• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

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

Does VDDQ have to match VDD? I keep them matched but unsure for higher voltages..
You can set vddq below vdd. Mine runs at 1.49v vdd, 1.43v vddq and 1.25v vddio now.
 
I finally nailed down 6400, only needs 1.55v to do it lol..
 
I do have one on lol, load is 39c :D

Also running tCL28 now, same voltage :cool:
39c isn't bad a all. You'll start to error out in the low 40s if you bring the CAS down enough. tREFI will do that as well.
 
I do have one on lol, load is 39c :D

Also running tCL28 now, same voltage :cool:
doing these things. :)


Screenshot 2025-04-13 162056.png
 
Didn't know Intel was also making Ryzen processors for AM5 :D (see thread title). For comparison reasons, I think it is beter to keep it to AM5 only.

I wanted to example A-Die. (See thread title)
Which runs identical on AMD and Intel. (XMP doesn't differe between systems)
Aida64 isn't even accurate for comparison reasons.
See everything you run latency, it's 2ns off every time.
So what really can you compare memory with?
2 like AMD systems?
Arbitrary testing really, who has lowest latency?
Either way, AM5 comparisons here is nill so far.
 
Last edited by a moderator:
My System:
Ryzen 5 7500F
Asus Prime B650M-K (Bios version: 3222 Current)
XPG Lancer Blade 2x16GB 6000Mhz Ram (AX5U6000C3016G-DTLABBK)
GALAX GeForce RTX 4070 SUPER 1-Click OC 2X V2
High Power Performance GD 750w Gold PSU (HP1-J750GD- F12S)
1TB KC3000 M2 SSD
512 GB SAMSUNG 970 EVO M2 SSD
480 GB KINGSTON A400 SATA SSD

I have been dealing with the Kernel 41 (63) error since March 7 when I installed the PC, getting 0 bug check codes. The error happens completely randomly with the system under load or not. I don't get a blue screen error. I have also tried things that I think may be causing the problem:

Removed the current protected socket, no change in the result, the problem persists.
Ethernet cable was old and I replaced it, problem persists.
Opened case and checked cables, problem persists.
OCCT power stress test without problem.
Cinebench24 test without problem.
Prime95 test without problem.
Furmark test without problems.
GPU driver clean install with DDU.
Windows UEFI and Secure Boot enable selected in bios.
Fast Boot is made disable.
Memory Context Restore is made disable.

XMP/EXPO Profiles were tried. Memtest with XMP/EXPO turned off was fine, but with the profile turned on, the system locked up and the test could not be completed. I have also attached the logs where the freeze was experienced by taking ss with the relevant names.
The EXPO Profile setting was manually entered and overclocked, Memtest test was completed without any problems at 6000mhz 30/40/40/76/116 and 1.350 Volts.
Although I used the system for a while with the EXPO profile manually entered, Kernel 41 and booting problem occurred again. Therefore, the RAM frequencies were reset back to base clock.

I was getting the error probably because the RAMs were in EXPO/manual OC. Since April 10, the pc has been running stable at base clock. So I don't have any idea how I OC my RAMs without any risk. There is a video in Youtube says 7000 series are desyncronized with FCLK 2000 frequency and because of that it says set it 2033mhz. I haven't tried it yet. I'm open to your advice.
 

Attachments

  • 4800mhz base clock.jpg
    4800mhz base clock.jpg
    917.8 KB · Views: 50
  • 6000mhz manual.jpg
    6000mhz manual.jpg
    911.2 KB · Views: 51
  • c24 multicore.jpg
    c24 multicore.jpg
    504.7 KB · Views: 40
  • Cinebench2024 CPU Single-MultiCore & GPU Render Tests - max 71.5 Derece.jpg
    Cinebench2024 CPU Single-MultiCore & GPU Render Tests - max 71.5 Derece.jpg
    403.1 KB · Views: 40
  • furmark test result.jpg
    furmark test result.jpg
    41.6 KB · Views: 43
  • prime95 30min test.jpg
    prime95 30min test.jpg
    1 MB · Views: 61
  • MEMTEST donma olan logların donduğu noktalar.jpg
    MEMTEST donma olan logların donduğu noktalar.jpg
    867.4 KB · Views: 34
  • spd1.jpg
    spd1.jpg
    450.4 KB · Views: 33
  • spd2.jpg
    spd2.jpg
    471.6 KB · Views: 42
Last edited:
XPG Lancer Blade 2x16GB 6000Mhz Ram (AX5U6000C3016G-DTLABBK)
Now testing ram on stability i have exactly the same sticks and turns out they are M die not A die!
 
My System:
Ryzen 5 7500F
Asus Prime B650M-K (Bios version: 3222 Current)
XPG Lancer Blade 2x16GB 6000Mhz Ram (AX5U6000C3016G-DTLABBK)
GALAX GeForce RTX 4070 SUPER 1-Click OC 2X V2
High Power Performance GD 750w Gold PSU (HP1-J750GD- F12S)
1TB KC3000 M2 SSD
512 GB SAMSUNG 970 EVO M2 SSD
480 GB KINGSTON A400 SATA SSD

I have been dealing with the Kernel 41 (63) error since March 7 when I installed the PC, getting 0 bug check codes. The error happens completely randomly with the system under load or not. I don't get a blue screen error. I have also tried things that I think may be causing the problem:

Removed the current protected socket, no change in the result, the problem persists.
Ethernet cable was old and I replaced it, problem persists.
Opened case and checked cables, problem persists.
OCCT power stress test without problem.
Cinebench24 test without problem.
Prime95 test without problem.
Furmark test without problems.
GPU driver clean install with DDU.
Windows UEFI and Secure Boot enable selected in bios.
Fast Boot is made disable.
Memory Context Restore is made disable.

XMP/EXPO Profiles were tried. Memtest with XMP/EXPO turned off was fine, but with the profile turned on, the system locked up and the test could not be completed. I have also attached the logs where the freeze was experienced by taking ss with the relevant names.
The EXPO Profile setting was manually entered and overclocked, Memtest test was completed without any problems at 6000mhz 30/40/40/76/116 and 1.350 Volts.
Although I used the system for a while with the EXPO profile manually entered, Kernel 41 and booting problem occurred again. Therefore, the RAM frequencies were reset back to base clock.

I was getting the error probably because the RAMs were in EXPO/manual OC. Since April 10, the pc has been running stable at base clock. So I don't have any idea how I OC my RAMs without any risk. There is a video in Youtube says 7000 series are desyncronized with FCLK 2000 frequency and because of that it says set it 2033mhz. I haven't tried it yet. I'm open to your advice.
Gotta try things!!

Try CR 2T instead of 1T
Try CL32 instead of 30
Try disable power down mode
Try additional vdd/vddq (1.40v?)
Try a different memory divider (2800mhz)

Running out of ideas....
 
Now testing ram on stability i have exactly the same sticks and turns out they are M die not A die!
But how? They seems A-Die in Typoon Burner, at least mines seem like that.
 

Attachments

  • typhoon burner.jpg
    typhoon burner.jpg
    387.8 KB · Views: 35
But how? They seems A-Die in Typoon Burner, at least mines seem like that.
For me Typhoon Burner is not showing which is it exactly but when using tRFC below 480 it will not start.
 
Last edited:
My System:
Ryzen 5 7500F
Asus Prime B650M-K (Bios version: 3222 Current)
XPG Lancer Blade 2x16GB 6000Mhz Ram (AX5U6000C3016G-DTLABBK)
GALAX GeForce RTX 4070 SUPER 1-Click OC 2X V2
High Power Performance GD 750w Gold PSU (HP1-J750GD- F12S)
1TB KC3000 M2 SSD
512 GB SAMSUNG 970 EVO M2 SSD
480 GB KINGSTON A400 SATA SSD

I have been dealing with the Kernel 41 (63) error since March 7 when I installed the PC, getting 0 bug check codes. The error happens completely randomly with the system under load or not. I don't get a blue screen error. I have also tried things that I think may be causing the problem:

Removed the current protected socket, no change in the result, the problem persists.
Ethernet cable was old and I replaced it, problem persists.
Opened case and checked cables, problem persists.
OCCT power stress test without problem.
Cinebench24 test without problem.
Prime95 test without problem.
Furmark test without problems.
GPU driver clean install with DDU.
Windows UEFI and Secure Boot enable selected in bios.
Fast Boot is made disable.
Memory Context Restore is made disable.

XMP/EXPO Profiles were tried. Memtest with XMP/EXPO turned off was fine, but with the profile turned on, the system locked up and the test could not be completed. I have also attached the logs where the freeze was experienced by taking ss with the relevant names.
The EXPO Profile setting was manually entered and overclocked, Memtest test was completed without any problems at 6000mhz 30/40/40/76/116 and 1.350 Volts.
Although I used the system for a while with the EXPO profile manually entered, Kernel 41 and booting problem occurred again. Therefore, the RAM frequencies were reset back to base clock.

I was getting the error probably because the RAMs were in EXPO/manual OC. Since April 10, the pc has been running stable at base clock. So I don't have any idea how I OC my RAMs without any risk. There is a video in Youtube says 7000 series are desyncronized with FCLK 2000 frequency and because of that it says set it 2033mhz. I haven't tried it yet. I'm open to your advice.
Try testmem5 run as admin with usmus1 config when you use expo 6000.

For me Typhoon Burner is not showing which is it exactly but when using tRFC below 480 it will not start.
If you have A-die sub 400 on 6000 should always work. M-die struggles below 480 and can at best do 460-470.
 
Try testmem5 run as admin with usmus1 config when you use expo 6000.


If you have A-die sub 400 on 6000 should always work. M-die struggles below 480 and can at best do 460-470.
OK, I just realized that whether I enable EXPO or not my FCLK Frequency always stays Auto, this time I tried EXPO with 2033mhz. I left my pc to make a Memtest86 with EXPO1 Profile (30/40/40/76/1.35v), It has passed the Memtest without an error but when I tried Prime95 Large FFTS test I got another Kernel Power 41 Error. After I got Kernel again, I change the FCLK Frequency to 2000mhz and tried Prime95 Large FFTS test, there was no error. Also I tried testmem5 with usmus1 config, there is no error for now. Im gonna check it if it happens in load especially while gaming.
 

Attachments

  • TestMem5 usmus config test.jpg
    TestMem5 usmus config test.jpg
    582.4 KB · Views: 35
  • testmem usmus1 config 6 cycle no error expo1 6000mhz - fclk 2000mhz.jpg
    testmem usmus1 config 6 cycle no error expo1 6000mhz - fclk 2000mhz.jpg
    539 KB · Views: 41
Last edited:
For me Typhoon Burner is not showing which is it exactly but when using tRFC below 480 it will not start.
tRFC1 is 884 and tRFC2 is 480 on me I don't know which one you mention there. This is EXPO1 Profile btw. We have same brand and model Rams but as you said you have another version of Die chips.
 

Attachments

  • ZenTimings_Screenshot.png
    ZenTimings_Screenshot.png
    45.1 KB · Views: 91
Last edited:
I find that lowering tRFC really needs some voltage. With my sticks.. it needs about what I ran with DDR4 all the time lol.. 1.55v

Edit:

TM5.png
 
* Not Stable * not tested * just booted * with increased TRFC 1,2,3,4 + lowered TREFI to see if CL30 is bootable. CL28 refuses
I will test with my compiler if it crashes.

This is just to share - that my DRAM is DR. I'm dissapointed for the Zentimings software. My DRAM is suddenly in the newer Version DR, and not SR.

ZenTimings_Screenshot.png


edit: I have now a reason to finally update = compile from source code libreoffice and rust in gnu gentoo linux for "stable" test purposes. It looks promising. 29 packages in total. When my dram was unstable in past for am4 or am5, the gentoo gnu toolchain showed it the fastest.

that topic motivated me to go on with overclock. I plan to make hardware changes to the chassis as i run now open chassis with cpu tower fan, gpu fan and one disabled exhaust fan. I want to actively cool my DRAM.

edit: It crashed in the middle of compiles. Not stable. But it showed me TCL 30 is possible with other values. Webbrowsing works which is not my standard. It needs to be tmpfs / compiler stable.

edit: trefi to 2 ^ 15. compiling libreoffice while writing and waiting.

Memory voltage worse = +8 ° Degrees Celsius (i need to fix my semi passive setup first)

Code:
spd5118-i2c-1-51
Adapter: SMBus PIIX4 adapter port 0 at 0b00
temp1:        +48.0°C  (low  =  +0.0°C, high = +55.0°C)
                       (crit low =  +0.0°C, crit = +85.0°C)

Room Temperature floor and wall and top with contactless infrared thermometer 20.2 ° degrees Celsius (plus / minus 0.5°C)

Edit - compiling crashed -> i need to order proper pc case fans first.
 
Last edited:
1. Stock EXPO Profile 6000mhz cl30 2. Optimized timings 6000mhz cl30
Stock.png
OC.png


For some reason my system does not like when using CL28 tCL or 6200 MT/s frequency and voltage increase or decrease will not change anything at all!

After RAM timing optimization and cpu curve optimizer -35 +100mhz boost on cpu i got fps boost in Fallout 4 from 138fps to 152fps (10% fps boost at the same 6000MHz CL30 and same voltage 1.35v) (cpu optimization only gives maybe +1-2fps and rest of all is ram)
 
Last edited:
1. Stock EXPO Profile 6000mhz cl30 2. Optimized timings 6000mhz cl30
Stock.png
OC.png


For some reason my system does not like when using CL28 tCL or 6200 MT/s and frequency voltage increase or decrease will not change everything at all!

After RAM timing optimization and cpu curve optimizer -35 +100mhz boost on cpu i got fps boost in Fallout 4 from 138fps to 152fps (cpu optimization only gives maybe +1-2fps and rest of all is ram)
Can you do an Aida64 Ram benchmark too? I just wonder how much your latency is decreased.
 
Stock EXPO Profile
Untitled.png



Optimized ram timings
1.png
2.png
 
1. Stock EXPO Profile 6000mhz cl30 2. Optimized timings 6000mhz cl30
Stock.png


For some reason my system does not like when using CL28 tCL or 6200 MT/s frequency and voltage increase or decrease will not change anything at all!
Your FCLK is 2133MHz.. Probably the reason why 6200 doesn't work.
 
Back
Top