responding to a comment about tRFC and tREFI timings

My Patreon: / buildzoid
Teespring: teespring.com/...
Bandcamp: machineforscre...
Twitch: / buildzoid
Twitter: / buildzoid1
Facebook: / actuallyhardcoreovercl...

[@JeKramxel]

You've improved your delivery a lot through the years, and you're getting your point across much better, these days. The knowledge was always there, but being able to explain things like these to a common audience, is a great feat! Also, don't get feel discouraged with people for them failing to realize the scale at which things happens in the IT world. People from other areas are just not accustomed to thinking like that, and have hard time to even imagine things at that scale. I've had the same problem, when trying to get my points across to people with a non-IT background.

[@thephlophers]

It's good. I could follow this, but I will say if it was even just clicking through some low-effort powerpoint slides it would be quite a bit better still than the classic windows calculators-and-notepad presentation method.

[@Mickulty]

You may not like the AIDA64 memory benchmark, but you can't deny that the result IS what peak performance looks like.

[@TheFoxyCoffee]

point is that the "peak performance" is irrelevant, as you won't encounter it in daily use. not to mention that those numbers that aida spits out are inflated by l1 cache, l2 cache, and l3 cache, so not really very reflective of the memory's own peak performance, but rather your system's

[@concinnus]

@@TheFoxyCoffee It's a meme, dude, don't take it literally.

[@Mickulty]

@@TheFoxyCoffee did i say it was a relevant benchmark De-snarking edit: I suppose since the joke was missed I should explain. "You may not like it, but this is what peak performance looks like" is a meme, generally as a caption for photos of things that look ridiculous. In the broader internet it tends to be preceded by "This is the ideal male body", but that's often dropped in the tech space because a modded gameboy is clearly not a male body. BZ (rightly) doesn't like aida64, and late in this video called the numbers it spits out "peak performance" (as opposed to realistic performance). Those two things together happen to fit the meme format well, so I was happy to point that out - and imply AIDA64 memory benchmark is ridiculous in the process.

[@gtijason7853]

@@Mickulty That was good bro !

[@nexushexus4365]

@@TheFoxyCoffee whoosh

[@LinkStorm13]

Thanks BZ, I've learned a LOT in those two videos.

[@Airwave2k2]

17:16 10 µs are 100kHz a little bit on the high side ;=) Looks like whales and bats have a range to "hear" that high of a frequency. Question is if there may not be a subharmonic that actually could come into enough resonance and lead therefore to an audible sound.

[@umeng2002]

Our primate brains aren't good at conceptualizing scale differences.

[@gamingvibrations5320]

"The developer decided they wanted a seamless transition" now I know what causes massive stuttering in Star Citizen! 😁

[@Vinterloft]

Separate PCIe 4.0x4 NVME drives for OS and game, or Optane, fixes that. You'll have stutters especially in quantum drive no matter what RAM you have until you have those sorted. That's why I always keep a wet fish handy to slap people who buy B450/ B550 motherboards for Star Citizen. RAM speed won't ever matter because it's a double precision game world that will thrash I/O even beyond 64GB of memory. And it's only going to get worse, not better. Contrary to popular belief Star Citizen is already incredibly optimized, it's simply just not feasible to optimize any more for PCIe 3.0 or slower storage. That's also why Star Citizen is the game that leaps ahead the most on a 5800X3D.

[@rapamune]

@@Vinterloft What is the benefit from the separation? technical explanation works, curious. i suppose you'd want the pcie lanes routed to the cpu (i.e. x570 over the b550 motherboards)?

[@clearlight293]

@@Vinterloft B550s have a PCIe 4.0 M2 slot with the lanes from the CPU; only the M2s with the lanes from the chipset are 3.0.

[@andrewryder3075]

AIDA's "Memory Bandwidth" should be called "Maximum Theoretical Throughput", since it's based on sequential read/write (which is NOT how data gets stored in RAM). It's even less "real world" than "Sequential read/write speed" on SSDs - (at least that happens SOME of the time).

[@samiraperi467]

It's not even that unless your timings are optimal.

[@andrewryder3075]

@@samiraperi467 Well, "Maximum Theoretical Throughput for current settings" is what I was implying.

[@andrey7268]

To add to your reasoning, Buildzoid, it's not like the world stops while DRAM is refreshing. CPU can still do a lot of work within its cache, and GPU is still able to render the frame. Though GDDR also needs refreshing, but when that happens the world still goes on - CPU can do its share of work. The real cause for stutters is likely not even shuffling data in RAM. Noone's really moving gigs of data in RAM - any half-decent developer will know to avoid this. It's loading and decompressing stuff from storage and pushing it to GPU RAM through PCIe that takes a lot of time - order of magnitudes higher than shuffling data in RAM. Then there's inefficiencies in game engines, like garbage collection kicking in or poorly optimized scripting or resource management (e.g. too many AIs taking too much time to process, too many objects to calculate physics on, too complex scene to render in a frame budget, etc.).

[@EpicBunty]

Nice. Tell us more.

[@VTOLfreak]

It's actually even less overhead as the DIMM can refresh one bank while the memory controller is reading/writing in another bank. this is called Auto Self Refresh (ASR).

[@ehhuum1480]

ASR is related to suspend-to-ram, where the mem controller doesn't have power. You're thinking of REFsb.

[@christopherjackson2157]

The trick is to only play fortnite on ln2 lol There's a fair bit of variation in the range of human hearing. I remember from some experiments I ran in uni that we had to go to around 20khz before absolutely no subject could hear a sound. We were studying language cognition in infants versus adults.

[@c3h8o69]

Thank you so much for this content, love it

[@VeiiTM]

I think the core factor lies in the understanding of the problem vs the question told The question is making a leap in arch design be anywhere comparable to consumer "gaming" or normal computational tasks Comment writer and like you mention seems "most community" see's tRFC refresh interval to be performance beneficial. This is an issue on understanding the scale. And you likely only ranted about it, tho visualized it quite well I was taught that "peak bandwidth" depends on the efficiency of the architecture (yes) BUT also the losses with it. In such case, the perspective is reverse ~ you always start with the full usable bandwidth outside of IMC design losses minus! dram-busy times ontop (Design allowed bitwidth already factor in losses and you always start with full 100% usable bandwidth ~ well "usable" but before on-dimm losses) Realistic usable bandwidth, then is much lower ~ but it's one you can shape (it's a platform difference here) For tRFC to make any kind of "noticeable" differences in FPS ~ it has to be a magnitude slower and everything else around it flawless. But ~ reality is, it won't. It can't. It by itself in comparison is too small and never a bottleneck on the user-timings design ~ yet a bottleneck on the transmission and bandwidth design Every timing secondary or tertiary needs to be on such a tight level ~ for a shift of tRFC to actually start and cause a "potentially higher usable bandwidth" or let's call it "fix of issue" Yet it won't. The only reason to focus on it, is when you target specific bandwidth values and target improvements by combinations. Not improvements by delay. The improvement that happens when "by accident" two of your timings are correctly aligned and not only pass on-imc & on-dimm package delay correction/slowdowns. Soo in laymen's terms, It can not show improvements in the human interpretative field or even FPS field. - As first all of your timings have to be flawless without autocorrection (which they 99.8% won't be, hence IC-Vendor timings & PCB-Vendor timings are dynamically asserted & non visible) - then your workload needs to utilize 100% of the memory bandwidth and ! the capacity per side. Be per IC size, so dataset size aware - and at the very end, be in a steady voltage and thermal controlled environment ~ soo discharge time doesn't out of nowhere change and tRFC usable, shifts In order to predict, measure, and visualize actual percentual improvements of a tRFC change. Just hence all these scenarios are very unlikely ~ it is a design timing, but not a performance timing. You never reach even close to the realistic bandwidth limits ~ to have to turn towards tRFC and tREFI changes AND then even see ceiling improvements It is and always was only useful for diagnose purposes ~ to differentiate between user timing flaw (DIMM autocorrection) and random-instability caused by it "Am i missing peak bandwidth points because my timings autocorrect" vs "Am i having big run to run variance because my refresh timings are too low and repeaten" A diagnostic thing with a diagnostic purpose on a controlled environment scale. So is also Aida64 & SiSoftware Sandra Inter-core bandwidth tests. You play with a different scale on different transmission delays. "Humans" , "Games", are irrelevant for this timing. Soo the perception build around this timing is wrong Gamers do not matter and random workloads ~ if you want to shape it If it's too low and you increase ~ if you see an improvement on the ns scale, wonderful. You fixed a flaw If you don't ~ then you don't reach efficiency for it to matter at all. Even at 400ns 👍 It's really more of a problem diagnose timing, than anything else & the potential problem increases by dimm density. Community tho has a flawed perception of it ~ soo ty for the video :)

[@VeiiTM]

EDIT: Shifting in this sense x user picked value (which is 12 digit decimal ns value, not whole digit value) - by the same factor of IC's you have on dimm Can be a good check to see if you are bottlenecked by it or by timings. Bottlenecked by only potential peak bandwidth ~ never by "usable performance". Well, words & meanings are hard SR usually has 8 ICs Shift it by 8 tCK ~ if access latency improves, wonderful you had autocorrection & maybe an issue why it's unstable Does it do nothing (often the case) ~ then all your timings are to blame (well always where). DR with 16 ICs, it's +/- 16 clock And so on In general, never expect performance out of it ~ it's not there for that and never was It's only usable for diagnose, hence in combination with perfect world tests it can or can not show if it should even matter for your case.

[@mohammad-zabih]

for years ive been confused with this trfc and timing meanings till i found this god of information , i hope you live like 300 years so next gen ppl can use ur knowledge as well as we all did man , TY

[@Si7encer_]

So when you commented at around 2:45 and stated that if you’re on 12th gen your timings should be around 32000-250000 was that in reference to DDR4 or DDR5? And if I am running Aida64 and getting better latency timings running 32000 as opposed to a higher number what should I be more concerned with? And what would benefit me more in gaming?

[@xlinnaeus]

lmao Taylor got put up in front of the class...

[@MIK33EY]

This and your previous video are a great explanation of things👍🏼🤝 For someone who is new to building their own pc’s, wants to learn RAM timings & overclocking/benchmarking and finds the BIOS completely daunting, can you recommend a good self teaching resource?

[@josefondas9080]

Hi, your videos are entertaining and really informative so I often use them for my school research. I once heard that DRAM degrades over time and it's losing its capacity. So after several years you should change it to keep your PCs performance. It seems false to me and I argued with the teacher, who said that, but I'm not able to find any statements that would approve or deny that without getting too deep so I can't understand the terminology and text. Your clarification (or even anoyne who knows the truth) would be really apreciated. Thank you EDIT: I'm convinced that if the capacitor failed (my imagination of loosing capacity) the whole DRAM would fail and cause BSOD.

[@Vegemeister1]

All chips that are actively used degrade over time, but it is a degradation of electrical characteristics, not a degradation of computing performance. A chip that is degraded too much will stop working at all or give incorrect output. The designers put in a safety margin so that even a chip that was not great when it was made will still work after many years. When you overclock, you are using up that safety margin, so overclocked parts should be re-tested for stability every now and then. However, it is possible to use things like error correcting codes to "fix" errors produced by a degraded chip. For things like flash memory where really sophisticated ECC is always present, a degraded part can still work but perform poorly because it is spending a lot of time correcting errors instead of doing useful work. But that wouldn't have been possible with consumer RAM before DDR5, and for server RAM with full ECC you would see a massive amount of corrected errors in your logs. This might also be a language issue. Capacitors have "capacitance", not "capacity". If you say "capacity" when talking about DRAM, people will think you mean the *amount* of RAM, measured in GiB.

[@RandoTark]

Yep, a lotta questions like this commonly fail to consider how much faster said component is actually performing said operations then we can conceive. I catch myself in that sorta "gotcha" here and there when thinking about "scenarios" like this question brings up. Once you realize its executing these things far faster than human perception, you realize quickly that "yeah it most likely doesnt cause what I think it might"

[@zazabean23]

Somehow people interested in input lag fail to realize that they use a system which doesn't run in their mind so their perception of nanoseconds/microseconds/milliseconds is irrelevant, yet its the most common flawed argument, to say such small numbers don't matter because they compare it against their perception of time. Once u scale the time to human perspective from CPU perspective it ends up being minutes, months and years. Also, your CPU does millions of reads and writes a second, so multiply that latency by same amount and now you see why it matters so much. RAM latency affects everything, every single memory operation your OS does, everything your game does, everything your devices do with RAM, so it affects input, ofc. Even if your framerate stays the same, RAM latency is the most impactful thing in entire PC after CPU clock/ucore speed that affects input lag, yet so many ignore it. To put this in simple perspective, RAM latency is equivalent of time it takes for you to remember something.

[@RandoTark]

@@zazabean23 hrm.. not sure how that relates to what I was saying. Also, while ram latency is part of the picture, it's a miniscule part especially when u are gpu bound. Furthermore, input lag is caused moreso by syncing issues and workarounds with monitor refresh and frame draw timing.

[@zazabean23]

@@RandoTark just because your pc can operate faster than you believe you can perceive that doesn't mean it's true. There can be 0 change in FPS and it would still affect system latency + other things.

[@RandoTark]

@@zazabean23 I think u are focusing on the end result.. which is fine, obviously thousands of variables add up to produce the end result of what we perceive in a pc. All I'm saying and like this video is saying. A couple hardware timings that operate in the area of nanoseconds for a tiny percent of time would not perceivably cause stuff like stuttering. Although it's easy to believe it could be when u fail to consider the time scales between human perception and the actual rate of execution of said timings.

[@zazabean23]

@@RandoTark Okay thats fine but I think it's its important to realize that even if you had a fixed frame rate it's measurable with the right equipment and humans can perceive it since it's not just one operation it's many.

[@LucaRuggier]

Hey BZ, love your video, just wanted to comment on the MHz vs MT vs Mbps. I am a computer engineer and honestly don't see anything wrong with using MHz. Hz aka frequency is just a rate measurement, it doesn't have to mean that a certain clock is oscillating at that frequency, If the RAM is making 3200 operations per second, that's still 3200 MHz. Just my 2cents

[@christopherjackson2157]

Yea I totally agree

[@ActuallyHardcoreOverclocking]

the thing is all the timings are in clock cycles. It's only the data transfers thar are done at 2 per clock cycle.

[@jdogdarkness]

@@ActuallyHardcoreOverclocking Can you please send me a link to your Ryzen memory bandwidth/stability/voltages google document? I've been looking everywhere for it.

[@wotertool]

I appreciate the video, quite informative and as stretched out as I am used to, however, the big takeaway is that now I do want to hear a DDR Orchestra, but from RAM sticks, not the germans.

[@LucyMai92]

I changed tERFI from 5.5k to 262k and memory bandwidth increased from 87 to 89GB/s and latency decreased from 66ns to 58ns. I would say thats a significant performance gain.

[@h1tzzYT]

no, if you only see it on only one program and not even in actual real case scenario, you know where it actually matters?🙄

[@DrWillz1996]

@@h1tzzYT it will absolutely improve fps in non gpu bound games. I guess it could be good for people playing with insane refreshrate monitors but other than that it's pointless.

[@pinnacleexpress420]

​@@DrWillz1996I swear there's more to it than just being gpu bound, could be wrong, but I'd say it most likely has a much bigger impact at 1080p, regardless of GPU utilization. I seem to have made gains in my visuals by overclocking RAM, reducing resolution via some 50% render resolution setting, and then cranking the anti-aliasing to the max, while saving on heat by undervolting my gpu. Ymmv.

[@mikk150]

Plus cpu cache still works under refresh

[@mrlithium69]

i consider 3.75% is kind of a lot, but that would never be a realistic assessment of the true impact of this. this is like 3.75% of 1% of the rest of the system.

[@Goldenhordemilo]

band camp good it sounds mint scale tracker spec

[@peterderbeste6817]

"going from no refresh to this ddr4 configuration, you would go from 1000 fps to 960 fps, because 3.7% of the time the memory is refreshing" I think you said. This is only true if your memory bandwidth is the SOLE bottleneck to your 1000 fps setup, wich i would consider slightly unlikely. But i really appreciate your point about the timescales, the refreshes are not responsible for stutters for shure.

[@russkubes]

He covers exactly this point around 18:58

[@peterderbeste6817]

@@russkubes that does not change the fact that the "3% refresh time means 3% less fps" is nonsense. also, thats not really the point he made at your timestamp

[@ActuallyHardcoreOverclocking]

since the video is about RAM it's convinient to just assume that any workload discussed in 100% memory limited.

[@Vegemeister1]

It works as an upper bound, and the point is showing how small the effect is.

[@markotodorovic5744]

Buddy can you make some guide how to OC ryzen 5950x for best performance 🎭 for gaming on MSI board! You are awesome thx for amazing videos!!!💙

[@alsteooaai]

hi, under € 350 what is the motherboard that barricades / limits less the (high) frequencies of the DDR5 RAM? With the highest maximum frequency limit of the RAM, which supports the highest possible frequency in OC pushed DDR5 (with top-of-the-range CPU and RAM working at higher frequencies than those generally supported by the Z690 mobos, which therefore in this case could represent a limit and block / run RAM and CPU at a lower frequency than the max they could reach, unless we are talking about Asus Apex or very expensive mobos), then mobo that does not impose a frequency limit on CPU and RAM and letting them work at their high frequencies. I foresee in the future to probably put an i7-13900K (keeping the same mobo) so the only thing I need is that it is a model of mobo that lets CPU and RAM work at very high frequencies without barricading and not limiting CPU and RAM to the limit frequency of the mobo. With high max frequency limit (the one indicated by the manufacturers in the data sheets), but above all true in a real scenario, because I see that in the data sheets some models have indicated 5333MHz, others 6000MHz, but some say that all mobos with 4 DIMM / slot RAM barricade / limit all to 6400MHz (frequency of ddr5, even if these and the CPU could theoretically work in OC even at a little more than 6400MHz), hoping that this is not true, I am looking for a mobo suitable for OC pushed RAM, the goal is to try to bring even more mature ddr5 banks that I will buy in the future (with 13900k having better IMC and SA) at the highest achievable frequency, perhaps between 6500 and 7000MHz.

[@BTAT2101]

On my 5900X /X570 Pro I have a trefi of 14029 or 7793 ns. I cannot change this in the BIOS somewhere. Can this setting not be changed under Ryzen?

[@denvera1g1]

Sees title "Responding to comment about tRFC and tREFI" My paranoid brain: "Oh no, this is about my comment on the DDR5 video where i reversed the two"

[@aeon7748]

by chance, i found that there os some combination of trefi and trfc that will give the best mouse feel, i think yes there is an effect, and no it isn't placebo, as I've tested this on my brothers rig and he confirmed this feeling when asked.

[@pinnacleexpress420]

I found a bigger effect from cwl and everything associated with it: wr, wrpden, wtr, and wrrd, as well as RttWr (!) and RTL/IOL. Im really surprised I never read anything about this personally. You ever figure anything out about aonpd?

[@EpicBunty]

For me when I put tfaw back to my stock jedec value of 24, everything became so much smoother. I realised that it's not just a matter of tightening them all and checking Aida. Also the iols and rtls are super important to lock in for consistent testing. Also raising twr gave me a lot of smoothness. Likewise reducing tras was also very noticeable.

[@PhilDaw]

more comment burn videos please

[@toonnut1]

Whats the most accurate benchmark tool to use for memory?

[@ActuallyHardcoreOverclocking]

Y-cruncher for bandwidth / SuperPi 32M for latency.

[@toonnut1]

@@ActuallyHardcoreOverclocking thanks for the reply. I've been trying to overclock my memory (i haven't got a clue lol) and it's passing y-cruncher and TestMem5 but crashing while gaming. It's so frustrating to see it passing benchmarks and stability tests but crashing while gaming.

[@Frej84]

I think he is answering what tools respond the most to latency and bandwidth changes. Not which programs catch memory instability.

[@toonnut1]

@@Frej84 well I've found out that black ops cold war is great for finding out instability lol 🤣😂🤣

[@lightcity8933]

Thanks king 😔✌️

[@killerrf]

So would a too high tREFI allow memory to pass a test like mem test then cause random issues in things like games or other day to day programs or not?

[@ActuallyHardcoreOverclocking]

If the TREFI is way too high memtest will detect it. If it's only a little too high it causes randomproblems.

[@killerrf]

Ok thx

[@EpicBunty]

​@ActuallyHardcoreOverclocking what are the best memtest profiles ?

[@H4x4t3hN00bz]

felt like I watched university lecture

[@noreng9333]

University lectures in mathematics usually assume you're well beyond an introduction to exponents...

[@penguinton7691]

Buildzoid do you know enough about what the different timings in RAM do to make a video? Gamers Nexus did one part of what different timings do but they kind of stopped after that. Since you overclock want to know which timings you find make a difference.

[@jdogdarkness]

@ActuallyHardcoreOverclocking Can you please send me a link to your Ryzen memory bandwidth/stability/voltages document? I've been looking everywhere for it.

[@benhall7574]

Hey Buildzoid, I have been using ADIA64 to test my ram as I have been tightening the timings, what would you suggest I replace it with to better test real world usecases?

[@DragonForce1393]

Some benchmarks, like geekbench, linpack, ycruncher and intel mlc

[@heni63]

Cool idea with the Ram Making sound haha any way to try or to make that ? I mean on the consumer or your site

[@hiibillymayshere4238]

Wouldn't drops in 0.1/1% lows be indicative of a mini "stutter"?. Considering it's showing the framerate cannot consistently stay at it's average. I notice improving my TRFC and TRC doesn't do much for the average but it increases my 0.1% lows and keeps them more stable. Conversely if I overclock the TRFC too much in particular I notice my .1% and 1% lows drop much more severely in times of demanding moments even in a easy to run game like league of legends, and on Destiny 2 for example, with a 980 EVO, I'll have times where it takes literally 15 seconds for my loading just to START, When I travel to a new area on a 5950x. I'm assuming that would be due to errors and the memory having to resend data that it failed too or current memory being corrupted and having to be resent/recached etc?

[@Muldeeer]

You should always test stability after overclocking, but errors I assume can lead to worse performance.

[@noreng9333]

There's no error correction in play for consumer grade memory. Lower tRFC and higher tREFI will always improve performance, until the refresh cycle gets too low, the data gets lost, and your computer crashes. Too low tRFC will crash very fast, even just 5 ticks too low will usually crash within 3 seconds when adjusting the timing in OS.

[@hiibillymayshere4238]

@@noreng9333 The computer isn't always going to crash immediately, there are varying levels of "corruption" or dysfunction. I once had TRFC and TRC values too low, no crashes (until maybe A day and a half later), but My sound would go out completely at random intervals, games would drop to 40 fps from like 200 periodically. Simple webpages that I just loaded fine on a previous startup took 25 seconds plus etc etc. Unstable memory can cause funky issues

[@ratinabox1065]

We have no good tools to test memory/OC stability. None of the ones I used made my system unstable. But I still end up with random bluescreens when I have more Network I/O going. After months of fiddling with clock timings and voltages of my Ryzen 2700x system I decided to go back to stock clocks and tweak my OS as good as I can. Maybe I simply degraded my CPU by having PBO with aggressive voltage settings in a 24/7 scenario for over 2 years, or my RAM is not B-Die and I was lied to.

[@Wasmachineman]

bz has come out of hibernation!

[@gtijason7853]

I use the time my memory takes to refresh for crying. Yeah I know my memory quite well lol

[@timtoomuch]

I keep hearing having a high tREFI will corrupt my ram is this true running 350 tRC 65000 tREFI on my voltage is 1.42v

[@Artur_Stoll]

So we can fix refreshrate in a game on 120fps for example and there will be no difference of long and short frames at all. Sry for butt Inglish.

[@Teatime4Tom]

Actually Hardcore Math

[@lietome2153]

Stutters in warzone was gone by upgrading from ddr4 3200 cl 16 16 16 38 to 3800 cl 16 16 16 32. 👍👍👍

[@PwadigytheOddity]

All logic goes out the window when you throw in a shitty outdated game engine. Saying "Refresh Cycle did it," vs "Refresh Cycle didn't do it" are both absurd, because stuttering in games can happen for the most obscure reasons. Maybe something in your system isn't running optimally. Maybe something in your system is running *too* optimally. Maybe something in your system is running at some weird window on a sliding scale that very specifically isn't liked by some line of spaghetti code that prefers some number of things happening in some arbitrary timeframe vs. another. Maybe the game (if Horribly designed P2P) is running part of its physics on someone else's system with insufficient cross-checking and *their* system is causing you to stutter. Maybe you have too many cores. Maybe you have too few. Maybe the price of tea in China is too high. Or maybe you perceived stuttering more because you died a lot that day in your multiplayer game.

[@Hwbot-Philip-Park]

Sir I am using z690 intel 12900k 4300cL 14 15 14 24 280 gear1 twr is 8 trefi is 262143 used 1.678v What is this diffrent 65535 131071 262143 my friends said 65535 is good but other is temperature is too hot and dangerous it is it right?

[@EpicBunty]

I think you should run sfc scannow regularly and just go with if. For me too high trefi definitely caused like a single frame stutter every few seconds.

[@Hwbot-Philip-Park]

@@EpicBunty It was a setting a long time ago, but now, after changing the cooling solution to water cooling, whether it is ddr4 or ddr5, all are used normally with 262143 trefi and there do not seem to be any problems ddr4 b660m-e mainboard 4400cl15 and asrock z790i lightning 8600cl36 all 256bit use now thank yousir

[@EpicBunty]

@@Hwbot-Philip-Park nice ! 256 bit of what may I ask?

[@lrmcatspaw1]

Dear buildzoid, I have a question that perhaps you can answer. Have you ever expirienced a system that Stutters with XMP, but does NOT with default ram timings. (not all the time, just from time to time). My 1700 ryzen had this issue with the triendz samsung b die. I now managed to upgrade to a 5900X since amd allows now x370 to use ryzen 5000, and for the most part its gone (its also more stable being able to play voerwatch without crashing). Its not ALWAYS gone. I can even get this stutter with games that I expect to be very easy to run like OSU! Its a new windows 7 install, so I am a bit puzzled about this. (the stutter is gone with default ram settings, non-xmp). Ty for your time.

[@Zeno-]

for one, why are you still on windows 7???

[@modyamro24]

Sounds like IMC instability to me. If your IMC can't run a certain frequency (in your case XMP) reliably it can cause stuttering, crashing, audio crackles and pops, etc. And since the original zen IMC wasn't that great, it's not hard to see that this could happen at even 3200 especially since B die is kinda "heavy" to run for the IMC afaik. The 5900X shouldn't have any problem with that.

[@lrmcatspaw1]

@@Zeno- Because its a gaming machine and I have less issues than on win10, especially for older games. Some wont work at all on win10. I also dont like the automatic updates. They have messed up modded games in a few instances. Especially Elder scrolls mods.

[@lrmcatspaw1]

@@modyamro24 Will look into it.

[@Vegemeister1]

A NEW Windows 7 install? That OS has not gotten security patches for two years, and feature updates have been stopped even longer than that. Your OS is older than your CPU, and it's a wonder that it boots at all.

[@Дмитрий-с3п4ы]

What about 7-zip and WinRar (alt+B) internal benchmark? Yes, I know about SuperPi and ycruncher.

[@Дмитрий-с3п4ы]

@tig dolob а в каком именно бэнчмарке ycruncher это? Под какой цифрой?

[@kamilmichon8061]

what do you think of z690 aqua oc ?

[@F34RALI]

Goat 🐐❤️

[@jeremymatthies726]

Hey Buildzoid, question....would this also apply to vram as well or would that be a whole other issue/factor? Great explaination though i was about 2 seconds behind in comprehension of what you were saying 🙄.

[@kaungkhant1990]

Poor Taylor what've you done :))

[@christopherjackson2157]

Lol. At least he asked about his misconception and got it cleared up.

[@winj3r]

What would be the best benchmark for memory, that showed what real world performance is?

[@CobraProductionsTV]

Depends on what your real world use-case is. If you game a lot I'd recommend using the built-in SOTTR benchmark. That game scales well with faster memory(On Ryzen atleast) For memory throughput I'd use the 7zip bench. --This is merely my suggestions and is in no way a definitive answer--

[@gtijason7853]

Y-cruncher for bandwidth / SuperPi 32M for latency.

[@lolnjeoglondajmejejplejlis3365]

there is a tool you can use to get rid of malicious comment bots of your channel

[@emmata98]

12:00 so the developers should make the map-parts way smaller?

[@HanCurunyr]

Take a look at both Fortnite and Warzone, for instance, both have gigantic open maps and no loading from traveling to and from any point of the map, but, fortnite does an amazing job in managing asset loading to minimize frame drop and Warzone does a terrible job at it, and it stutters and starts pulling 3GB/s from a NVME constantly.

[@emmata98]

@@HanCurunyr Fortnite has cartoon graphics... Also we are speaking of loading in the background...

[@HanCurunyr]

@@emmata98 You can change fortnite for Pubg or Super People, or AC Valhalla, AC Odyssey, any giant game with a giant map, Warzone's code for map loading/streaming is shit, one of the worse in the market, every other game, even elden ring, does it way better, the map's size is not the constraint, is the way the dev makes the loading

[@emmata98]

@@HanCurunyr like I wrote...

[@fracturedlife1393]

Dumb normie now seeing the link between TRFC and TREFI (not OCed intel). This was good.

[@xxovereyexx5019]

waytoodank

[@TrueRegulators]

so did you just explain for 21 minutes how metric prefixes work?

[@xlinnaeus]

very interesting for someone who's never heard about metric prefixes before

[@Airwave2k2]

TBQH milli, micro and nano is a thing the average person doesn't compute.

[@ActuallyHardcoreOverclocking]

There's a reason I moan about YT comments and this is pretty much it.

[@Wasmachineman]

​@@ActuallyHardcoreOverclocking never change bz.

[@TwiztedJugallo]

What he said

[@tinymud3324]

It's So Small it's i-relevant - LOL

[@nwk6258]

4:50 *laughs in csgo source engine* wanna bet? Ur right about the tighter timings at lower frequencies for ram not being optimal. Enemies shoot at me so fucking fast at low freq tightest timings possible, standard OC on ryzen 3800x locked at 4.4. It’s smooth as shit but physically impossible to kill people in every situation.

[@nwk6258]

Csgo is more ram dependent than any other game out there in modern times. *

[@Zeno-]

...I think you just have to get better :P

[@killerrf]

@@Zeno- “ 8 hrs a day bruh!”

[@nwk6258]

@@Zeno- buddy, I’m in main and made playoffs twice, I know how to shoot at enemies. I’m telling you certain timings let you kill people easier, some make it way harder. We all play a different game.

[@zazabean23]

you're using a CPU that has ram latency issues though. What you're saying can be proven with basic math & understanding of human perception. To solve your issue sell AMD cpu buy a intel gen that has good c2c & ram latency & can reach high clock speeds, buy and tweak your ram into oblivion, get amd gpu, tweak and crank to oblivion, get a good display, & if you can get a fiber line otherwise you're learning the game wrong.

[@crochebotamrelapute8816]

Thanks for this interesting video. By the way i like your video, enjoy our 666 likes. Huhuhu :D

[@fandomkiller]

now i see why u were pissed off the other day, fuck the comments bz, dont read them lol

[@Mom19]

Does this mean my 4090 ti 8000$ Room heater will have 1 frame more if I buy more expensive and faster refreshing memory? Jk

[@Metalhead-4life]

Anybody had an issue running their bdie kit TREFI at max 65535? (using a trident z 3600 c16 4 dimm kit)

[@ms.stability]

so my hyperx 3600 CL17 with tRFC 630 r not that bad;) like i thought thx