You can't even tighten the timings as they are already pushed to the max performance out of the box. Good luck trusting your system and stability to memory thats been beyond cooked and comes pre nuked from the factory.kek
>>109286071Usecase for overclock your ram?
>>109286071Goes to show that just because you steal the tech from Samsung as usual, lacking the fundamentals of the technology because you have no R&D in it explains why they can never master the tech they steal.
>Throughout 2026, we've seen motherboard manufacturers verify CXMT's DDR5 with official BIOS optimizations that allow it to run beyond 8,000 MT/s at this point. OEMs such as Dell and HP are using CXMT RAM in their region-bound systems, and even proper PC hardware companies like Corsair are using CXMT modules. Lexar, Kingbank, Netac, Asgard, Gloaway and more are also producing retail DDR5 kits with CXMT chipstl;dr - CXMT DDR5 runs exactly as advertised. It can run at 8,000 MT/s, just like it says. It can be overclocked a little, up to 8,600 MT/s, but it does not scale, and can't be overclocked with higher voltage.Literally non-news.
>>109286088And this is also why China can't break into Apple's flagship iPhone screens because they stole OLED and folding screen tech from Samsung as well with zero understanding of the fundamental ways these technologies work and how they are supposed to be made to maintain quality.Apple yet again rejected BoE made OLED screens and even other Chinese phone companies still use Samsung folding screens for their flagship phones because Chinese made ones fail in quality, both initial and long term reliability.
>>109286088I recently just watched a video about how they tried the exact same thing with CNC machines and both national champions shat the bed
>>109286071They literally cost the same as cartel's memory
>>109286122You're omitting that these are cherry picked samples and not representative of the actual product that ships to end users.
>>109286088Chinks are dumb. Bosnia should be creating new chips
>>109286149Link?
>>109286071I litterally don't care we need more ram.
>>109286174>Buys chang ram today, system starts crashing due to instability 6 months to a year later.Hope they have a long warranty period. kek
>>109286088Cope from Mutts is something else
>>109286071It's just the beginning
>>109286202They won't be able to make DDR6 as that will require equipment changs don't have and I doubt they will be able to steal that tech from Samsung since Samsung is finally starting to crack down on IP leakage.
>>109286155>>109286122Not to mention no one knows how stable these specially binned chips ran over time. Overclocking to reach records uses techniques that normal people can't do.Bottom line is that the product that actually ships to consumers is shit. End of story.
as long as it meets spec who cares.>the absolute fucking state of XMP/EXPO overclockingi'm literally having to override my SoC voltage at 1.25 because the msi motherboard wants to boost it to 1.31 which somehow prevents the system from actually rebooting normally. just am5 things.
>>109286155My non-Chinese memory from one of the big three can't even do 6000 MT/s without absurdly high voltages. Every module that runs 8000 MT/s is cherry picked.
>>109286240EXPO is fucking trash on AM5. When I enable it, it dials in crazy high voltages and the system isn't even stable.
I'd rather have 32GB of base spec DDR5 then 16GB of XxX Bad Dragon Anal Dominator Tactical Gayming RAM
>>109286071shalom goyim, buying this rams is a highly anti semitic act and you should be put in jail for it
>>109286252Its not that it can't do 8000, it can't even do its rated speed consistently which means their process is shoddy and yields shitty so they push through crap that would never make it out of the fab from quality DRAM companies, into retail products.There is a reason why Apple will only use CXMT dram for phones made and sold in china for a reason and even then they haven't even committed to that decision as they need to test it further. Selling low quality parts would ruin their brand in the west.
>>109286088People will defend corporate slavery, a business owning its employees after they quit and effectively turning them into a poison to sue other businesses who choose to employe them, just to dunk on China.
>it only has 95% performance at 10% the cost that could break the DRAM cartel>hahaha what a failurethis is some major cope
>>109286125Apple stole every single thing they’ve ever made
>>109286424I'd agree but you don't mention that it can't even maintain its rated speed because its unstable. The quality of the chips aren't good. I personally want cheap DRAM but I'm not going to trust my system to something that can't even do its advertised speeds without issues. Cheap is better and desired but more than that, I want a stable system and would be willing to pay for it. But I'm going to hold off anyway cause prices are stupid for DRAM and storage right now.
>>109286210DDR6 wont be exists until 2035
>>109286450Try next year.
>>109286349iPhone is a rebranded Samsung phone.
>>109286471100% gonna be delayed indefinitely
>>109286182My brother in christ at least they'd have a system to run.
>>109286071>isn't as performant>BAD Performance>Horrible Performancejesus christ anon chill out I will post in your thread even without the clickbaity claims
>>109286071Fine with me.Frees up the better DRAM for use in sticks. These things can go into systems where the user will never overclock the RAM, embedded system, systems that control the RAM much more strictly.And all the while, CXMT will be improving their production quality.
https://www.tomshardware.com/tech-industry/samsung-engineer-sentenced-to-7-years-in-prison-for-selling-chipmaking-trade-secrets-to-chinese-chipmaker-ex-employee-supplied-10nm-dram-data-to-cxmt-for-usd2-millionHE BUYED BOUGHTED CHINA COMMIE MEMORY BUILT ON STOLEN TRADE SECRETS
>>109286071>horrible performanceIt runs at JEDEC verified levels, that's good enough overclocktard.
Despite what your favorite youtuber says, none of this Chinese RAM stuff matters one bit, because they will not sell under the market price and there's basically infinite demand for memory for years to come.
>>109286071Guaranteed these audiophiles fail at an A/B test.Muh timings, fuck off. That hasn't been a worthwhile thing for two decades.
Who is SK hynix and how did him die?
>>109286171https://www.youtube.com/watch?v=HUBB8DoiB80
>>109286071>buy cxmt ram>insert ram into computer>computer worksI don't see any problems here.
I don't care about performance, i care about that the price is current ram price -10 bucks.Nobody is gonna buy this garbage that is using amd pricing strategy.
>>109286071CXMT is basically worse Samsung ram and Samsung ram was already bad for DDR5.
>>109286155>these are cherry picked samplesYou obviously have no clue how hardware companies differentiate their products if you think being cherrypicked is something of note here. Literally ALL "performance" RAM modules and CPUs are cherrypicked.>>109286793It's either that or giving more money to the cartels.
>>109286686>>109287039>buy DDR5 6000 dimms expecting to run them at 6000...>ends up running below stated specs or system crashesRead the article and you'll knte that its hit and miss if they hit their reported speeds. Thats embarrassing.>>109287676>Literally ALL "performance" RAM modules and CPUs are cherrypickedNo, they are binned. Plus, the actual dram you can buy from cxmt can't hit those speeds which goes to show its shit quality. seethe more zhang. next time try innovating yourselves instead of stealing.
>>109286125why did Scamsung have the shitty green lines for many of the screens they supplied to phone manufacturers then?You'd think BOE would be the one with those issues
>>109286071RAM is so heavily binned that overclocking as not been a real option for a very long time. If the chip could reach higher speed than marked, the corsair jews would put it in a higher bin and charge twice as much for it.
>>109286443no one cares what ram you use in your system, faggot. Go buy overpriced ram from the big 3.
>>109287850They didn't, those were screens made in china.
>>109287884>triggeredHuh. You sure seem to care.
>>109286071So the summary of this article is that CXMT chip isn't as performant as SK Hynix chip.But is it better than Micron chip? I remember early batch of NVIDIA Turing GPU could spontaneously commit sepukku because the Micron's VRAM chips are defective.
>>109286088Technological progress only benefits the upper class. The faster China can steal it the faster you can get some before they inflate all the benefits away.
>>109287896Kek. Here comes the cope. Many of the OLED screens were made by Scamsung during the late 2010's and early 2020's for many phone manufacturers.And many of them wound up with the notorious green and pink line issues. Only revisionist shills like yourself tries to pretend it never happened
>>109287896All Scamsung OLED panels have a QR code behind them that links to Samsung OLED patent page.Even chinkshit garbage have that.
>>109287936I think the point you're missing is that Samsung made many of those "rigid amoled' screens in China for Chinese phones. They are selling that plant as a matter of fact.https://www.oled-info.com/samsung-display-looks-sell-its-tianjin-amoled-module-factory-demand-rigid-oleds
>>109287850>they supplied to phone manufacturersincluding themselves?>>109261022>>109260936>>109260341>>109259111>>109257469
>>109287915Hynix > Samsung > Micronthis also applies to GDDR6. AMD GPUs out of the box perform faster with Hynix than Samsung.
>>109286071The timings are meh, but they do clock fairly high 8000+ at moderate voltages, samsung/micron can't do that reliably.
>>109287971Those issues were present in many of Scamsung's own phones, not just Chinese phones. Even their flagship Galaxy phones people pay out the ass for. How come BOE didn't have anywhere near as many of those issues?In fact, many phone brands switched over to BOE after Scamsung's massive fuckup
>>109286071>first generation manufacturing batch can't compete with 20+ years of experience yetHold the fucking presses and forget about Spiderman for a minute!
>>109287998HDDR6, Samsung is slightly faster and runs cooler.
>>109286433Well then if they do, then China should steal from them the knowledge on how to steal, since they are clearly better at it.
>>109288012Samsung cannot enable fast timings in the driver or else they get memory instability. Hynix can.
>>109286071>Chinese product is worsewow shocking
>>109288006Every screen developed problems at one point or another. No manufacturer doesn't have issues.Even Chinese phones using Chinese screens have green line issues. https://community.oneplus.com/thread/1644002770442780680That said BoE has massive QA issues now that LPTO screens are the latest and greatest screens but vastly more demanding to produce. This is why Apple went with Samsung, their bitter rival, as Samsung (Korea) has the tools that only Korean companies can buy.
>>109286083>he runs above JDECngmi
>>109286071It will get better over time. People usually forget how crap chink anything was 20 years ago and now I only buy their cellphones and other electronic stuff. Even their cars are getting much better.
>>109288111>No manufacturer doesn't have issues.That's because as anyone with two brain cells to rub together figured out long ago, OLED always was and still is a dead end technology and the fact that manufacturers got away with convincing retards that it's superior has set display tech back by 2-3 decades.
>>109288175Nearly all enthusiasts run EXPO/XMP which is above JEDC.
I don't understand how to manually overclock RAM, can ChatGPT help?
>>109288193You need a motherboard/BIOS that supports overclocking.
>>109288193yes but gemin
>>109286349>Apple will only CXMTneat, CXMT delivers top quality then. they only do certain markets due to trump being retarded, hope you realize that
>>109288193Read an overclocking guide then copy someone elses timings once you figure out which ics you have
>>109286071>can't tighten the timings because it's already in the 8000MT/skek is that a bad thing?
>>109286071>inconsistentso your saying there's a chance
>>109287820>buy DDR5 6000 dimms expecting to run them at 6000...>ends up running below stated specs or system crashesThis happens with memory from the big guys from Korea and USA too. DDR5 6000 is an EXPO/XMP overclock. JEDEC spec is 5600. There is literally no guarantees that DDR5 6000 will work with your specific CPU/motherboard. If you got a bad bin for CPU or there's some reason the MC doesn't work well with your DIMMs you're gonna see instability. Literally seen it happen multiple times and I've never even seen CXMT memory.
>>109287820>buy DDR5 6000 dimms expecting to run them at 6000...>ends up running below stated specs or system crashesWelcome to DDR5. I've built three different DDR5 based systems so far because I'm actively avoiding dealing with this generational shitshow (two AMD one Intel), all three with different memory and motherboard manufacturers.All three can not run above JDEC or even JDEC with tightened timings when using 4 DIMMs despite being on the mobo's QVL and being advertised with much better than JDEC speeds/timings.So to make a long story short there are no 4 DIMM consumer DDR5 boards period. You might get lucky, great for you, when not even QVL kits consistently manage to POST with their advertised and """"""""""""""""""""""""""""tested"""""""""""""""""""""""""""" performance what the fuck do you expect from ChingChongJEIUFHUIAEghuzgFAIHWI LLC in Not A Paperstreet St. Shenzhen?
>>109286443>it can't even maintain its rated speed because its unstable. The quality of the chips aren't good.[citation required]that's not what the article nor the source said
>>109288446Sincere question.Is it not worth it to build a PC unless you specifically can use overclocked memory?I always avoid XMP/EXPO because nothing else in the world uses it so I feel no reason to and also have no real strong feelings about the problems with each platform. Always seems weird when people freak out so much about memory speeds
>>109288479They're advertised for those speeds though
>>109288479Memory speeds/timings don't matter if you have an expanded L3 cache part like x3d or the upcoming novalake bLLC
>>109288479>product A claims to do X>product A claims it works with product B>product B claims it works with product A>product can not actually do X or work with product BWhy does that matter? Because that is called committing wire-fraud amongst other things.And yes timings and speed matters.>but muh bing bing wahoo benchmarksSuch a shame the only thing you can do with a PC is play this weeks gacha shit.
>>109288479Not him but I've never used EXPO or XMP. I always do manual timings and clock for my memory. Currently doing DDR5 5600 CL30. 1.1V (or 1.15V, I forgot). EXPO/XMP is really bad if you want stability. I did test EXPO profile with my memory, DDR5 6000 with that profile, but I had a game crash after playing for like 6 hours. Completely unacceptable.
>>109288479Overclocked memory is a tiny performance increase for a fat stack of reliability problems. You build a PC to upgrade what you want when you want without pathologically rebuying stuff.>>109288533It's not fraud retard. Motherboard vendors make very narrow promises wrt memory compatibility. They never said this random new chinkshit module would work with their product. The memory vendor might, but it's on them to make good not CXMT.
>>109288632>It's not fraud retard. Motherboard vendors make very narrow promises wrt memory compatibility. They never said this random new chinkshit module would work with their product. The memory vendor might, but it's on them to make good not CXMT.even big three kits on their compat lists sometimes don't work in expo
>>109286071terrible article. they simply made a single paragraph tweet into a full article full of speculation and misleading claims
>>109288500>>109288533IDK if your not getting my question or dodging itI personally don't believe in the concept of running OCed memory because I don't believe it is fit for production regardless if a backwater company advertises it to be so.My question is the existence of companies selling overclocked memory or your ability to OC one of the primary reasons to be interested in building a PC. If you where forced to run JEDEC would you still build and use a PC?Again I ask because people write walls of text about how bad these platforms as if they are almost intenible. I'm not saying there aren't problems but it seems to be self inflicted.
>>109287820>they're not cherrypicked, they're cherrypickednot an argument
>>109288644They usually list EXPO configs as OC (not guaranteed). Memory compatibility is insanely kiked. If ECC were common nobody would be buying any of this shit because it doesn't work a lot.
>>109288719>Intel® XMP 3.0 certified>AMD EXPO™ certified
>>109288479I can't tell the difference between default 5600 Mhz and 6000 Mhz EXPO.Hell, when I bought my computer, the DDR5 JEDEC standard was still 5200Mhz
>>109288738That's not the motherboard mfg. The full chain has to be certified. This is why servers have special modules that cost 3-4x more / GB.
>>109288719>>109288738People really need to understand that XMP/EXPO memory is completely non-standard. It's something the DIY industry came up with no real agreement in reliability and compatibility standards.Even the so called "certification" is literally just self-certification to make sure the profile can load. It's meaningless for actual reliability/compatibility.I wish I understood why people but so much faith in the DIY memory companies in the first place. I understood what companies have been really selling since DDR3 and I never took a real consideration it was for anything serious beyond extracting your money. Companies like Gskill, Corsair, etc are selling a promise and as far as I'm concerned they have little to zero relation to the actual memory fabricators, so why did people ever believe in the first place I will never understand.
>>109288944JEDEC literally exists so server manufacturers don't have to dig to far to get compatible memory.Modules are more expensive mostly because they are capital goods and have a higher standard of reliability with ECC and buffering.
>>109288632>doesn't know what a QVL isSad. Many such cases.>>109288659>I personally don't believe in the concept of adhering to gravityOkay?
>>109288982>I personally don't believe in the concept of adhering to gravityOkay?So your PC doesn't run if your memory isn't overclocked?What the fuck are you trying to say?
>>109288978>because they are capital goods and have a higher standard of reliabilityWow how do you think they go about establishing that? Could it be they go to specific board mfgs and have their modules certified? Asking rhetorically because that's exactly what they do.>>109288982>doesn't know what a QVL isNot everything listed on a QVL sheet is guaranteed.
>>109289025What are you trying to say nigger?>hurr I don't believe in overclockingAgain: Okay?How is that relevant to anything said so far?
>>109288533Good luck bro. Memory manufacturers aren't hiding it.
>>109288446(apologies for boomer screen photo, the machine currently incidentally airgapped)I've been tuning 4 (unmatched) 32GB (single rank) sticks of Samsung M-die on a 9950X3D and managed to get it running at 6400 MT/s, but it took almost a month of tweaking to get it totally stable. Now it'll survive 24h+ of Karhu, AIDA64 etc with no problems. This IC is pretty dogshit (those timings are extremely tightened down, despite appearances) but I still get pretty decent performance out of it just by running it so fast. I have no fucking clue if 1.4VDD is even safe on this memory, EXPO is 1.3VDD and the highest I've seen for any other Samsung M-die kit is like 1.35V. It absolutely needs it though, 1.35VDD won't even make it into Windows.Getting this memory stable was by far the hardest time I've ever had tuning PC hardware.I also had a matched kit of 4x32GB (dual rank) SK Hynix A-die, and that worked okay at the advertised EXPO 5600 MT/s speeds, but no amount of tweaking voltages/drive strengths/RTTs ever got it even vaguely stable at 5800 MT/s or above. Gave up on it after a week (then sold it for 4x what I paid hah).Really, AMD boards would be better off if the majority of AM5 boards only had 2 DIMM slots. The average user shouldn't even be tempted to try running 4 sticks, and 2-DIMM boards have significantly better signal quality (and therefore top speeds) just by avoiding having hundreds of extra unterminated traces. It's criminal how rare 2-DIMM ATX boards are.
>>109287998I still remember when during the RTX 30-series period, 3060 Tis started turning up with SK Hynix memory instead of the usual Samsung. Crypto miners especially really liked it because it got a significantly better hash rate than Samsung memory, but then that generation of graphics cards started dropping like flies after a few months because it turned out that SK Hynix GDDR6 had awful durability.
>>109289048>Could it be they go to specific board mfgs and have their modules certified?Not reallySome brands have tested/branded modules but under no circumstance is it the expectation that memory just doesn't work or be unreliable just because you didn't use specific modules.Again, that's what JEDEC is for. If your a Google, Dell, HP, etc and have a system validated for 5600Mhz DDR5 RDIMMs you can reliably source DIMMs from basically any manufacturer who sells DIMMs under the same standard and have an expectation they will work just fine as everything electrically is the same.This is a vast difference compared to 6000Mhz CL30 DDR5 XMP/EXPO in which each DIMM manufacturer runs the DIMMs differently with different timings and voltages. This is impossible to validate against because how the DIMMs function electrically is subject to the DIMM vendors whims
>>109288479JEDEC is fine for writing emails and editing simple spreadsheets>>109288632>Overclocked memory is a tiny performance increase for a fat stack of reliability problemswhere do you people come up with this shit? have you tried it yourself? (rhetorical question)
>>109289588>where do you people come up with this shit?Maybe because of all the complaints
>>109289588JEDEC speeds for DDR5 are higher than the OCd version on the bottom for that DDR4. Worse timings, but DDR5 has 2 independent 32bit channels per module for a total of 4 which reduces latency a lot.
>>109289621People who are obessive about RAM don't operate anything rational.
>>109289614complaints from who? redditors?>>109289621CPUs don't need that much bandwidthbut you're free to try it yourself and see what happens
>>109289588>where do you people come up with this shit? have you tried it yourself?I unironically have and I also memtest / stress test everything before using. Which is how I know the tiny performance difference isn't worth the trouble.
>>109289642show me before and after screenshotspost your memory timings and memtest results
>>109289633The fuck are you saying? The literal statement of the post I'm replying to is that JEDEC is only adequate for spreadsheets while at the same time posting a benchmark that shows how much higher bandwidth and lower latency increases framerate beyond 200 FPS at a cinematic 720p. JEDEC profile for DDR5 5600 is better than the benchmarked DDR4 that's also massively OCd. So if it's only suitable for spreadsheets, then the OCd DDR4 is not even good enough for that. Forget playing games with DDR4-4300, it's way too low performance.You're lucky if you can edit a word document.
>>109289673>JEDEC profile for DDR5 5600 is better than the benchmarked DDR4lmaoonce again, just try it yourself, seeing is believing
>>109289679What the fuck do you want me to try, exactly? Stop vague posting.
>>109289689your own systemshould be obviousyou can pirate the game, or ask me to try some other workload, go ahead
>>109289698... You want me to benchmark Cyberpunk at potato settings? You do know it's not gonna tell you anything useful, right? I'll do it though, why the fuck not.
>>109289730it won't tell you anything useful if you don't care about CPU performance in gameswhich you clearly don't because you think JEDEC is finebut i'm glad we're at least having a productive discussion instead of people arguing about abstract hypotheticals
>>109289652Bro I haven't done that in 2 years. Move on.
>>109289750told ya
>>109289747>it won't tell you anything useful if you don't care about CPU performance in gamesNTA but if you have to benchmark at such low settings then what's the point of you're not playing that those settings?Don't tell me it to give the "hypothetical" performance for other games or future games if you don't want to argue hypotheticals.
>>109289781>Don't tell me it to give the "hypothetical" performance for other games or future games if you don't want to argue hypotheticalsi mean i can benchmark other games that are actually played at those settings as wellbut it does, game engines generally scale similarly across the board (as evidenced by the fact that X3D is such a huge boost)it's just that some games (like wow) can't easily be benchmarked so you have to use other games as proxies
>>109288632>Overclocked memory is a tiny performance increase for a fat stack of reliability problems.If you don't know how to OC then perhaps, in actuality the downside is that it it takes way too long to narrow down the exact tightest timings you can run fine.
>>109289821pro tip: you don't actually have to run the tightest possible stable timings to get a huge performance boostif you're on AM5 you can use buildzoid timings which takes 5 seconds and gets you 80% there with 10% of the effort
>>109287882>RAM is so heavily binned that overclocking as not been a real option for a very long timeYou're either a clueless fag or a chink shill running damage control. Hynix M and A dies both scale with voltage and overclock extremely well. There's tons of unrealized potential in them under ANY XMP/EXPO profile currently on the market. That's part of the reason why AMD have introduced new EXPO Low Latency profiles that tighten up a bunch of timings out of the box, though even those can tighten a lot further.
>>109287925Somehow me physically getting a faster computer "only benefits the upper class"
>>109286471>DDR6Gee can't wait to buy a single 8GB stick for $2000 plus taxes each
>>109286151The good chips of the cartel are basically sold out.
>>109289747Here is what my office PC does. I did close the spreadsheet application during the test, but unfortunately the CPU was pegged at its thermal limit the entire run. I could raise the thermal limit by 20C or increase fan speeds by about 50% and it'd do better, but I feel like we are dealing with absurdities here.
>>109290058why crop out the CPU? if it's a 9800X3D that's not particularly impressivefor 14th gen or non-X3D ryzen it's goodhere is what a gigatuned X3D does: https://youtu.be/_3GFm-QN4Ag?t=113and gigatuned 285K: https://youtu.be/MmflN2KJIZc?t=106you will never reach those numbers because that dude has an industrial chiller and bins 100s of chipsbut you can get close
>>109289781It's a benchmark like 3dmark, furmark, unigine, cinebench. no one actually plays cyberjunk lmao
>>109290151The CPU is a 9700X. The screenshot in this >>109289588 doesn't have the CPU shown so I didn't think you'd care.
>>109290254>The CPU is a 9700Xcoolnow apply this and see what happens: https://www.patreon.com/buildzoid/posts/low-effort-rank-77403831>doesn't have the CPU shown so I didn't think you'd careit does, clearly says it's a 14700K
>>109286071i mean if they shit out 3200MHZ DDR4 ram i'll be a happy man>b-but muh overclocking!!! wahhh wahhh!!!Eh, 3200MHZ just works, already got 32 giggies and i'll stay on AM4 platform untill like the 30's/40's atp
>>109286088Good. The fuck Jyung gonna do
>>109290312Yeah it does, but after spending time downloading an 80GB game and only remembering the crop factor, I didn't remember the chicken scribble on top of the cropped screenshot.My computer would be unstable if I set it to those settings. Not unbootable, but some application crashes maybe once a week or once a month. 6000MHz is 100% not possible with 0 crashes. Also, I don't think anything would happen, because even if I set the memory to 6000MHz with those settings, the CPU would still be at maximum temperature. If I wanted to get more performance, I'd first have to at least make sure the CPU isn't temperature capped - or if it is then it's at least close to TJMax, which it isn't. I could probably get better results by just running the test in winter.
>>109290416....is your pump speed set properly
>>109290536I have air cooling with 1100 RPM max fans not CLC, and my temperature limit is also set to 75C.
>>109289915What can you do with a physically faster computer that you couldn't do 20 years ago?>but efficiencyinflation
>>109290416>the CPU would still be at maximum temperatureno idea why you're throttling in cyberpunk but it's probably not throttling down to base clockif you're at 5-5.3 GHz right now better cooling isn't gonna make a big differencehonestly just seems like your system is kind of a clusterfuck and that's why you're defending bad memory settingsbut if you're happy i'm happy
>>109290686encode videos? Move files around?You are confusing CPU with computer.
>>109286071Give it a generation or a few. Cheap shitty DDR5 is still better than no DDR5.
>>109290803My system was built with 3 goals - low power consumption, silent operation (if I hear it, it's garbage) and low maintenance (this is also why I use ceramic TIM even though it's less effective). It wasn't throttling down to base clock.The argument: >>109290058When you look at this benchmark of a 9700X running ebin 5600 MT/s DDR5 - do those numbers look so fucking awful that you need to only reserve that computer for spreadsheets, given that it's completely inadequate for gaming or not? That's basically what you were suggesting. In my situation, I have no reason to believe the memory is even the current bottleneck because I have low performance cooling, but even if I was able to get lets say 50 extra frames out of the CPU just by dialing up all the voltages on the memory controller (I run mostly stock voltages or undervolt) - do you really think that it'd actually be a worthwhile difference that would translate to a make or break gaming/non-office task experience in real life? My graphics card can't even run cyberpunk above 100 FPS stable with settings I'd realistically like to play the game at.Even stuff like 7zip benchmarks perfectly fine for normal day to day use. Could it be better? I'm sure. Does it need to be? I don't think so. Your view is way too extreme and one dimensional. OCing memory does increase performance but it also comes with downsides. For one, it takes time to dial in stable settings, higher voltages will also eventually damage the memory controller. I've seen a few people with cooked MCs before. Maybe it'll take like 10 years to do the damage, but I still use my 10 year old X99 MATX system.
>>109290810People were doing that fine in 2006. I don't see your point.
>>109291122They were doing it poorly, I can move all the files from 2006 in a feasible amount of time.
>>109289821That's the point though. None of these niggers itt are putting the time in to do that. They just pick EXPO Certified(tm) timings or copy parameters from a fucking chuuba and assume it works. Then blame something else when their computer starts fucking up. Nobody wants to have their computer in a change, benchmark, test loop for 3 days.
>>109291253and this has benefited you because...
>>109291281Because it has saved me time. The same task takes less time, the larger tasks that were impossible are no possible. Why should excellence need a reason to exist?
>>109286071CXMT is an emerging player in the RAM market, we know that, we weren't expecting it to be state-of-the-art. I mean, we're currently in a RAM shortage with high demand, are there any better ideas during the RAMpocalypse?
>>109291297>excellenceis this a euphemism for societal decay or something?
>>109291374I'm not Russian lmao
>>109286071>hisense>tlcstarted shite now took over Sony's TV section
Samsung memory fabrication posted 19x higher profits then their best year on record.It's no joke when I tell you none of these companies give a shit about who else enters the market.The smart ones are developing the ram tech of the 2030s with some of this extra money.
>>109286071Oh no! How much?
>>109291036>When you look at this benchmark of a 9700X running ebin 5600 MT/s DDR5but is this JEDEC 5600 or are you running the EXPO timings at 5600? even AMD shows a substantial difference between JEDEC and EXPO in their marketing materials>Your view is way too extreme and one dimensionali'm extreme and one dimensional because there's literally no reason not to tune your memory except lazinessalso this is 4chan, you have to call the other guy a clueless retard or your argument holds no weight>it takes time to dial in stable settingsapple sells macbooks that just work out of the box>higher voltages will also eventually damage the memory controllerlol ok i think i discussed this with you beforethere are only like 5 real people left here after the hackanyways i'm sorry you're insecure about your build>>109291272>copy parameters from a fucking chuubatfw no overclocking chuubas
>>109290416>>109290556>>109291036why even complain about your own sustained performance if you deliberately run a downclocked/thermals/noise limited setup. just say that's your design point and everyone gets itanyways the memory controllers & PHY of the zen-type cpu design thrive on optimized latency more than absolute clock speed. the IF link is a limiting factor for parallel bandwidth, it's essentially a cap on the maximum stable RAM speed before requiring worse latency penalties with the ratio divider option. Unlinked IF:RAM speed provides little benefit. So keeping the ram system set for best latency at the expense of 5-8 watts of core output is usually a winning trade, all use cases averaged.you wont miss the 150mhz unless all you do is competitive game
The better CXMT dies are all going to domestic chinese AI / Phone industry. By decree of the party the captive chinese market is more important. You are buying the memory shitbins that didn't qualify for State Owned Enterprises or Hauwei/Xiaomi/Lenovo if you're buying chinese dimms. People like Corsair at least get the mid tier bins.
>>109292314How much? "Here's some cheap garbage that works" prices or "fuck you goyim I have a monopoly" prices? Only one is going to entice me to buy
>>109292069The market prices math on that suggests Samsung only ever had 5-10% profits on memory before this AI shit.But its funny you think enough of that $$$ millions will develop mram or 3 bit logic wells or whatever is going to supersede DDR7.What memory needs is a new design that can continue to scale, or somehow virtually multiplies each sram unit's capacity. This 14nm shit has got to stop.
>>109286071do consumer fucking cares? we want fuck huge, affordable rams
>>109289355Huh, I didn't know anyone even sold 4gb dies in single rank Good job on tightening that at all, must've been hell to validate stability for that capacity. You sure it's stable without error correction happening though? Because 3200mhz uclk/mclk and 2133mhz fclk is pretty aggressive
>>109289355keep your sticks colder than 50c and the refresh interval it can hold while set tight doesn't deviate. ddr5 and newer all force the refresh with tempif you want to get extreme about this solder more and better caps along the path for lowest esd highest capacity, matched inductance. cleans noise from the signal lines.passively the best you can do is keep the circuit cold
>>109292874>I didn't know anyone even sold 4gb dies in single rankI didn't either. P/N is Kingston KF560C36BBE2-32. Only have it because Amazon had it for $150/stick in March and I impulse purchased 4 when it was in stock for like 30 seconds. Shame it's not 2x2R DIMMs instead.>must've been hell to validate stability for that capacityYes.The worst part was that training used to be super inconsistent. I could pass Karhu for 48 hours, reboot the machine, change no settings, and then fail in ~40 seconds. The key to fixing that turned out to be enabling the "DDR5 Robust Training Mode" option in the AMD submenu. Boot times (no MCR) are still only about 30 seconds, surprisingly. But I didn't realize that until later on, because I was doing the sort of>tweak>retestloop, and assumed that the different settings were why stability would fluctuate.The other tricky thing was that Auto VDDG fluctuates from 0.9 to 0.95 between boots, and 0.9 often doesn't even POST.>You sure it's stable without error correction happening though?Nope. But, AIDA64's synthetic bandwidth tests increased by ~3% when I went from 6200 to 6400, likewise -3% when I dropped from 6200 to 6000 as a test once, so silent error correction can't be *too* bad.>Because 3200mhz uclk/mclk and 2133mhz fclk is pretty aggressiveI also extensively validated a 3100 UCLK/MCLK and 2166 FCLK config, that one was easy. Couldn't get 2200 FCLK working. So, 2133 still has a bit of room, and keeps the 2:3 ratio with UCLK that helps latency just a little bit.>>109292933I think the max I've ever seen was 52C on the two innermost RAM sticks when I was running 1.45VDD and sustained TM5 1usmus. Right now it's at about 48-49C, 13 hours into TM5.Karhu and AIDA64 only get it up to about 45C.I have considered installing a RAM fan, but it doesn't *quite* need it.On that note I could probably max out tREFI to 65535, it worked before on 3100 MCLK, it's just a little loose for safety right now.
>>109286071The mainland chinks still lag behind.
>>109286071I literally don't care how bad the performance is as long as it's cheap enough that I can put plenty of it into my system to prevent paging to hard disk.
>>109286071the vast majority of people don't need overclocked ram. so this doesn't matter for like 97% of users.
>>109286083fpbp. The primary consumers: corporations, absolutely will not run this stuff outside of specs.
>>109293328>>109293351The real issue is its inconsistent even at its non-overclocked rated speed. e.g., its unstable thus caveat emptor.They stole the IP but lack the manufacturing expertise and skills to actually make it properly and consistently.
>>109286071>overclockingWho gives a shit. I'd be surprised if even 1% of PCs run overclocked RAM.If it's stable at stock speeds and is available at prices that aren't outrageous, it's fine.>but muh long term supportYour choice is this, or having no RAM whatsoever long term. At least until the AI bubble comes crashing down, or the RAM cartel is broken up.>but muh stolen secretsLike I care about the cartel's well being. It's corrupt from top to bottom, they've all done it before and this is par for the course.
>>109293927>he didn't read the article, just the headlinesits not stable when not overclocked either. Its luck of the draw. If you're lucky, your kit works at slower speeds and timing, if not you're gonna have a unstable system even when not overclocked.ITT: The zhang bots be in full damage control mode. kek
>>109293952>he didn't read the articleI skimmed it. It was a waste of time and said nothing about instability at stock clocks, only that performance was slightly lower.If you want to disprove me, point to the line that says so. Otherwise I'm just gonna assume you're talking out your ass.
>>109293952>he didn't read the article, just the headlinesmaybe if the OP fucking posted the article instead of just the headlineI'm not going to bother googling it.
>>109287925>me getting stuff for cheaper is le bad
>>109293952>its not stable when not overclocked either. Its luck of the draw. If you're lucky, your kit works at slower speeds and timing,So what you're saying is it either werks or it doesn't, just like cartel RAM.
>>109292114I'm running different timings compared to JEDEC. They are worse than EXPO. Running custom timings for the JEDEC profile clocks is very easy to do, because you can run everything at stock voltage and type in almost any numbers that don't look like an overclock compared to whats written on the RAM itself.>insecure about your buildWhat is that even supposed to mean? You're acting like there's 0 risk so I'm being irrational and not doing something that's completely free. Even though it's clearly not completely free. Just a 400MHz increase on memory requires 0.2V increase on almost everything memory related. If you limited yourself to "reduce your memory timings a bit" instead of "OC your memory to some arbitary level that's the bare minimum to run video games", you'd have a lot less pushback and you'd help people a lot more.Just to be clear, I don't think this is necessary either. Running just plain JEDEC profile is fine, you will lose performance but not many people will run 720p low and compare it online. The whole point is that EXPO is dogshit, you need to be skilled at RAM OC in the first place to use EXPO and XMP profiles. Having a PC that crashes or PC that slows down after a few years is a worse experience than having a PC that performs less good than someones overclocked setup unless you know what the symptoms mean and how to fix them.>>109292223Statement of fact is not the same as a complaint. You won't miss it in competitive games either by the way.Sometimes I wonder if I'm the only person who actually plays video games, because the minimum system requirements to play video games in the "tech enthusiast" space are completely incoherent. You really aren't gonna notice a difference between 300 FPS and 500 FPS in league or quake III or whatever. Even if you do have a 480Hz VRR monitor.
>>109286071Apple recently applied for permission from Daddy to use them as a supplier so the promise of gaybux from California will force them to tighten up their manufacturing and QC. Apple doesn't want to get a reputation for unreliability on their overpriced fagshit.
>>109294343No, thats not the correct conclusion. CXMT ram is way worse than the big three as shown.Its okay chang, you'll eventually figure it out once everyone else moves onto DDR7.
>>109286125Retard buys item for x amount of dollars, retard thinks item is supposed to last forever and can't process x dollars = y time of usage
>>109294912Whats that have to do with OLED and China's inability to produce the latest panels?
>>109294918
>>109294935yes, the dude here >>109294912 moved the goal post after getting triggered by people noticing that China's tech industry is just a subsidiary of Samsung.
>>109294944Kek retard folding his girly arms and responding scornfully
>>109294949Zhang-gu, relax. I'm sure one day in the far far future when Xi is gone and your people hopefully free, you'll eventually maybe be innovate and no longer steal technology from superior peoples.kek, who am I kidding... that'll never happen.
>>109294955Sorry you're so upset
>>109294960why would I be upset? my people don't have a reputation of stealing and cheating. kek
>>109295025>>109294960
>>109295039>so angry people everywhere know chinese steal and cheat. so mad that zhang can't talk.I accept your concession.
>>109295049>>109294960
>>109295025>my people>don't have a history of stealing and cheating
>>109295049Reminds me, did you know Chinese also steal Korean battery tech? Chinese cars are known to have issues with batteries bursting into flames so they stole the tech LG made to make EV batteries safe and to also package them into cars more efficiently.China truly is just looking up to Big Brother, Korea.No wonder its common in China to say "wow, that must be Korean" or "You are looking Korean" today when referencing something good.
>>109295025but you're a nigger anon...
>>109286071>I got exactly what I paid for and I can't just magically get double the performance by tinker trannying with bios settings>LE BAD PRODUCTkekobvious memory cartel propaganda
>>109295082Nawh cuh, having to steal and then boast like you're something special is peak N tier behavior.You changs really need to learn to apply yourselves like your two superior neighbors, Korea and Japan.
>>109295177Weak. Sorry you're so upset
>>109295184What? You don't agree that Korea and Japan are your superiors? Why are you always stealing from them then?Try to answer without sounding mad.
>>109295192>>109294960
>>109295192sperging unc, that's rough
>>109295211>>109295232
>>109292314shit binned hynix a-die and samsung b-die still scales well with voltage so it's not a binning thing
>>109293853>inconsistent even at its non-overclocked rated speedyou/tom's made that up lmao. nowhere in the original source >>109288652 did it conflate these two points as a critical flaw of cxmt's memory
>>109286071Samsung and SK Hynix use EUV for their DDR5, CXMT doesn't CXMT's DDR5 is more impressive for that reason
>>109286071based chink'Os, price hikers on dilate watch
I play vidya at 1440p, whats the use case for overclocking ram?
>>109294668the rest of your post is paranoid rambling but yes, i do actually need 8000 FPS in quake>>109295292shit bin B-die can't do 3200 lolDDR5 is more consistent, in general>>109295354what vidya? cinematic playstation ports? then it doesn't matter
beat my score
>>109295325You stupid idiot. No one uses EUV for DDR. Its fucking 10nm and DUV is used.Kek, stfu if you don't know what you're talking about.
>>109295408retardback in 2021 btw why act so confident?
>>109295487>>109295408Logic and memory "nm"is different, Samsung and SK Hynix both use EUV, Micron and CXMT don't but Micron already left the market
>>109295383b-die inherently scales with voltage across most of its primary/secondary/tertiary timings
>>109295383shit bin b die cant do 3600 at standard low timings, good stuff can run 4000+ at 1.5v+but b die ddr4 is so old, and so out of production, the price is outrageous
>>109295280wow this is desperation. sorry you're ai upset
>>109296522Did we hurt your feeling by exposing your blatant IP theft?What good was it when you can't even capitalize on it properly? Your DDR5 is bad. Do better. Maybe hire more Samsung engineers like you did to get DUV working.
>>109288652That's modern tech journalism for ya. They thrive on that shit, especially when it's anything Windows related. Making a mountain out of a molehill so that unwashed masses emotionally click the headline and repost it online is how you make easy money in today's outrage economy.
>>109296548>welel absolute fucking retard. Sorry you're so upset, pussy
>>109296609The only ones who would be mad are wumos changs who can't stand truth of chinese behavior getting exposed. Stop stealing and learn to innovate on your own.
>>109286071>>109288652they're shilling for cartel jews and making shit up from nothing? what a surprise
>>109288111>Scamsung OLED >QualityTell that to all the phone companies they fucked over with their dogshit green line screens, including themselves. BOE is unironically superior to scamsung
>>109296716Why does Apple reject them then? Answer me, chang.
I underclocked my ram to 3600
>>109296752Maybe Apple are retards, the fuck should I know. None of your coping changes the fact that scamsung OLED is a fucking shitshow with dogshit quality that gives phones green lines over the slightest overheating, retarded faggotRefer to >>109287992
>>109286071RAM is like pizza nowadays. It's better to have shitty pizza than no pizza.
>>109286342Do you get paid to lie?
>>109286083Epenis benchmarks, that's it.
>>109286471DDR6 will be CAMM2 only on the desktop. UDIMMs are getting phased out.
>>109287850BOE did. Big problem for the OnePlus 7or 8 and some oppo phones.
>>109286083JEDEC speed DDR5 is barely any better than DDR4 XMP.
>hurr durrr is le bad!>still get mad assblasted for a bad chink productwhat the fuck is happening in this thread
>>109303144you got it wrong. its the assblasted anons (who probably owns big memory stocks) who are making shit up to smear the chink product
>>109302982XMP is overclocking
>>109286083because i want to.
>>109286088>trillions in lossesHow?>all the RAM is sold goyEven with stealing tech they are selling RAM Samsung literally does not have to sell. Sammy can't have lost a single cent.
>>109302982JEDEC is guaranteed stable and working on DDR5. Overclocking on DDR4 isn't.
>>109301419It's hybrid camm/dimm
>>109294133Are you agreeing with him that china rules which is what he said or are you retarded and misinterpreting his post
>>109286071>expecting third worlders who are only good at copy and pasting to be good at anything other than copy and pasting What did you expect out of China? Next thing you'll say is you thought they could do reliable ECC or NAND.
>>1093043863600 cl 18 is guaranteed cl 16 with binned ram, 14 with fancy b die
>>109302852No. OnePlus 7 and 8 were using Samsung OLED display. As well as a bunch of Sony Xperia phones. All of them had the green line problem
lel. seven years after its release, china still can't do ddr5 without our machines."china will overtake us"yeah, alright.
>>109305349Ddr5 came onto the market in September 2021 which isn't even 5 years ago
>>109305356stop lying. it's was being developed since 2018 and mass production started in 2020.https://news.skhynix.com/sk-hynix-launches-worlds-first-ddr5-dram/"Since SK hynix announced the development of World’s First 16 Gigabit (Gb) DDR5 DRAM on November 2018, the Company has provided its major partners including Intel with sample products, and has completed various tests and verification of its functions and compatibility."
>>109286071Thread was paid for by the RAM cartel
https://www.pcmag.com/news/memory-from-chinas-cxmt-cant-quite-compete-with-the-big-players-yetWhile CXMT's chips might make up for some of the missing capacity that gamers and electronics manufacturers are clamoring for in 2026, they don't necessarily measure up in performance. Safedisk found it was possible to reach an 8,600 MT/s overclock on the 6,000 MT/s Kingbank kit, but several design weaknesses emerged. SK Hynix memory was faster on raw performance at the same speeds and timings, although Safedisk didn't release comparison benchmark numbers.Voltage scaling is almost non-existent, so you can't just increase the voltage to improve stability and unlock additional performance. Tuning sub-timings doesn't work particularly well, either, meaning getting anything under the stock Column Address Strobe (CAS) speed is unlikely.
In summary, CXMT == Tofu dreg of DRAM
>>109305388It's a big deal for embedded stuff where a number of products simply stopped being viable over the past six months.
ram speed is almost non important.https://www.pugetsystems.com/labs/articles/7200-vs-5600-mbps-ram-for-core-ultra-200s-plus-processors/it's better to buy slower 32gb than faster 16gbhttps://www.pugetsystems.com/labs/articles/when-does-ram-capacity-impact-performance/
>>109305372Back then it was only used for niche arm servers with 0.00001% marketshare. Servers that matter i.e. epyc and xeon were still using ddr4 back then. Ddr5 only became widely available after alderlake launched
>>109286471>2027 mass adoptionIf they rush it they will begin adoption in Q3-Q4 2027, and it won't be standard until 2029-2030
>>109305414Where the hell do you need DDR5 in embedded stuff?
>>109305446ia
>>109304386>JEDEC is guaranteed stable and working on DDR5lmao
>>109302982>JEDEC speed DDR5 is barely any better than DDR4 XMP.And DDR4 JEDEC is barely faster than overclocked DDR3. and the same is true all the way back to DDR1.
nobody actually runs at JEDEC
>>109306102Literally every PC outside of gayming PCs run JEDEC.
>>109306102The vast majority do. Self-built and boutique gaming PCs are a very tiny minority of the lot. Most serious OEMs don't want to deal with customer complaints and returns from random overclocking fails. They need the PC to just chug along for whatever warranty period they prescribe.
>>109286071i need 2tb ram to run latest llm local in full resolution.i dont care if its somewhat slower than others.
>>109306102dear anon run default voltage see image on screen tier stuff
>>109306102I use JEDEC on my NAS because the extra speed makes no difference, but the 1-2 watts saved is measurable when the machine runs 24/7 all year long.
