AMD Ryzen 7000 Series [Zen 4] RAM/Memory Scaling feat. Tuned Memory
Vložit
- čas přidán 13. 08. 2024
- Support us on Patreon: / hardwareunboxed
Join us on Floatplane: www.floatplane...
Actually Hardcore Overclocking: / actuallyhardcoreovercl...
Buy relevant products from Amazon, Newegg and others below:
AMD Ryzen 5 7600: geni.us/qGJVbZ7
AMD Ryzen 5 7600X: geni.us/Hn7m9
AMD Ryzen 7 7700: geni.us/6VP8wD
AMD Ryzen 7 7700X: geni.us/fAfH
AMD Ryzen 9 7900: geni.us/RRL3Fa
AMD Ryzen 9 7900X: geni.us/Rjlq
AMD Ryzen 9 7950X: geni.us/lTmCsO
Intel Core i3-13100: geni.us/DfDsan
Intel Core i5-13400: geni.us/lQZFpt
Intel Core i5-13500: geni.us/goRuq2
Intel Core i5-13600K: geni.us/lJgCa
Intel Core i7-13700K: geni.us/Gpuqh3
Intel Core i9-13900K: geni.us/AieGiL
00:00 - Welcome to Hardware Unboxed
01:14 - DDR5 Memory Timings
02:01 - Watch Dogs: Legion
04:29 - Hitman 3
05:02 - Horizon Zero Dawn
05:31 - Shadow of the Tomb Raider
06:44 - The Riftbreaker
07:09 - A Plague Tale: Requiem
08:07 - Marvel’s Spider-Man Remastered
09:05 - The Callisto Protocol
09:28 - Final Thoughts
AMD Ryzen 7000 Series [Zen 4] RAM/Memory Scaling feat. Tuned Memory
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 - Věda a technologie
For more information don't forget to visit Actually Hardcore Overclocking: czcams.com/users/actuallyhardcoreoverclocking
What kind of difference with a different CPU or motherboard?
@@robertlawrence9000 They've discussed why doing videos and tests that way isn't the purpose of what they're doing or showing us, in previous videos.
@@mick7727 I know they said with low CPUs and with 4090 at 1080p but I'm talking about ram may work differently on different CPUs and mobos. I'm asking for confirmation
Do you have the DDR5 5600 Tuned Timings as well?
steve can you do a DRAM stick comparison with 2 VS 4 to see how the thing had change from zen 2 , 3 to zen 4 ?
"The thumbnail was not photoshopped"
I'm pretty sure it was photoshopped, he's missing all of his gang tattoos!
This is SPARTA!
That's from just after the new studio build surely.
Overclocked Man.
not photoshopped... he used affinity photo... so it was "affinityated"???
I think it is incredible how often the tuned 5200 memory was about equal to the non tuned 6000 memory. Thanks Steve for more incredible videos
I cant read "Thanks Steve" without the intel woman voice saying it in my head
this one: czcams.com/video/b8h5WGyz4P0/video.html
@@TheNubaHS LOL
I saw their video of ryzen 3000 with tuned memory using DRAM calculator, it was one of the best and most helpful tech videos I saw on the internet.
@@TheNubaHS Thanks Nuba
I wont say it's incredible, as it was kind of the same with DDR4 on Zen 3. It just shows how some brands just push big numbers to match some marketing in lieu of really providing improved performance.
Yeppers - Buildzoids 4 part guide was a huge help. Even if he considers it low effort content or tedious and boring, that information allowed me to figure out where the trouble spots where (and there was many) and take the stable at reduced mt/s speed XMP/BIOS derived defaults to rock solid stable at actual rated speed & performing exceptionally well at rated speed.
Mind a link?
@@colors5108 Links get deleted by CZcams I found. So just head over to Actually Hardcore Overclocking - it is Buildzoids channel. He has a good amount of stuff on the subject.
link would be nice!
Aint no way this dude said yeppers
@@sigmamale4147 No sir - I wrote it. ROFL
This was amazing. What’s remarkable is how big of a difference there was on 1% lows across all the tests. Just looking at the lowest 1% low vs the highest 1% low on each chart, the difference is at least 20% on most charts and exceeds 50% in a couple of them. That’s such a massive difference just from manually tweaking some settings.
What we really need is an update to that Ryzen memory tuning utility that would provide good memory timings for every kit on the market.
With DDR5 that really isn't needed as there are only a few memory ics at the moment and it is very easy to get a hynix kit
Would love to see some tuned high frequency scaling tests with Raptor Lake
It's planned.
@@Hardwareunboxed feel free to ask anything about ddr4 dual rank Samsung bdie. The timings go so low and intel CPUs scale significantly with the bandwidth and latency improvements
@@Hardwareunboxed FINALLY POGCHAMP !!!
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 so I prefer to stay on 6800. Responsiveness is amazing. Also got nice boost in some games.
MVP
You should include those “tuned” timings alongside stock ones for the future. A lot of people like to see stuff like this and it really helps people.
I think the ones that don’t say tuned are stock
Seems like a waste of time. Barely anybody overclocks their stuff these days and for a good reason. Marginal performance improvements aren't worth the hassle. In most realistic configurations (i.e. not 4090 on 1080p which is stupid) ram won't make any difference whatsoever.
@@dat_21 You so don't have any kind of high tier gpu. I play in 1440p 240hz with 4090 and it makes massive difference having low tier ram vs high tier and even at 4k on some games 4090 will make cpu work hard enough that ram does make difference, having 120fps vs 150 is massive in terms how the game feels on proper gsync monitor
@@xTurtleOW I don't know about that. If you have money for 4090, you have the money for 6000CL30 kit for sure, and beyond that it is in the diminishing returns area. I don't think overclocking is really worth it at this point. And difference between 120 and 150 isn't massive, it's very very slight at best.
@@dat_21 bro a 20% gain is not marginal. In some of those games the minimums are drastically higher
The gains in the 1% lows are massive for the tuned kits. Great to see.
This was great testing. Love to see some "low effort" BZ
I love how well your videos are made. Even though this is 7 months old I am able to use it (and others from you) to put togther a system that is primo. Due to a sale locally, I just built myself a 7800X3D on an MSI Tomahawk B650 motherboard with DDR5 6000 CL30 Trident Neo RGB and am confident I have about as good as it's gonna get. Thanks for having these comparisons!
So AMD tendency to like tight timings seems to remain relevant!
Nice to know! And good stuff as always.
And thanks to the Bullzoid also for those memory timings! He talk a lot in his videos, but he also know what he is talking about! I raise my hat! ;)
Really informative and beneficial. Thank you Steve and for making Builzoids info into bite size chunks
I nearly had a heart attack laughing at the thumbnail! Good job!
😂
Yeah, and it's a real picture of Steve, when he's on the wilderness in Australia, hunting those venomous beasts they have over there.
Right, Steve?
After seeing the vid of Steve building his "shed" (it was a house) by himself, I wouldn't be surprised if that pic was real 😂
@@gustavo_vanni Koalas?😜
@@johnscaramis2515 Exactly, these beasts are the most dangerous things to ever exist.
Great physique, Steve ! Did you build that muscle lifting all the new gen gpus ?
The improvements to 1% low fps (the statistic that matters most when looking at fps, imho) with tuned memory in Spiderman are insane. That's prolly around the same kind of "generational" boost that spending the premium for the x3D models will bring. Hope you'll take a look at that when the X3D parts are tested (including a X3D+tuned memory bench in Spiderman).
Yeah we really do need that data now, I’m 😮
OMG awesome! Thank you so much for putting this video out. Nerds everywhere rejoyce! It was a struggle to find memory timing videos for last gen ryzen. So thank you :)
Wow that tuned memory really makes quite the difference! I'd be interested to see how much less of a difference it makes when you have a ton of L3 cache like on the upcoming x3d chips.
Great video! Please revisit the Buildzoid Hynix custom tuned timings with the incoming X3D parts!
x3d chips tend to not care about memory at all. Going from 2666cl16 to 3800cl14 with custom sub timings gains you a staggering 1-2% depending on the game
@@tqhr I think we will see the same for 7800X3D, maybe to a lesser extent than in 5800X3D due to more cache in general.
But I wonder how the dual chiplet X3Ds will react
@@tqhr and that is exactly the reason i'm suggesting it.
Thanks for including the Kingston Fury Beast DDR5 6000 CL40 in this teat. I am looking for this RAM as it is one of the few DDR5 6000 RAM that exists here and are not as expensive as the rests. I'm not a fan of memory tuning as I'm not good with tinkering the PC so I will always leave them on stock. However, seeing the boost in performancce with the timings provided in this video made me want to try it out. Hopefully it will be stable.
stock? not even XMP?
Their 5600 CL40 kit is also M Die, should have similar ability to tighten timings down very well
@2intheampm512 their 5600 kit is far less likely to be mdie than their 6000 kit.
@@MrDutch1e All Kingston DDR5-5600 and up use hynix and about half of that is M-die. You just have to know which part numbers are M-die. If you go up to 64GB they even have a kit at 5200MT/s that has M-die.
Check out Buildzoid's "Easy memory timings for Hynix DDR5" video from a couple weeks ago. He made it for all hynix dies not just M-die so any of Kingston's DDR5-5600 and 6000 kits will work. Buildzoid basically has you leave the first 3 primary numbers alone so if you got that CL40 kit it would stay 40-40-40 ect. Just follow his advice and you'll be set.
Great video, thanks for the hard work!
I can see people watching this many times over :)
Would be interesting to see how 13th gen responds to the tuned timings in an Intel edition
Love the thumbnail. Steve is a BEAST!!
Pumping iron while he waits for the results of those epic benchmark marathons.
Been a Buildzoid fan for a good long time. Thanks for shouting him out and using his knowledge in this video it was cool seeing the collaboration between two of my favorite channels
I'm thrilled to see you point this out! I bought a system like this recently, and on a home-built semi-random memory bandwidth microbenchmark, going from base DDR5 settings all the way to buildzoid-inspired tuned settings netted an unbelievable 46.5% semi-random bandwidth increase. Obviously games won't be quite as memory sensitive, but it's great to see a prominent youtuber like yourself do such a public service announcement; these differences are very significant, and so easily achieved; it'd be a shame if users miss this opportunity simply for lack of publicity.
XMP/EXPO are honestly pointless now with ddr5. You can a fraction of what you get when manually tuned and don't get the guaranteed stability of jedec.
It will be really interesting to see how the X3D chips scale.
It won’t matter nearly as much for X3D. It won’t be grabbing data from system memory as often.
The buildzoid timings give 8% increase to my 7800X3 running on D6000MTs CL30.
I know it's necro but really wanted to correct the misinformation from the other guy who responded.
the thumbnails are getting better and better
These guys are going above and beyond with their benchmarks
They are a pristine fortune for the community
Can you share all the timings you ran for the SR and DR DDR5-5600 tests plzzz :D
Those tuned 5600 DR kits seems to be the sweet spot if 6000 and above is out of reach
Especially the DR kits
doesn't matter - either needs manual tune and 99% of people not capable of any manual tune.
@@mroutcast8515 The video gives you the figures to tune with though?
@@xPlasmaz Yep, that was infuriating tbh, not even the 5600 kit model names are mentioned.
@@xPlasmaz lol, they will be different for every kit IC and even silicon quality. Different IC's will be capable of different timings.
Damn this is a video we need to save us money, DRR5 timmings are all over the place is good to see a video that will actually show the difference between them. And the funny things is some of these slower rimming ram is just a few bucks cheaper so you would think you aren't losing much. Thank you so much for this.
the issue is that most brands doesn't advertise what RAM chips they use on each product so you need to get a full part number to search for otherwise it's a lottery.
Curious how much the X3D cares about memory ^^
should be less dependent on memory
I saw your video of ryzen 3000 with tuned memory using DRAM calculator, it was one of the best and most helpful tech videos I saw on the internet.
SK-Hynix (A-Die) here.. got it this week for $165 ($185 with Tax) Model is G.skill Ripjaws S5 6400MHz 16x2
Last week I tried " 30-36-36-30-68" 6200MHz at 1.4v & used sub-timings for maximum performnce.. on my Ryzen 7700x
6400MHz didn't work well.. I think Bios will hit 6400 MHz in next few months.. but not now (Asus still on AGESA 1.0.3)
But let me tell U.. at 6200MHz aggressive timing.. it's can beat even better frequency OC.. I won't be surprised if it's can match 6800 kits
I heard some people say 6400 is manageable with AGESA 1.0.4 update (MSI..etc) but only time will tell.. will try 6400 when Asus release it
Note:
With the right kit (Hynix) All boards can do 6000.. majority of boards can do 6200.. but not all boards can do 6400+
The sometimes huge 1% low uplift is particularly important, since it will be more relevant at realistic resolutions.
Amazing work guys, thank you.
Kinda wish you covered Infinity Fabric tuning (2000MHz+) and DDR5 clocks over 6000MT.
Either way, the gist seem to be latency/sub-timings are king with Zen4.
Hopefully when you review 7000X3D you also have a small section dedicated to the things mentioned above.
Because, Infinity Fabric tuning on a 7700x is pointless, you gain nothing, other than maybe some stability issues.
Yeah I remember AMD talking about infinity fabric clocks not needing to be linked to memory for best performance anymore. But I have seen zero testing of this at all.
In the video that hardware unboxed is taking the timings from by buildzoid, he said to use infinity fabric at 2033mhz and ddr5 at 6000mhz, as anything above is unstable and not worth the effort
@@DKTronics70 You assert that, but we want proof of it.
Exactly what I was looking for that was lacking from the previous video. The conclusion of subtimings having a difference of 10%, which could very well be a margin from a good/bad motherboard-auto or manual tuning, was what was missing in the stratification of the previous data.
Impressed with that 5600 DR Tuned performance, wow!!!
I have a severe flashback to the time when Zen 1 was just released in 2017. With people cracking on infinity fabric, ram speeds and timings
Seems like 5600 tuned with tight timings is the way to go.
Why was the thumbnail photoshopped to make Steve LESS buff than he is in real life?
Top-notch stuff as per usual guys! Glad to see you've got ripped while benchmarking all day long.
Great follow up piece. My ram is tuned higher and the scaling continues in many games. It's a shame the motherboard manufacturers left so much performance on the table. I imagine aggressively tuned kits will be on average pretty close to the upcoming 3D parts but we'll find out soon if that's the case.
This was like the 1 thing everyone wanted to see for overclockers. How much extra performance can I get with those timings on top of frequency.
Thank you Steve for the hard work here. I could not imagine dealing with manually tuning these kits with the additional AM5 memory training these motherboards require, which simply sit there for 3-5mins on a black screen. It’s a gruelling process and you wonder if it even is working.
Thank you for highlighting the importance of secondary & terciary timings Steve, since XMP profiles still don't contain the most important timings like tFAW/ tRRD's & the SCL's. More content with buildzoid please it's awesome 😁 getting most out of your RAM&CPU with timings is super useful! Regards 🤟🌹
Trfc and trefi make a much larger difference in games than the timings you mentioned.
@@MrDutch1e was just pointing out some of the timings 😁 a combination of important timings makes for an optimized setup 😄 most reviewers/ youtubers don't focus on secondary & terciary timings at all... XMP is dog****
@Doomx12345 ya especially on new ryzen platform. Trfc defaults to over 800 lol. Brutal. Trefi is a bit better than intel by default tho. Which is surprising they didn't screw it up since it wasn't even on am4 platform.
@@MrDutch1e do you have a link covering that? What’s a generally good start point for that? I have Samsung DDR5
Love the thumbnail.
I don't have the patience for full tuning and don't play at 1080p anyways, but I will say that I am glad that my 5600 kit I bought randomly turned out to be Hynix kit - juicing it to 6k and juicing down the latency was pitifully easy, haven't failed on a setting yet
Steve is the man! Steve doesn't disappoint. Thank you Steve =)
Well, that was extremely useful!
Tuning your memory obviously makes a huge difference with Zen 4. It's nice to see that even the cheap 5200 memory works really well once tuned, but also interesting to see (also considering results of the previous video) just how much performance is held back by most memory if it ISN'T tuned. You really aren't getting your money's worth from a Zen 4 CPU unless you tune your memory or use certain premium memory kits, and tuning is probably always going to be worth it, even with a premium kit. Of course, the X3D chips might turn out to be a major exception to this, just as the 5800X3D wasn't.
I could also add that, chances are that your graphics card and settings are such that you won't actually benefit that much from tuning your memory, at least not yet, but it does seem like a bit of a waste to not be tuning your memory on a Zen 4 CPU. It's basically just free performance, and the amount of time and effort it takes to tune your memory really isn't that hard. You can just look up Actuallyhardcoreoverclocking, or another resources, to get easy to follow directions for memory tuning.
It's how I ran my 8350 bulldozer at 2133 to 2400, 4.7ghz on water, ram had heatsinks. I reached 5ghz but I just kept it at 4.7 for stability
Very eye opening and awesome work! I’m guessing frames per dollar using tuned memory makes only a small difference, but maybe worth checking in case a top end card can be matched by a less expensive one with better tuned RAM 😅
Thank you. Finally some ram tuned benchmarks
Wow, the 1% really jumps for the manually tuned timings. This was really comprehensive! Not just looking at the DDR5 6000 timings, but also 5600 and 5200.
I wanted to improve my DDR5 5600 kit, so thanks for that.
Great thumbnail, you guys just keep getting better and better each round :).
It's very informative and in line with what i've seen in another reviews, that's why the results with Zen 4 varies so much.
If the 5800x3D is an indication, then memory bottleneck shouldn't be a problem with the X3D parts.
Is there a deadline for the review?
The 5800X3D still scales with memory but not by as much. I was still able to get another 8% to 10% more performance just from memory tuning though.
@@WereCatStudio (X)
@@Rem_NL If you doubt then check HWbot rankings for 5800X3D in R20 and R23. Have in mind that Cinebench scales poor with memory so 3% to 5% gains in CB is much higher in games.
@@WereCatStudio I agree. The extra cache acts as a latency buffer so it's less sensitive but when this buffer is maxed out by memory intensive computation, the usual timings do matter.
@@tilapiadave3234 you also have to take into consideration that part of the hype is due to the fact the 5800X3D is the ultimate upgrade for AM4 when AM5 is overpriced. It's quite like the i7 4790k or if the 5775C was unlocked.
Seems consistent with BZ’s general thesis on DDR5 that not only is frequency not the most relevant, but even the primary timings are less relevant for performance than tertiary timings
Gone are the good old late DDR3/early DDR4 days of “just get some motherboard and some memory, it’s not going to affect anything”!
To be fair, this was never true for DDR3/DDR4 either. We just didn't have enough people who knew their shit and weren't pushing any games to 240 Hz. Where memory tuning really matters are competitive titles, especially because you want to keep those 1% lows as high as possible. I recently went back and tuned some DDR3 on an FX8350. Got me a 20% improvement in TimeSpy CPU score going from 2133Mhz auto timings to manually tuned. That being said, as these were green PCB 1333Mhz CL9 1.5V sticks, having to push 1.7V+ into them caused some temperate related stability issues eventually.
Even on DDR4 with 64-banks per channel (2x DR or 4x SR 16-banks) "primaries" ("inner-to-bank") timings did barely anything, the "speed of switching between banks"-timings ("tertiary") gave much more performance boost even then, especially tRRDS+tFAW timings.
Primaries are more important when you have less count of banks per channel, ie 32 per channels (DDR4 2x SR 16 banks or DDR4 4x SR 8 banks or DDR3 4x SR 8 banks etc) and even more for 16 banks per channels (like DDR4 2x SR 8 banks - which is crime that they don't show the banks counts in characteristics, see Linus' video "The Dirty Way Manufacturers are Downgrading Your PC")...
On DDR5 the "standard SR" is 64 banks - so "primaries" do almost nothing. But soon there will be DDR5 with only 16 banks per stick - and their performance again will be worse with the same frequency/timings and "primaries" will give them more boost.
You guys must have heaps of fun coming up with the thumbnails. What a crack up!
Now I'm counting on you making a video how to do a proper tune on G.Skill 6000 30-38-38 kit :)
Nice info! Same video but with Intel 13th Gen next please!
Would like to see the same for intel 13th gen, 2-DIMM boards can do 8000+ with Hynix A-die 👍
It's planned.
Steve living out his best life in the thumbnails!
Steve bro! You are jacked!
Man, this is the fkng content I wanna see !!! 👍😎
brutal job
PS: Steve, you owersaw his 1 little trick - not 2000 FCLK, but 2033. As he said, for some reason it just works the best and gives additional ~2%... just sain
And second edit - I found it: czcams.com/video/dlYxmRcdLVw/video.html
Buildzoid is an absolute gem. Deserves more recognition
With latest bios (which I think came in december if I recall) , my ddr5 expo 6000 it doesn’t stay stable, tests find issues, have to run it default speed, no expo profile works. 2 sticks, 32GB total, in the correct slots and installed correctly. Waiting for another bios update that helps…
I am living this nightmare. Same kit probably.
4k benchmarks please will using faster Ram really matter
Very interesting results, will definitely tune the timings on my 7000 series cpu :)
This is the reason, why this channel is great. Important topics that are too tedious for others to explain and supported by facts. Thank you so much!
7:48 From the 4800 kit to tuned dual rank 5600 we see well over 50% increase in 1% lows. And 6000 to tuned 6000 30% increase in 1% lows. Those are massive numbers on 1% lows, and really 1% lows is the important part to improve.
8:43 from 6000 to tuned 6000 almost 50% improvement on 1% lows.
There seems to be a mistake on the Buildzoid's timings. If tRRDS is 4 then tFAW should be 16. tFAW is tRRDS times 4 as seen on other kits.
To my understanding tFAW affects bandwidth quite a lot so if you can't get it down to 16 the next values to try are 6 and 24. Now their cycles don't match.
HWUB is now my favoriite tech youtube channel. i have been asking for mainstream reviewers to go over ram tuning results for ages now and it finally happened.
Reviewers that only set XMP and forget are the bane of my existence. leaving (in some cases) next gen performance gains on the table while still spending insane amounts of money on premium kits and motherboards.
Buying 13900k and using 6000mhz ram XMp. That is a lot of performance left on the table.
Thats a big amount of work guys.
Thanks for testing!
Altough i would‘ve appreciated a tuned 6400cl 32 in this list because i‘m using one myself with an 2133mhz If😅
Same, I think it might be faster but I have no clue and no time to test. If it’s not it might be better to clock it down with cl 30 timings but rn I sadly have no reference :/
@@leucht i got about 10% more performance than on 6000cl 30,38,38,96,96!
Tested a lot of timings and clocks because my ram ran not stable with older bios versions
Love these kinds of tests. I would however have just liked to see at least one or two tests at 1440p and 4k as most gamers with that kind of hardware will not be running at 1080p. I understand you are doing the tests at 1080p to get the point through but for someone running at higher resolutions we have no idea if there are any gains to be had.
That thumbnail though.. Wow Steve.. You've been working out havent you lol ;)
Thanks for sharing the timings! RAM OC is hell. Being able to just plug in timings is great.
Edit: comments say don’t plug in values off the internet.
Can confirm.
Except kit to kit variation is so great that you can't just plug in the values
Rule #1 of memory OC: Do NOT copy other peoples ram timings. This is especially true if you arent experienced at memOC
I mean Buildzoid shared them weeks ago.
@@OCDyno This
wow great work!
Any info on how RAM (speed and CAS) scales with production workloads?
(I assume not as much as gaming)
Was about to comment this
As always it depends on the specific application. It strongly matters when you have a significant amount of data to compute, like encryption, encoding, conversion, compression or checksum tasks.
The difference with gaming is that those productivity tasks rarely require realtime constraints and you can run them in the background or when AFK.
i hope you get some well earned rest after this so you have energy to do it again for the X3D variants soon enough :X.
excellent content as usual!
Hey guys would you share your memory timmings as your video cuts before the memory trimmings screen ends 😊
Is there a good tutorial on how to tune memory timings?
Copy the timings/voltages shown in the video into your BIOS. Like this: czcams.com/video/KOqhyVNPhaM/video.html&ab_channel=HardwareUnboxed
If you want a quick tune with big gains - reduce tRDDS+tFAW+tRDDL - these 3 timings on many sticks are responsible for ~50% of total possible boost from tuning all the timings combined.
On AMD CPUs you can always safely set tFAW to the minimum possible value. IDK about Intel, so to be sure set it to no less than tRDDS*4.
tRDDS value on 95% chips also can be set to the minimum possible value. And tRDDL is from 0 to 4 (DDR4) or 8 (DDR5) bigger than tRDDS, need to test your chips specifically.
Good timings for DDR4 on AM4: tRDDS 4, tRDDL 4, tFAW 6 or tRDDS 4, tRDDL 6, tFAW 6.
(Test your settings in the end with tm5 with extreme profile for many cycles.)
I recently replaced my horrible 4800 CL40 that I got for basically scalper money at Alder Lake release. Got a 6400 CL30 kit. Went from 730 fps to 840 fps in Valorant, so finally playable.
nice joke XD
Yeah, I need at least 1 kilofps to not get a headache from the stuttering
So brave
Sadly I can't hit like more than once. Thank you for your hard work and research on this kind of stuff. I am waiting on final parts but putting together an AM5 system and am now much happier that I went with the 5600 CL28 RAM and now even have the exact timings that provide the best results. Lets hope mine is stable :)
Hey, where did you find the complete 5600 tuned timings?
Steve is ripped! 🤣 Just for the improvement on 1% lows alone is well worth exploring tuning your ram.
Amazing video. Do more of these OC tests, please!! 🙏
That thumbnail is gold
Thank you for including the 5600 CL28 kit!
Hopefully DDR5 will mature is a bit more and we can get the 5600 timings on the 6000 kits.
I want to overclock to 6000mhz with ram 5600 kits using high voltage. Is it possible? I have not found tests on this topic. I would like to know before buying..
This is such an informative video. I am in the process of building a AM5 gaming rig, and I never tune my memory as I try to go with the best tuned kits out of the box. Thanks!!!
Platform cost still an AM5 issue.
If you don't plan on upgrading for a long time yes but not if you reuse the same board across multiple generations.
@@Alex-qq1gm That's on the assumption that amd will make the am5 platform last as long as the am4. 2026 is rather relatively close.
@@mangatom192 its only 3-4 years away when they stop support
@@Alex-qq1gm I tend to help to many people with choosing the parts.
for all entry to mid range builds I cannot recommend ryzen 7000. It's just too damn expensive, part of this is because of local shops and importers but the other is because of the prices of components that manufactures require. I find myself recommending ryzen 5000 and intel 13th gen with DDR 4 much more then 7000. it's just not worth the cut that other parts will suffer due to buying ryzen 7000.
LOL, your imagination in doing thumbnails knows no bounds.
😄
Pretty excited for this platform just bough motherboard and ram and waiting on the 3D v-cache cpus to be released or pre order fingers crossed I’m not let down
Buildzoid is a powerhouse. Love the guy
You are spoiling us , we love it , and ofc we want more.. I would love if maybe the future cpu reviews u make come with 2 recomendations of ram kits : 1' the best out of the box kit for that specific cpu and 2' the budget tuned kit alongside with the timings table. I know about silicon lottery and all , just asking for a 'safe tuned guide ' and not borderline maxed out waitin for catastrophe to happen timings.
How do you infer anything about 4k gaming in your final thoughts from these tests when you test in 1080p, though? That didn't seem to make much sense.
I didnt knew Steve was so ripped under that shirt. Thats some quality tshirt keeping that muscles under control
Thank you very much for sharing this. This is a complicated subject and your presentation provides a lot of clarity to it.
Thumbnail notification squad, reporting in.
Thanks, got myself some M-Die. Will use those settings!
Thanks for al the hard work 💪💪 ❤
Lifting all those 4090s had some serious gains!
Wow, this is complicated stuff. I remember the good old days of get the highest number you can but which is also a good value, try not to mix brands or capacities, and call it a day. Then I learned maybe enable XMP or EXPO, depending on your motherboard/cpu, and i felt like an expert. Now this, this is beyond my abilities :) Anyway, I am certain people smarter than me will benefit greatly from this video, and I'm glad you guys made it with this level of detail. This must have taken FOREVER to make with all these tests, which we certainly appreciate. Have a great night Hardware Unboxed.
Top. Notch. Content. Thank you so much for the time you put into this! Funny enough I just saw Buildzoid’s video yesterday and have been working on tuning my RAM and this shows it’s absolutely worth it.
Great video. Follow up vid including tuned + overclocked memory and infinity fabric OC?
Perfect timing in preparation for Zen 4 X3D :)
X3D chips don't care about memory speed or latency because the cache takes over much of the latency sensitive role memory plays for achieving high fps with traditional CPUs.
@@vx9330 Yep that's the theory - but good to see it benchmarked too
Very nice, great to see all the charts comparing these changes.