Ryzen AM5 Hynix 24Gb M-die DDR5 timings 2x24GB/2x48GB DDR5-6000 to DDR5-8000 V0.2
Added 2024-12-22 00:00:14 +0000 UTCThis is a more comprehensive version of the easy Hynix DDR5 timings but specifically for 24Gb M-die. Also include dual rank timings which the original easy Hynix timings didn't.
V0.3 improves tRAS and tRC settings.
Comments
run some real benchmarks instead of AIDA. And yes the 8000 setup with 2000FCLK will have less bandwidth than 6400 with 2133. You could try 8000 with 2133 FCLK but that usually ends up performing a bit worse in games.
Actually Hardcore Overclocking (Buildzoid)
2025-10-22 02:07:26 +0000 UTCHey BZ - Using the 6400 vs the 6000 timings (which, surprised the 6400 work for me) on my 9800x3d with an Asus x870e Apex yield statistically identical results in Aida64. Seem reasonable to you? If so, since 6000 takes far less VSOC and VDD, any reason to push to 6400? Anything I'm missing? With the 8000 timings, also boots, and latency is less (by ~1.5ns), but read bandwidth much lower. Am I doing something incorrect?
Stephen Hicks
2025-10-22 00:53:17 +0000 UTCFor 7800 I'm getting errors in y-cruncher (VT3) about 4+ hours into the test... I've tried raising V_DRAM to 1.45V (I can get other tests to pass with 1.35V) and I've tried reducing RFEI to 10000 so I'm quite sure it's not temp related... Any other timings/setting I can change/loosen? MOBO is Tomahawk x870e and mem kit is F5-6000J2636H24GX2-TZ5NRW.
Peter Szmrecsanyi
2025-09-03 14:31:57 +0000 UTCwhat motherboard?
Actually Hardcore Overclocking (Buildzoid)
2025-09-02 02:01:17 +0000 UTCDang, so far both 7800 and 8000 settings crashed after about an hour of stability testing. Suggestions on what else I should tweak, or should I just stick with 7600?
Fr0zen
2025-09-02 01:31:47 +0000 UTCuse the timings for whatever speed you're using.
Actually Hardcore Overclocking (Buildzoid)
2025-07-31 10:03:47 +0000 UTCHi, if the RAM module is 6400 but using 6200.. should one follow 6400 or 6200 timings?
Lukas Ani
2025-07-31 04:21:59 +0000 UTCWhat kind of latency should I be expecting in aida64 with the 'easy' timings at 6000? I'm seeing 78-79ish on a 2x48 build with 9950x3d which is a little higher than the low 60 I got on 7950x with 2x32. Should GDM and PowerDown both be disabled?
Frank Razenberg
2025-07-08 10:13:40 +0000 UTCMy 2x24GB 6400 CL32 kit seems like it can't run tRFC 576 but 592 works; could it be temperature related?
Deddie
2025-04-15 11:34:11 +0000 UTCUnfortunate, guess we’re in the same boat. At least my current sample undervolts well with CO, but might just try some other samples to see. At the end of the day 6000MHz is still not bad if you can stabilize it.
Gary Liao
2025-03-18 04:22:52 +0000 UTCCTCED596G6800HC36DDC01
Gary Liao
2025-03-18 04:21:15 +0000 UTCI just upgraded to a 9950x3d from 7950x3d on an MSI x670e ACE with latest beta bios and I couldn't even get the 2000 FCLK, 3000 MEMCLK fallback timings posting with a 96GB 6600CL32 Dom Plat kit. BZ timings would code 36, and fallback just boot loops during memory training or code 00s. I did finally manage to get the MSI "Hynix Optimized CL6000" (whatever those timings are) posting.
Rich Evans
2025-03-18 04:09:24 +0000 UTCwhat's part number of your memory kit?
Actually Hardcore Overclocking (Buildzoid)
2025-03-14 15:23:18 +0000 UTCJust got a 9950X3D, installed on a Asus B850i with latest bios. Using a 96gb 6800 MHz cl36 kit. With the fallback timings listed here I can only get 6000MHz stable somewhat stable at 1:1 with 2000FCLK and 3000 UCLK. Requires at least 1.275V SOC to get there. Based on your chart here and some videos it seems like this would be a “loser” in terms of IMC quality. Would it be worthwhile to return and try my luck with another sample?
Gary Liao
2025-03-14 03:13:25 +0000 UTCEven if dram temps aren’t hitting 50 degrees?
Stefan Neamtu
2025-01-18 08:15:01 +0000 UTCI’m also on a 7800X3D, not a 9800X3D. Not sure if that matters with this.
Echo10001
2025-01-14 08:29:50 +0000 UTCHi BuildZoid. I’m newer at RAM tuning. The title of this also says 2x48, are these timings good for a 2x48 kit? I’m using a Corsair VENGENCE 95gb DDR5 6400 CL32 kit. Are the “BZ Lazy timings” hard to get to working or generally are they ok? Secondly, when you say “this timing only applies to multi rank”, should I still set that timing to the number u recommend? If not, what should it be.
Echo10001
2025-01-14 08:29:09 +0000 UTCJust a quick update. Turned out RAM wasn’t stable after all, but instead something was bugging. After a reboot, karhu instantly caught errors and crashed the system. Raising vdd to 1.55v and vsoc to 1.1v (might be unnecessary) made it possible to pass karhu (2d), prime 95 (3h) and occt (1h). I might lower trfc from 778 to <700, but I remember you saying trfc needs to be somewhat high for 8000 to work.
w ww
2025-01-12 19:15:41 +0000 UTCWould have to let it run for longer but as for now no. Besides of trfc <700 crash, seems like it isn’t related to trfc. As trfc in Auto didn’t have any changes.
w ww
2025-01-10 14:26:35 +0000 UTCyeah I'm not sure why this would be having issues. Does Prime95 Large FFTs also crash?
Actually Hardcore Overclocking (Buildzoid)
2025-01-10 13:56:19 +0000 UTCSince my RAM looked to be stable on memtest, karhu & OCCT. I went to CoreCycler for some CO, using YCruncher_Old. Temps during CoreCycler are ~47C, during Karhu after some soaking in time ~65C A dedicated fan is already ordered, but otherwise the top fans are configured to the RAM temp. I am not sure if trfc is the issue, I just know trfc <700=crash. Here are the voltages: VSOC 1.05V, VDD 1.5V, VDDQ 1.35V, VDDIO Auto/1.465V. Can’t wait for your next ram timings reaction.
w ww
2025-01-10 13:51:14 +0000 UTCThat's very strange. What are you stress testing with and how hot are the DIMMs getting?
Actually Hardcore Overclocking (Buildzoid)
2025-01-10 13:37:54 +0000 UTCI just leave the Rtts on AUTO.
Actually Hardcore Overclocking (Buildzoid)
2025-01-10 13:37:12 +0000 UTCI am having a hard time setting trfc, everything <700 = instant crash, so I am trying to set it step by step higher to find stable. Until now <900 is not stable. Not sure if DOCP tweaked changes anything compared to DOCP I. X670e Gene, 7950x3D, 2x24GB 8000J404824G, running at tcl36, trcd48, trp44, tras40
w ww
2025-01-10 13:25:28 +0000 UTCThanks! On Asrock (x670e Taichi+7950x), do you set Rtt values manually or a preset (agesa/competitive/aggressive)? Running 2x48Gb 6200mhz CL28 with your timings atm, passed 6h of TM5.
Bruno Filipe
2025-01-02 02:26:48 +0000 UTCDon't run max REFI without cooling.
Jack E
2025-01-01 19:19:18 +0000 UTCI wish you could test the Phoenix's APU in near future 😉
Irwan
2024-12-31 13:16:17 +0000 UTCYeah I haven't tested these with the APUs at all. Though I'd guess the timings probably aren't the issue.
Actually Hardcore Overclocking (Buildzoid)
2024-12-30 17:12:15 +0000 UTCDear Buildzoid, I have a pc that using B650M Pro RS, KingBank 48GB 6800MT/s, and 8600G. I've already tried the are the configuration... But, I just find out that the most stable one is the 6000... Maybe do you have the more specific configuration for the AMD AM5 APU? Thank!
Irwan
2024-12-30 14:41:58 +0000 UTCThe most useful $5 purchase I have ever made! Keep up the good work!
Ahmet Dağ
2024-12-27 22:30:03 +0000 UTCCan you substitute 50000 refresh for 65535 in "Lazy" and eliminate the recommendation for ram fans or does that come with the combination 1.43v and refresh?
Nathan Sherman
2024-12-24 16:58:15 +0000 UTCGood stuff. We all appreciate the hours invested. Looks to be some fun holiday experiments once the final pieces of the testbed arrive when I earn a break from my charts/studies.
Jeb
2024-12-23 23:07:49 +0000 UTC
![ゲイAI[チョモランマ]](https://nokimo.com/istorage/88150.jpg)
