Intel 13th gen Core Series RAM/Memory Tuning & Scaling, i5-13600K, i7-13700K & i9-13900K
ฝัง
- เผยแพร่เมื่อ 8 ส.ค. 2024
- Hetzner: hetzner.com/hub/ex-line
Actually Hardcore Overclocking: / @actuallyhardcoreoverc...
@ActuallyHardcoreOverclocking
Support us on Patreon: / hardwareunboxed
Join us on Floatplane: www.floatplane.com/channel/Ha...
Buy relevant products from Amazon, Newegg and others below:
GeForce RTX 3050 - geni.us/fF9YeC
GeForce RTX 3060 - geni.us/MQT2VG
GeForce RTX 3060 Ti - geni.us/yqtTGn3
GeForce RTX 3070 - geni.us/Kfso1
GeForce RTX 3080 - geni.us/7xgj
GeForce RTX 3090 - geni.us/R8gg
Radeon RX 6500 XT - geni.us/dym2r
Radeon RX 6600 - geni.us/cCrY
Radeon RX 6600 XT - geni.us/aPMwG
Radeon RX 6700 XT - geni.us/3b7PJub
Radeon RX 6800 - geni.us/Ps1fpex
Radeon RX 6800 XT - geni.us/yxrJUJm
Radeon RX 6900 XT - geni.us/5baeGU
Video Index:
00:00 - Welcome to Hardware Unboxed
00:25 - Ad Spot
01:15 - A Brief Overview
03:50 - DDR5 Memory Kits Tested
04:49 - Memory Timings Used
06:48 - Marvel’s Spider Man: Remastered
08:26 - The Callisto Protocol
09:00 - A Plague Tale: Requiem
10:06 - Horizon Zero Dawn
10:25 - The Riftbreaker
11:01 - Shadow of the Tomb Raider
11:35 - Watch Dogs Legion
12:10 - 7 Game Average
13:09 - Final Thoughts
Intel 13th gen Core CPUs Memory Scaling & Tuning Benchmark, Core i9-13900K
Disclaimer: Any pricing information shown or mentioned in this video was accurate at the time of video production, and may have since changed
Disclosure: As an Amazon Associate we earn from qualifying purchases. We may also earn a commission on some sales made through other store links
FOLLOW US IN THESE PLACES FOR UPDATES
Twitter - / hardwareunboxed
Facebook - / hardwareunboxed
Instagram - / hardwareunboxed
Outro music by David Vonk/DaJaVo - วิทยาศาสตร์และเทคโนโลยี
Check out Hetzner servers here: hetzner.com/hub/ex-line
Thanks for explaining DDR5-8000/8200. Can't wait new stuff from my favorite gaming channel... ;-)
Could you show how various games "fill up" the x3d cache via Windows Performance Monitor. Example: csgo when runing on Intel CPU takes about 20MB of cache, see Performance Monitor window. With larger cache, do games shift more game data from RAM to the cache? and by how much given the available x3d cache can be as much as 144MB. I guess it is up to the application to take advantage and up to benchmarker to show (please) the x3d cache is filled or not, aka used.
I swear! It’s uncanny how every time I want/need to know more about a certain topic… one of my fave tech content creators makes a video about it! Awesome work guys, thank you. This is just what I needed to know about my soon to be installed 13900k
Hope you're getting some Hynix A-die RAM for it.
@@teemuvesala9575 How do I know whether it’s Hynix A-die or not? Going with G.Skill Trident Z5 RGB 6400 or 7200
@@alpinestars-wm8rv shouldnt be the trident Z5 6000 be enough, in the test if the 6400 was faster it was like 1-2 percent, probl not worth the price, 7200 if u have the money I guess, but I would probl not go for it
It does not matter if Samsung or Hynix, bot used on gskill chips, if u run them just with the xmp. If u manually want to overclook them even further, according to internet it is very likely u get hynix with the 6400 kit of Gskill, but if u do not want to overclook past xmp, samsung chips will do just fine.
@@Lukas_Pailiv yes… food for thought! 👍
Hopefully you are not gaming at 1080p with a 4090. If you are just sell your crap lol.
Love these can't wait for ddr4 vs ddr5 memory scaling
@Toroidal Zeus important to note he is using command rate 2 for the ddr4 and not command rate 1
I’m mexican & love to do food videos while smoking weed on my TH-cam channel, to cure people’s boredom fam
@Toroidal Zeus I use cr1 DR @4000mhz. My CPU doesn't have the greatest imc. But still DR is better than SR as you said.
Yeah but it's with a 4090 at 1080p. lol. SO logic goes out the window.
@@benjy117 4090 is probably best 1080p gpu tbh
"Once again, I've called on an Actual Hardcore Overclocker (Mr. Buildzoid 😉) to help out with this content..."
I chuckled harder than I should have at that channel name tie-in/pun. Well done, Steve. And thank you Buildzoid for your countless hours of work to help this community get the most out of their hardware!
Besides scaling, its a good showcase of how latency can beat out clock speed in memory tuning. I have the GSkill 6000/30 kit and am running the Buildzoid settings with 12900k. Works great.
I have a 12900K with GSkill 6000/36, I am going to have to see what I can manage with that since it's clearly worth trying!
How big a noticeable impact do you think GSkill 6000/CL32 would make over 6000/CL30 with a 4090 for 4k120-240hz gaming and productivity?
@@Hyperfluxe Not sure, but I guess I'll find out soon enough haha. The timings are all different between the two kits so there could be quite a difference
@Hyperflux Follow buildzoids timings, cas latency doesn't matter, it's a crude way of checking the die that's being used but as I said, you can literally just look up what die is in the kit. Make sure it's a hyinix m or a die. If you're playing high refresh esports games, ram performance matters more, I don't see why you can't just get an A die kit at a higher clock if ur spending that much as is. The nain factor for productivity is ram capacity within reasonable levels of speed
RAM speed is only as good as the speed that the CPU memory controller can handle. Anything higher is not doing anything. Anything lower, and you are not getting enough.
The ad for Hetzner is nothing but shots of a Core i9. Oddest ad I've ever seen.
Ahahaha.. i usually skip ads so i didn't even notice. Wonder who made the ad. Such an odd choice.
Something tells me that they like to use the Core i9-13900 for servers. I'm not sure why...
and they think that is what impresses us. Its just a bad ad. Are its uptime and performance KPIs good and the price competitive? they didnt give me any hook to even want to check them out
True, weirdest hetzner ad I’ve ever seen, servers with 13500? 🤔🤔🤔🧐🧐🧐🤔🤔🤔
This is the video I needed as I plan on building a new rig next month! Thank you!
G'day Steve,
I'm really liking these memory relating to gaming videos you are doing with Buildzoid's input because yeah Buildzoid is more into the Science of Max Performance. I guess in Petrol Head terms you love Touring Cars & Buildzoid loves Top Fuel Drag Racing.
I also watch AHOC & what you were saying about not all CPUs being able to be stable or even post with really high MEM Specs Buildzoid has shown with his different intel CPUs as he has some good & terrible ones, he has also shown that Motherboard choice can also limit MEM performance.
For anyone interested Buildzoid's Channel is not just OC Based but also contains a lot of educational content explaining how to test for stability if you do OC & how different parts work like Memory Timings, VRMs & other stuff too.
LOVING these in-depth spec hardware comparison videos! Speaking of memory timings I have a question; I heard that because of the infinity fabric AMD use, that it's a bad idea, even with the latest 7950X3D CPU, to go over 6,000MT/s RAM. But having looked at some of the high-mid range motherboards, I see they list support for higher RAM speeds, like the ASRock X670E Taichi says supports DDR5 6600MT/s, the ROG STRIX X670E-E says supports 6,400MT/s, and the Gigabyte X670E AORUS Master says supports 6,666MT/s.
Other than being more expensive, is there any reason I shouldn't get say a 6,600MT/s 64GB (2x32GB) kit (with the lowest CL I can find) along with an ASRock X670E Taichi motherboard and a 7950X3D CPU?
As a bonus question, which DDR5 64GB (2X32GB) 6,600MT/s RAM kit would you recommend for the 7950X3D setup, if performance is more important than price?
Thanks in advance. :)
7:10 'DDR5-6000 (CL30-38-38-28) Tuned' doesn't match the Buildzoid's manual timings from 4:49 (CL34-44-44-28) . Very interested to know which is correct.
Buildzoid creates excel files to see when a request to memory bank can be sent-reset etc, those timings are important, going off by a single digit can mean you miss optimal timing when memory bank is ready for use. But he has a tendency to ramble, which he openly admits, finding the one video where he actually posts the timings in many of his 1 hour videos, is an epic quest I've yet to complete
Good catch. A pretty big blunder on his part. I would assume that he would let us know the video where Buildzoid blesses us with these timings as I would really prefer getting them there. He made one for AMD but I don’t know about Intel DDR5.
@@nathanwhite617 Buildzoid recently said that he wont be making a video for Intel with DDR5 timmings, since stability seems pretty random. AMD either works or doesnt, Intell will pass a full test of testmem5 just to have a couple of errors over an HCI memtest overnight.
@@ccrodrigues yeah so I saw. That sucks. I wish he would do like a 6800 something that is really safe just to be better than XMP for people that really don’t have any experience.
Thanks for all the testing! In the future, I would love to see a "practical shootout" with something like an i5-13600K, comparing XMP 6000Hz vs tuned 6000MHz, across a wide variety of games. A lot of review content focuses heavily on hardware, but it would be cool to examine which games benefit the most from memory tuning, as gains on one CPU will usually occur on others.
2-10% depending on the game
Something more practical, like testing a 4080@1440p you will barely see the difference. Maybe 5% if you are lucky.
@@dat_21 LOL, don't you know that you'll never be able to reach CPUs full potential if you're doing 1440p? 1080p will show how much potential performance difference you can get in CPU bound scenarios, 1440p not so much.......
@@DragonOfTheMortalKombat That's the point. If you are building optimally you are not going to be CPU bound most of the time.
Now this is a good video. While 5-6% doesn't seem like much the 1% low came up as well giving a smoother experience.
I think there's a mistake at 4:50 with the tuned 6000 primary timings which look to be 30-38-38-28 (not the same as the 7200 kit).
15:09 Thank you for saying this! Stability isn't "mostly stable" or "almost never crashes" it's an absolute, stable. I don't have anything against people testing the limits and living on the edge, but the number of times I've heard things like "It's perfectly stable, two passes of memtest says so. Sure Prime95 crashes but that's just Prime95 that's unstable. I almost never have a game crash, and certainly no more often than anyone else." And if it's not Prime95 then it's some other test or program that crash, and yet they claim the system is stable.
Tweak your memory and OC all you want, but if it ain't 100% stable in every test, game or program then don't claim your system is perfectly stable.
Now is it something you can live with? Perhaps it's truly only Prime95 stress test with certain parameters that fail or some other program you don't care about, then fine. But don't claim it's stable!
A problem with things like this is that when you approach the razors edge it's not always easy to determine when you're actually over the edge. Memtest in all it's glory won't tell you. You can have a system memtest stable but still bluescreen or crash in games or other applications. Prime95 torture test can be very good at exposing this, but again it's not perfect. No single application is. It's up to you the user to determine what is stable enough and how much performance you are willing to give up.
To me a few percent of performance is certainly worth the peace of mind, but then I'm not a competitive player or OC fanatic. The moment my tweaking show the least sign of causing instability I will back off, probably a bit further than most. But then I have a background in working with servers and mission critical systems so I'm kind of fanatical about the stability of my machine.
Well, your first mistake is believing the people about their OCs on the internet. People were exaggerating their results since the beginning, even in early 2000s. And sillicon lottery is certainly the factor too. Just because you have a same CPU/RAM model, doesn't mean it could OC just as well.
These days I'm less fanatical about the performance of a PC, I just run everything at stock. If I need more performance, I just buy better hardware that offers at least a 50% boost. Otherwise it's too much hassle.
@@dat_21 Smart man, especially how hard these chips get pushed at the factory. There isn't an insane delta left to squeeze out. Maybe 10-15% at most with all factors involved. But that's certainly smaller than a cpu upgrade each year.
Thank you for sharing this. I do find the variance between the 6000 and 7200 kits to be of particular interest.
Trust me there is barely any difference. You're better off using DDR4 3200Mhz
@@dazeen9591 lmao
@@dazeen9591 Clueless
@@ThunderingRoar no u
@@dazeen9591 no u
Curious if there was any noticeable difference when GPU bound. I would think not but would like to know for sure. Also would have liked to see what type of effect if any the faster memory has on production workloads.
Would be nice to see priority given to cpu reviews followed by memory timing and then motherboards so we know what to pick when cpus release. A lot of people by now have already bought there cpu ram mobo. Tho i totally get this is a lot of work and i appreciate it and it's sometimes not possible.
Im confused on the graph you show buildzoids timing for 6000 as 34-44-44-28 then when you switch over to the gaming graphs it shows 30-38-38-28. Wondering which one it is.
I'm running ddr4 but it's good to see the timings so people can copy and paste. It took me a long time to tune mine
Just as I was researching the topic I found this video. Thanks again guys for providing such valuable data. Cl30 6000MT/s seems to make the most sense to get, especially around 160€ currently where I live.
Also ppl should know DDR4 tunings are quite different to DDR5. Primary timings with DDR4 are dependent on tight, low as possible timings to get max performance. Not so with DDR5, which is more dependent on secondary timings for max performance.
i would love to see a ring bus oc aswell, lowering the latency even further.
Don’t worry il react to the video on stream tomorrow 😁
you -> 🤡
Good info. I do appreciate the time people put into tuning memory and it's very interesting to see the results, but based on these results I think I'll stick with my DDR5 5600 with stock XMP settings on my PC equipped with a 12th gen CPU.
would be interresting to see a comparison against a high end DDR 4 setup like B-Die 4400 because it has been shown before that lower timings oft produce better framerates than higher clocks
Very few cpus and motherboards can do 4400. That’s the very high end of 4400 gear 1
They are only going to show results that 99% of users can actually achieve. Not too 1% only hardware that can.
@@killerrf If "They are only going to show results that 99% of users can actually achieve." then they wouldn't show 4090 content. There are cpus and motherboards can do 4400 that are less expensive than a RTX 4090.
@@sammiller6631 not even close to the same. Anyone can buy a 4090 and plug it in and it get results like reviewers. No fiddling. 4400 is not plug and play. 4090 will produce 10 fold results than what ram overclocking will without the headache if your stable or not or testing for it which can take literally days
@@killerrf my 13600k and Prime Z690M Plus D4 (190€) can do 4400 just fine
@@ketos8315 well if your gear 1 good for you. Your one person. Not the 99%. What timings and voltages did you fiddle with to make it work? How high is your SA voltage ?
Wow, tuned 6000 looks like it's the way to go
Absolutely. It's a great mix of price-performance.
You'll notice nearly across the board for DDDR5 that 6000-6400 will be the range to go for depending on use case and cpu.
Ty!
I'm running the tridentZ 5 6000 cl30 on my 13900k and its running flawlessly with a mild overclock, zero bugs/crashes...highly recommend it
Do you use the timings provided by Buildzoid? Did you have to raise VCCIO or VCCSA? Which is the exact kit you use?
I wish there was a tuned ddr4 for comparison.
Thanks been waiting for this video............
Steve : Great content and thank you for the benchmarks! A note on stability after tuning the memory. There are specific stresstest software you can use to verify stability. TM5 with anta777 ABSOLUT preset is one for the memory, and Y-cruncher VST to test the IMC.
yup. not practical for reviewing obviously, but i disagree with the conclusion that it's "not worth it" due to potential stability issues - those can be easily overcome if you invest the time in proper testing.
anta777 Extreme / Absolute are not good for DDR5. Use Kahru memtest, and Ycruncher with VST/N32/N64.
@@gk-if3qt I wouldn't call days / weeks of stress testing "easy". It's really annoying to do and often it takes hours of testing in multiple programs for one lousy mistake to pop up.
So buildzoid'z memory tuning basically give more performance boost than prev-to-next intel cpu generation swap. Adourable. Need same test with AM5 tho
There is a video on HWU with AM5 memory tuning with a tuned DDR5 6000 memory kit from Buildzoid, you'll need to go back a month or two on their videos to find it. I used the setting on my own ram kit and it works a treat.
AM5 is dead.
They starved the IF bandwidth compared to EPYC SKUs
They did do this same thing with Zen 4 CPUs.
Any benchmarks to show differences at 4k?
The DDR5-6000 CL30 seems be a popular choice for a lot of people right now on both platforms, especially when you get a kit on the Hynix die that can be easily overclocked.
There's no way you get other memory chips than Hynix M-Die with 6000 CL30 kits. But also overclocking is almost non existent in these kits. Right now we have M-Die for lower speeds but tight timings, or A-Die for highest speeds on DDR5.
@@JorgeAlarconL This is not entirely accurate. I'm currently running a 64GB 6000 CL30 dual rank kit @ 6400 30-39-39-28 @ 53ns on a 12700K + Z690 + 3070OC ITX setup pushing HDR10 4K120 to a C1Oled. Going from 6000 XMP to 6400 tuned with very tight timings on a dual rank kit is a world of difference in user interface in regard to smoothness/slickness/snappiness. While the gain in framerate is relatively low, the actual gameplay experience is vastly superior while also maintaining the ability to effortlessly multitask in the background without disruption/distortion. Games in ref: FH5, MS:FlightSim, CS:Source, and settings are Ultra/Extreme or completely maxed out.
@@jakehutchens you’re not hitting 4K 120 with that card, sorry
@@frostybe3r maybe in minecraft lol
@@frostybe3r s/he might in two of those games listed, but not MSFS
would be nice to see some of the sub timing its hard to see where those gains are coming from
@Hardware Unboxed - when you do the Intel/AMD heads up can you throw a couple older titles in their as i seem to remember Intel losing support for some older games starting this generation? thanks!
I wonder where my DDR4 4000 @ C18 would fall in these charts? Lower clocks but much lower latency, I would expect it to outperform the 4800 and potentially the 5200 kit too? (I am running i9 13900ks / 128GB DDR4@C18 / RTX4090 / Aorus Elite Z690 DDR4)
Should be able to get that down to c16 at lease. I did 4133c16-16-16-32.
Cool video
I've had XMP/XOC memory be unstable at their "stock" overclock, so I think I agree with the conclusion, stability is king.
From what I've seen the higher speed jedec like 5600 cl46 is actually really good on ddr5
Yup even if it appears stable, sometimes you'll still experience stutters and whatnot if it's close to the point of instability
I was messing with my corsair ddr4 3600 14 16 16 36 with my 5800x the other day. I was trying to get a tuning baseline and for some reason I ran the occt memory test with the XMP profile enabled and nothing else touched and there were a some errors. With the timings tuned and voltage increased it was doing 3733 14 14 14 28 with no errors. It had been running in my 13600kf system at 3800 14 14 14 28 and passing all memory tests but I never thought to even try testing it at the stock XMP settings. So yeah, in the words of buildzoid XMP is trash...but also I can't imagine most people having the patience to tun the timings.
This was very nice, exactly what I wanted to see, and now I am eager to see the default motherboards configurations test.
BUUT, I do have to admit, I am quite disappointed for not seeing FactorI/O in this series of tests. Given just how much of a performance improvement it gets from 3D Vcache, I would expect to see big gains from better memory bandwidth and/or timings too.
Factorio is very niche. To get the benefit of the 3D CPU you need to live and breathe factorio. It's decently optimized to run on average hardware.
@@dat_21 True, you get the benefits only if you have biiig bases. Because you have basically no benefit from more than 60 UPS. And if I remember correctly, on an already pretty big base, the simulated 7800X3D went to 400 UPS, while the "normal" CPUs were at 200 to 250, something like that.
@@Winnetou17 it's more interesting to benchmark the turntime in the endgame of 4X titles like CivIV or Stellaris. Or do this on Chess/Go engines.
@@PainterVierax The thing is that we already saw the improvements in Factorio for 3D Vcache, that's why I was curious in the first place. And to be fair, I still play Factorio from time to time, though not with huge bases, while Civ I haven't played in a good while. And never played Stellaris or Chess/Go.
I guess here maybe we can have Paradox games too, in the late game they can slow down as well,
I think you have an error in the charts vs buildzoids tunings:
@ 4:50 buildzoid tuned ddr5 6000 primaries: 34-44-44-28
@ 6:50 spiderman "tuned" ddr5 6000 primaries: 30-38-38-28
Shouldnt they be the same?
Yeah, that really confused me as well.
Even besides general stability, one thing you really should have brought up is that aggressively tuning your RAM, especially setting the tREFI that high, can and likely WILL corrupt system files. That timing is especially sensitive to temperature, so filling your PC case with heat from the GPU absolutely WILL bring it past the safe threshold. Keep tREFI below 50k. It gives relatively great performance gains, but it’s not worth corrupting entire libraries of data.
Trust me, I had what I thought was a stable
memory OC that passed all the stress tests, but after a while there were multiple file folders that were so corrupted I couldn’t access them at all and had to run some repair utilities. Don’t risk it.
Also, all due respect to Buildzoid, but setting the tRAS that low is kinda ridiculous and guaranteed to be unstable in the long run. It really shouldn’t be lower than tCL+tRCD, as not only are there limited gains but it WILL NOT be stable over time.
Overall, I’d say many of the timings he gave are only for hardcore memory overclockers. Don’t abuse your RAM, folks, because it’ll take its frustration out on your data.
This comment should be freaking pinned.
TREFI below 50k is rookie numbers though. One should always properly stress test the system before starting to use new settings for 24/7 including high temp situations - the common stress tests don't take GPU heat into account -, and go from there.
@@Cuthalu 33k TREFI gives me TestMem absolutnew errors at 72-73C. I max out at 70C in stress tests during the hottest summer days but I dropped the cooling to test temperature-sensitive timings. I'm sitting at 30k TREFI atm. tFAW is another temperature-sensitive timing, at 16 it may cause errors for m-die around 70C.
Why no tuning on the 6400 kit or was that done on a previous video?
Long story short - buy decent RAM to begin with and don't bother tuning.
Thank you for the effort of making the video, I know it's not financially viable hahah
XMP profiles vary though. Primaries could be similar, but secondaries can vary and adversely effect the performance. Besides, most Hynix A die can be overclocked to 7200mhz. It's decent option instead of Shelling out another 100 bucks for an XMP profile that might not load up.
What I got from this if you are on AM5 tuned DDR5 6000 is what you want so buy a high quality kit that speed. No need to go faster or more expensive on the AMD side.
Time stamp 4:55 -> why should I use CL 34 timings in the Buildzoids table for DDR6000, when I already have CL30 stock ahead of the tuning ??
You should have look into your videos, before you post it in TH-cam.
Can you please test warzone 2 or competitive esports games. the 1% lows gains (if any) would be interesting to see imho
This. It's those games where this performance is generally the most needed. Considering A LOT of people play competive games, they should be tested far more often. For example, OW2 has a high stress benchmark run for workshop that would be easy to test with, but no HW sites use it. :(
great vid
I love tuning memory for the benchmark scores, but I had a similar thing where a couple months later I was having individual chrome tabs crash. I ironed out that issue and still use tuned timings, but I definitely understand wanting a PC that just works.
@RTX Gaminghub I think it was system agent voltage. I had it at 1.25 and raised it to 1.3. I thought the lower voltage would be fine because it didn't start completely crashing until like 1.15 but that wasn't the case
I think if you are raising the two ram voltages by 0.05, then, to be safe, you raise the other three 0.05 (so 1.25 -> 1.300 for the two and the other is 1.35->1.4 just like the two ram voltage).
Great to know 👍
Currently at 6800mhz 32-39-39-28-67 on Kingston Renegade RGB 6400 with stock 1.4v. 13600kf and Asus Z790A-WIFI. Could go higher but I would have to bump the voltage and lose timings for diminishing returns so I prefer to stay on 6800. Responsiveness is amazing. Got nice boost in games.
I’d like to see dual rank vs single rank DDR5. I’m currently running 64GB dual rank 7200 CL32-44-44-44-32
You aren't verifyed?
damn im impressed with the 6000 tuned times
def goin for that
15:41 Degradation or Windows 11 to get BSOD??? DDR5-5600 DR looks good to me. I wonder if you had been tuned too if they will the best bang for the buck.
Why are there no timings for the 6400 kit? Or am I blind?
When will the testing video of the Premium B650 motherboard be coming?
Can you show how much games "fill up" the x3d cache via Windows Performance Monitor. Example: csgo when runing on Intel CPU takes about 20MB of cache, see Performance Monitor window. With larger cache, do games shift more game data to the cache? and by how much given the available x3d cache can be as much as 144MB. I guess it is up to the application to take advantage and up to benchmarker to show (please) the x3d cache is filled or not, aka used.
Did you leave the RING clock stock?
Hi! What memory controller (IMC) and system agent (SA) voltages did you use for DDR5-6000 "optimized"?
Run OCCT small data set for 50 minutes and keep lowering
Okay, but what about 2 vs 4 sticks? Will 4 sticks of faster memory be slower than 2 sticks of slower memory because of the using 4 slots?
The thing about memory tuning is that it's difficult to simulate full system load conditions during memory testing. With tuned memory all of my crashes are during gaming intensive gaming loads when the 350+W GPU heats up the rig. And then you have no idea which timing to loosen and the more time you've spent tuning, the higher the sunk cost seems...
Trfc and then trefi are the most temp sensitive timings reduce those if you are suffering instability due to thermals.
I followed BZ's "guidelines" for stabilizing 7200Mhz and bumped SA PLL, Memory Controller PLL and Ring PLL to 1.035V and had 0 BSODS ever since; admittedly I've only played 5-6h of MS:FS but it was usually enought to crash twice with the PLLs at stock 0.9V. Definitely worth a try
You mention “Stock 7200 kit”. Are you literally installing them and enabling XMP? That’s it? Stable? Would an ASUS ROG Z790 HERO MB (13900k) allow this?
It should yes. it is best to check the mobos QVL list though to see compatibility with Ram kits.
Would any of this have impact on boottimes?
now I am wondering about the final battle: 7950X3D/7800X3D vs 13900KS using buildzoid's timing. He is truly great for leading us down the hardship of RAM OC. Finding out these values is a really time consuming and boring process, he did it so most of us don't have to.
Buildzoid ran so we could walk XD
It can't be a final battle on a board that isn't 2 dimm like dark or apex.
@@fhpchris I mean, that's a rabbit hole that never ends.
There are some legendary Ryzen chips that can do 6600mhz? Should we assume it's not the "final battle" if he is using 6000mhz? With intel there are some kits doing 8400. Should we assume that it isn't the final battle if that isn't being used.
Some intel chips can do 6ghz all core. etc etc.
The final battle is wherever the 99th% percentile of chips can achieve.
IE buildzoids 7200mhz kit with maybe some other tuned cpu parts.
@@Savitarax My 12900ks can do 7200 on a z690 hero (4 dimm). I don’t think 8000 on an apex is something you need a legendary cpu to do. My AM5 rig with a 7950x and asus x670e extreme can’t even post 6000 Xmp. I feel it’s unfair to show 6000 xmp because you literally need a good cpu to even do that. AM5 is terrible.
@@fhpchris "My PC can do this so let me ignore everything you said about 99th percentile chips. Also let me throw in a shot at AM5 being terrible because it doesn't allow me to waste money on 7200 RAM that I'm still trying to convince myself wasn't a waste of money. "
I really want to buy some hardware unboxed t-shirts. Wish there was a merch store.
same
They had merch for sure, idk if they still do
Nice
Does anyone know the difference between g.skill U model and J model ddr5, Qualified memory lists seem to favor the U models.
Where are you guy's find buildzoids tunes timmings for ram over 5600/6000. I planned on having a 13900k on a msi pro board.
I do plan on tuning
I wonder why the change from MSI board to Gigabyte? Any particular reason? I’m looking to buy that MSI actually. On paper it’s better equipped than this Gigabyte model. Great Video. Subscribed!
I’m looking at Kingston Fury Beast 5600mhz and 6000mhz ddr5 40-40-40 latency for my 13900k build but there’s also Gskill Z5 6400mhz at cl32 timing. But can the 13900k benefit from the higher mhz or timings? Is there a point to this or just stick with 5600. Probably not going to bother with overclocking as there seems little point. All are compatible with the MSI z790 mag tomahawk qvl
just stick with 5600, beyond 5200mhz is not benefiting intel, especially 2k / 4k gaming
I think CPU with less cache will benfit more from Ram tuning perhaps?
Looking forward to seeing how DDR4 4000 CL14 stacks up 👍🏻
This
based on a frame chasers video raptor lake DDR4 plateaus around 4000 CL16 with latency in the low 40ms range. They couldn't get any gains beyond that. It would be interesting to see more games benchmarked though. Just make sure its in gear 1 because gear 2 really destroys performance, for ddr4 anyway.
its all about secondary and thirtiary timings brother
Are you tell me i can switch adjust my 7200 g skill with these settings and they are 100 percent stable?
Hi!Top best motherboard b650(m) best price mid-range?
RAM timings just blow my mind tbh.
i have 13900ks is bad the 7600 ram speed ? z790 nmaximus hero
Frig I been watching videos and reading forums trying to justify spending $35 more for slightly higher clocks and slightly lower latency. Your video and benchmarks just saved me $35 lol
Can you include the 13700k, i am most interested in that one and would like to see the results you get
Very nice video, would have been nice to see tunned b die so there was a best of ddr4 on there
From the indapendant community testing on linus tech tips forums about 30 people found tunned b die was as fast as tunned 7200 out side of spider man
Have you considered adding a game that's outside your usual suite which is more memory sensitive for this type of content? Something like CoD might show a bit more results.
if i overclock some 4800 ram to like 5600 ram or 6000 ram is that good
is it me or there is something wrong with tuned timings? They arent the same in "Memory timings used" and benchmarks.
Yeah this video has incorrect info in it and really should be taken down or corrected
Can u do a video comparing this to ddr4 raptor lake?
This video prompted me to do a quick google search and is it true Gear 1 mode isn't possible with DDR5 even at 5600mhz for the 13xxk series? If so then...wtf. Kind of disappointing.
The 1% lows are where the gains are and they are noticeable when gaming.
I can settle for fourth
what about i5 and i7? maybe not that detailed, but they are also great for games and it is more common to get lower end memory with them than on the high end
Thank you for telling people not to OC their RAM 🙌
Meaning?
The 13900KS with the Aorus Extreme motherboard seems to outperform the MSI Tomahawk board by a big margin.
Had you put the 13900KS on the Aorus Extreme it looks like it would’ve outperformed the 7950X3D when you did your comparison.
Happy Days! 👍
Single rank here meant you had one 32gb kit, and dual rank you had two 32gb (=64gb) kits?
no, each 16GB memory had 2 memory ranks of 8GB inside, this don't matter much in DDR5, had more impact in DDR4
I thought it was weather the memory was on one side or two sides?
How can I check if my memory is single or dual rank then? I guess dual channel is a different thing?
maybe it would be useful to include an x3d chip with its best tuned setup as well next time so you can still compare
Fixing the 1% lows is well worth the tuning. Just make sure you stress test thoroughly
Tuning timings doesn't require any stress testing in my experience, as long as PC boots up, doesn't show any weird issues or BSOD and doesn't make the windows startup repair appear, your PC will be stable. Stress testing is more important at every step when increasing voltage and frequency though.
@@DragonOfTheMortalKombat Incorrect. Your computer can boot fine and then crash seemingly at random while running any task. Just ask anyone, or watch the video above
@@Im_Behind_You I know that, but it didn't happen with me. I have experienced such a high level of RAM instability that it literally corrupted windows and applications, but it only happened with frequency tuning, not timings tuning. But then again, it can be different for everyone. I am just saying that it may not be necessary to stress test for hours after changing every value/setting.
@@DragonOfTheMortalKombat too high tREFI value corrupts files 100% don't know about other timings, but definitely some do. It's very hard to pinpoint exact timings as file corruption takes alot of time so it's hard to test.
Exactly the faster ddr5 has much better 1% lows providing a smoother gaming experience
I concur. I've been dabbling in RAM overclocking for years. Every now and again I'll revisit trying to do it successfully again only to find even the latest technology still doesn't stop the stability issue. Theres nothing worse than the PC crashing on boot up and destroying your windows install causing you to re-install the entire OS. And that happens with RAM overclocking.
RAM overclocking was always like that. Very risky and most unstable of all. And these days there is barely any gain from overcloking the RAM in games. And in any "serious" application you absolutely want 100% stability, therefore RAM OC is out of question again.
If you do it properly this does not happen though... It is a tedious process though and should be done methodically.
@@dat_21 100%. From my experience I actually found the PC to be more "zippy" when overclocking the RAM vs fiddling with the timings. Going from 3200Mhz to 3600Mhz when I was using DDR4 was noticeable. Also going from 6000Mhz to 6400Mhz with my current DDR5 was more noticeable than any timing changes. But even with just a MHz overclock and no tuning at all there will still be random crashes. I don't think 1440p or 4k see anything from RAM overclocks.. Just the CPU dependant 1080p.
@@Koeras16 If its tedious, then it's not fun anymore.. RAM is volatile.. When the RAM crashes you lose data even if the program you were using was auto saving. Which can literally break the program or even windows.. And if it doesn't totally break it, that one crash sets the PC up for another crash which goes on and on and snowballs into a broken windows install. You can think you've done the tuning "properly" then a crash 5 days later starts it all off. And for what? 5% fps maybe..
@@ironmaiden5658 Are you on AMD platform by any chance? Their CPUs scale with ram frequency better than Intel. Personally I don't think I would notice a slight RAM frequency bump in any scenario. DDR4 to DDR5 maybe, but it's a massive bandwidth boost, so it's not surprising.
So either 5600 cl28 dr OR 6000 cl30 but tuned
~ $100 considering the entire system (~$3500.00) build really isn't all that much for a ~ 5% gain.
Buildzod's timings are quite stable as long as you have Hynix.
I've been running 2X16GB G.Skill (Hynix A-Die) 6400MT's 30-38-38-28 (Buildzoid timings for Ryzen 7000) FCLK 2133 for a while now gaming / many / various benchmarks. R9 7900X
Improvements over "EXPO" have been impressive. Aida64 Mem Latency went from 67ns to 59.1ns
Read / Write/ Copy improved quite a lot as well to 87K MB/s / 93K MB/s and 84K MB/s / Gaming performance has certainly improved as well.
The only content I'm missing is how all these modern cpus perform when streaming with NVENC. NVENC is a gpu encoder but when you stream on a single pc, you still use more cpu resources than usually and it always effects performance in some way - sometimes a lot, sometimes almost at all. Especially when you use cameras and fancy effects for your OBS scenes. I didn't find anyone on TH-cam who would perform such type of benchmarking.
You lose performance specifically because of obs. If you use shadowplay and nvenc you don't have that performance loss.
@@Pro4TLZZ even with NVENC there is some performance loss. Saying from my own experience. However I have a quad core cpu only.
How can you tuning ram 6000 mhz like that😮😮 can u tell me
Do you have video about that I need timing
@hardware unboxing
Buildzoid🐐🔥🔥
13900ks, z790 apex and 8000-cl32 daily here, its godly! TG ram as Gskill is very inconsistent this time around, and have worse coolers (no pads for heatsinks ect).
I'm more interested in DDR4 vs DDR5 performance comparison for Intel 13th Gen. It would put to rest a lot of discussion regarding platform costs and if it is wise to pair Raptor Lake with DDR4.
Fast DDR5 is faster than fast DDR4. You're welcome.
Steve did the DDR4 vs DDR 5 memory comparison already.
DDR4 is good for low-budget builds or if you already have 32GB of fast DDR4. Otherwise there’s no point as DDR5 is fairly cheap now.
@@Rachit0904 a low budget build doesn't 13th gen though. 12400 and 12600K for example are fast enough and can be had for cheap.
@@roB3rnd when there's DDR4 3600 b-die kits out there that can clock up to 3900 15-15-15 and lower, then that's a legitimate reason to spend less, spend a bit of time tuning and get par or almost on par performance outcome vs DDR5 6000. But if you want to chuck on XMP and call it day, then yeah DDR5's the way to go 🙂