about?
@@Wasmachineman My apologies, I had no idea these were roast videos. I thought this was a science and technology channel.
You can't handle an ounce of critique? If my timings were shit, I'd like to know. Honestly, never go to university. You'll quit just because the teachers call you out on your shit. If it is a good Uni.
Honestly, if I was Buildzoid and seeing the same mistakes made time and time again...don't people read about how to do RAM OC? You might even learn something.
You can see him mentioning secondaries and tertiaries in just about every video about RAM he overclocks IIRC.
I would say, watch more BZ videos, watch from beginning to end. And even then, do your own research. Find something written about timings, explanations. Look through the specs of your RAM, the manufacturers PDF. It's like 200 pages. And do your own testing. Lots of it. Only you have your hardware. We can't do it for you, even if we wanted to.
@@dainluke Most of those settings don't do as much as one might think in day to day usage. XMP is honestly fine for most people. Dig deep into those tertiaries if you appreciate and enjoy the process. This is by consequence just a video about optimal timings and comparing memory setups of different platforms. And yes, it is extremely tedious to optimize tertiaries and probably never will be "worth your time".
@@MrPunkassfuck I wasn't featured in the video. But other than that, I've watched almost every BZ video since I found him in 2019. He has said on numerous occasions that he doesn't think setting tertiaries is worth the possible instability and effort, if you're just trying to get a decent OC rock solid (in terms of data rate, primaries and secondaries).
I do OC my own stuff and I enjoy OCing, I also never ask for help because I don't expect help. I adopted D5 and I've been doing my own settings for months now because there was no help to be had. There's a difference between criticism and shitting on someone. A fairly large one at that. It's okay to be knowledgeable and to instruct someone when they're making mistakes, but I don't think anyone who submits XMP or mediocre settings realizes what they're signing up for. He should explicitly state it. That way it doesn't just become a roast fest.
Your comment is tremendous. The pretentiousness. It's astounding how you just assume I've never read a data sheet or researched RAM or OC'd myself. Amazing, truly. No for sure bro I know nothing. I can't be as cool as you and your crew of RAM OC Gods running 7200 32-42-42 with your 57X1/55X3/53X8 12900KS configurations.
On Ryzen it really doesn't matter. The only gain you get is instability and erratic behavior. Just get 3600 kit with tight timing and load docp settings and leave everything on auto.
Good Old Gamer has tested many Ryzens lately with tuned timings and the gains are pretty good.
Sup
Man I wish twitter actually sent out notifications. I have a 2666 10400 I would have liked to post
next time maybe?
WHY WHEN I WRITE RYZEN DOWNCLOCKS AT 75C IT GETS REMOVED FROM R/OVERCLOCKING. ARE YOU STILL SHILLING FOR AMD?
@@erisium6988 what that got to do with what i write?
Hey @buildzoid. TBH I've stopped watching after first example. There's nothing to gain from it for me. How do I know? Because I've watched first two episodes before and it was the very same tone. Waste of time. Let's cut to the chase - I have generic ryzen setup - some bad b-die bin and 5800x. You said on previous episodes "it's not interesting, it's boring, etc." yeah but let's face it - there's a lot of people with such "generic" setup and timings taken straight from DRAM Calculator. When you want to improve and get somethin better (than DRAM calculator) from your kits, you go to the youtube, buildzoids channel, watch reacting video, seeing setup like yours "oh lets hear what he says" and you saying what? Timings tight, timings loose, fine Voltage, le fin. Nothing more. Instead of saying: try changing it, try changing that, I would've do this that way, etc. Don't be that guy.
mine 1-1-1-1 :>
that's illegal even for SDR ram :)
that's a DNS server not ram timings.