SK hynix A-Die (Overclocking thread) only for RYZEN AM5 users

[Krit]

Hello everyone! I wanted to save some time please share your Zen Timings SK hynix A-Die / RYZEN 7000 or 9000 Series 100% stable custom end result screenshots. I need these examples for a faster overclocking success as there are no such a thing as a dram calculator or something similar. From scratch it will take a days to get the best results and full stability so i need a faster way to get it done.

Use Zen Timings it's a simple app and shows all needed info about timings and voltages.

Simple example.

ASRock B650 Pro RS + XPG Lancer Blade 2x16GB, 6000Mhz, CL30, AX5U6000C3016G-DTLABBK, EXPO enabled.

 

[_roman_]

these are my daily settings - long term stable - compiler fool proof

I quote my post from 1.April from topic: https://www.techpowerup.com/forums/threads/ddr5-cudimm-explained-benched.329420/

I really have no knowledge about overclocking. I do not tune it to dead.

Other settings are windows 11 pro 24h2 stable. But will crash the gentoo compiler. = unstable.
I tried to figure out a year ago how to get all those windows test software and how to use it. Useless in my point of view.
When my gentoo crashes during compiling. And it will - I know its unstable. I can play for 12 hours a day in windows 11 pro without any crash with those "crash" settings.

I only change one setting and test it for two weeks. If something wrong I revert via uefi profile on my flash drive for uefi updates. (asus and msi have a feature to save all uefi=bios settings to a usb pen drive)
e.g TCL 30 = won*t work - i revert = does not boot - needs more voltage most likely. I do not want to go over 1.35V DC yet.

-- I asked ages ago in another forum / topic to determine the chip type. Mostly all settings per hand on a cheap - entry mainboard - with a cheap entry processor.

-- Settings were 1.25V stable - I just raised to 1.30V DC on purpose. // Anything else crashes my compiler. Raising Voltage on my work box is not to my liking for maybe no benefit. // I'm well aware of that DDR5 runs with 1.40 or 1.45V DC and with active Fan cooling for DRAM.

Note: Do not buy this particular mainboard - It has an issue with powersaving mode with intel wlan.

edit: as someone else asked for memory temperature

spd5118-i2c-1-53
Adapter: SMBus PIIX4 adapter port 0 at 0b00
temp1: +38.2°C (low = +0.0°C, high = +55.0°C)
(crit low = +0.0°C, crit = +85.0°C)

spd5118-i2c-1-51
Adapter: SMBus PIIX4 adapter port 0 at 0b00
temp1: +38.0°C (low = +0.0°C, high = +55.0°C)
(crit low = +0.0°C, crit = +85.0°C)

SPD5118 - SPD HUB for DDR5 Memory Modules

The SPD5118 hub device is a DDR5 Serial Presence Detect EPPROM with Hub function and integrated temperature sensor is used for memory module applications.

www.renesas.com

 

[ir_cow]

SOC and CLDO and VDDIO is often wrong in Zentimings just so your are aware.

 

[A Computer Guy]

Hello everyone! I wanted to save some time please share your Zen Timings SK hynix A-Die / RYZEN 7000 or 9000 Series 100% stable custom end result screenshots. I need these examples for a faster overclocking success as there are no such a thing as a dram calculator or something similar. From scratch it will take a days to get the best results and full stability so i need a faster way to get it done.

Use Zen Timings it's a simple app and shows all needed info about timings and voltages.

Simple example.

ASRock B650 Pro RS + XPG Lancer Blade 2x16GB, 6000Mhz, CL30, AX5U6000C3016G-DTLABBK, EXPO enabled.
View attachment 394438

What is your goal more specifically? For example: getting highest RAM frequency with MCLK:UCLK at 1:1?
Does MCLK:UCLK ratio matter in your case?

 

[Krit]

What is your goal more specifically? For example: getting highest RAM frequency with MCLK:UCLK at 1:1?
Does MCLK:UCLK ratio matter in your case?

Tight timings with 6000-6200mhz frequency. I don't like high voltages.....

6000-CL26
6000-CL28
6200-CL28
6200-CL30

What usually is stable FCLK for Zen 4 ?

 

[Dr. Dro]

Quick recipe:

tRAS should be tCL + tRCDRD
tRC should be equivalent to tRP + tRAS + 2 clocks.
Raise tREFI to 65535 clocks, and test.

28-36-36-36-64-102 2T 65535, 1.40V would be where I would start. If consistently unstable, reduce tREFI to 32767 clocks. Large interval here improves access latency as the memory refreshes less often, but as memory cells leak charge, it will get unstable very fast if the chips are hot.

 

[VuurVOS]

Tight timings with 6000-6200mhz frequency. I don't like high voltages.....

6000-CL26
6000-CL28
6200-CL28
6200-CL30

What usually is stable FCLK for Zen 4 ?

I used this sheet for calculating the memory timings.

DDR5 Timing Calculator for AMD V0009

docs.google.com

Since my dimms are rated for 1.45v, I will run them at 1.45v!

 

[Dr. Dro]

I used this sheet for calculating the memory timings.

DDR5 Timing Calculator for AMD V0009

docs.google.com

Since my dimms are rated for 1.45v, I will run them at 1.45v!
View attachment 394440

Nice! This timing set is perfect, my starting point was a bit more conservative, but ideally this is what you want for C28, and what you should be gunning for.

 

[Cowboystrekk]

This is a weird one. A-die dual rank 16gb. 8gb on each side of the sticks. They max out at 7400 so not as good as A die 16gb single rank.

I used this sheet for calculating the memory timings.

DDR5 Timing Calculator for AMD V0009

docs.google.com

Since my dimms are rated for 1.45v, I will run them at 1.45v!
View attachment 394440

You can run rfc way lower probably rcd 34 or 35 snd rp 30-32.

 

[Space Lynx]

I don't know if mine is M-Die or A-Die, but here it is... I was really hoping for better ns reading, 66 seems a little high? @ir_cow @TheLostSwede

 

[ir_cow]

@Space Lynx you have a X3D CPU. frequency and latency does not matter much at all.

 

[Cowboystrekk]

I don't know if mine is M-Die or A-Die, but here it is... I was really hoping for better ns reading, 66 seems a little high? @ir_cow @TheLostSwede

View attachment 394462

You can compare my timings vs yours Raise ypur refi to atleadt 50000. Try rfc 380. If it works you have A-die. If you can lower more than 470 you have M. You can probably run cl28 if you raise vdd to 1.45v+. You can probably lower rcd and rp to 35 or 34. Maybe you can turn gear down mode off. I bet those changes will get you close to 60ns.

 

[Space Lynx]

View attachment 394463

You can compare my timings vs yours Raise ypur refi to atleadt 50000. Try rfc 380. If it works you have A-die. If you can lower more than 470 you have M. You can probably run cl28 if you raise vdd to 1.45v+. You can probably lower rcd and rp to 35 or 34. Maybe you can turn gear down mode off. I bet those changes will get you close to 60ns.

i just did 1.35 to 1.36v for a little extra stability. not sure it matted. but 1.45v is a little too spicey for me. i'm alright with cl30, but i might consider your other changes, thanks!

 

[Dr. Dro]

View attachment 394463

You can compare my timings vs yours Raise ypur refi to atleadt 50000. Try rfc 380. If it works you have A-die. If you can lower more than 470 you have M. You can probably run cl28 if you raise vdd to 1.45v+. You can probably lower rcd and rp to 35 or 34. Maybe you can turn gear down mode off. I bet those changes will get you close to 60ns.

Mm, your AIDA64 is quite out of date, though. You should update it to get more accurate results. Not entirely sure 6.90 is even optimized for Raphael 3D chips, maybe, just maybe for the normal ones.

Protip: You can click save instead of busting out the ol' camera lol. Come out neat like this:

 

[Cowboystrekk]

Mm, your AIDA64 is quite out of date, though. You should update it to get more accurate results. Not entirely sure 6.90 is even optimized for Raphael 3D chips, maybe, just maybe for the normal ones.

Protip: You can click save instead of busting out the ol' camera lol. Come out neat like this:

View attachment 394464

I debloated my win11 which improved latency by 1ns, but it removed my screenshot tool, lol. I have cracked version of aida, hence cant upgrade I tested with latest and got very similar results.

 

[freeagent]

Anything more than 6200 1:1 is a bit dicey for me. I can almost nail down 6400 1:1 but still fails after a couple of hours. Kind of a bummer. 6200 is ok, not really much difference over 6000 though. So I just clamped down on tRFC gave it a lil extra voltage and that is it. I know this ram isn't G.Skill.. but I do think its a bit weak. 1.5v is routine over 6200 lol.. if I want decent performance as compared to others.

 

[VuurVOS]

View attachment 394451
This is a weird one. A-die dual rank 16gb. 8gb on each side of the sticks. They max out at 7400 so not as good as A die 16gb single rank.


You can run rfc way lower probably rcd 34 or 35 snd rp 30-32.

Lowered tRFC and tREFI without lossing stability. When I changed rcd and rp to 34, my system was vey unstable. Didn't try 35.
Rather a little bit more headroom than an unstable system during the summer season.

EDIT: Lost the ability to reboot after a long memory stress test without stability errors and long cooldown....

 

[Braegnok]

 

[Cowboystrekk]

View attachment 394527

Lowered tRFC and tREFI without lossing stability. When I changed rcd and rp to 34, my system was vey unstable. Didn't try 35.
Rather a little bit more headroom than an unstable system during the summer season.

EDIT: Lost the ability to reboot after a long memory stress test without stability errors and long cooldown....

That rdwr is very low. Try 15 or 16. Rcd and rp 35 probably works. Only rfc and refi are affected by heat to my knownledge.

 

[MiraeNN]

I ran tests with TestMem5 and Y-cruncher, and this turned out to be my most stable configuration. I couldn’t boot at CL30. When I tried lowering the tRDRDSCL value further, TestMem5 started throwing errors, so I didn’t push it anymore. These are my current settings. It could be better — Im open to your thoughts.

Patriot Viper Venom RGB DDR5 7400 CL36 2x16 A-die​

 

[Cowboystrekk]

View attachment 394586

I ran tests with TestMem5 and Y-cruncher, and this turned out to be my most stable configuration. I couldn’t boot at CL30. When I tried lowering the tRDRDSCL value further, TestMem5 started throwing errors, so I didn’t push it anymore. These are my current settings. It could be better — Im open to your thoughts.

Patriot Viper Venom RGB DDR5 7400 CL36 2x16 A-die​

I would raise voltage to 1.4 on vdd. Then you can probably run cl30. Rcd and rp can probably run at 35 or 36. Rfc at 390 should run fine.

 

[ir_cow]

Change your tREFI. 65535.

 

[ShrimpBrime]

Anything more than 6200 1:1 is a bit dicey for me. I can almost nail down 6400 1:1 but still fails after a couple of hours. Kind of a bummer. 6200 is ok, not really much difference over 6000 though. So I just clamped down on tRFC gave it a lil extra voltage and that is it. I know this ram isn't G.Skill.. but I do think its a bit weak. 1.5v is routine over 6200 lol.. if I want decent performance as compared to others.

A-die scales pretty good.
My now testing settings at.

1.60v
6800mt/s CL28-38-38
(still working on this, I never said 100% stable here, but it's close.)
My ddr5 samsung kit/s (b-die) are no where near this tight or responsive to v-dimm at this speed.

 

[freeagent]

Does VDDQ have to match VDD? I keep them matched but unsure for higher voltages..

 

[ShrimpBrime]

Does VDDQ have to match VDD? I keep them matched but unsure for higher voltages..

Yes, they are matched.