Overclock.net banner
24,661 - 24,680 of 31,499 Posts
Hi,
Nice to meet you. I came across your forum while looking for information on the best timings for DDR5 RAM on AM5, and it seemed natural to reach out to you.
I'm not an expert on timings, but I have a decent understanding of how PCs work and their BIOS configurations. I would be happy to share my current timings and perhaps ask for your advice on how to optimize them.
Currently, my configuration is:
  • 9800X3D
  • X870E Aorus Master
  • Corsair DDR5 6000 MHz CL28
  • 7900 XTX
The data I have gathered so far is as follows.
I look forward to your help, and thank you in advance!
Congrats for your system!!!

You should set these 2 DDR5 rules:
tFAW = 4x tRRDS
tRC=tRAS+tRP

Additionally, RAM Oc'ing demands loads of patience, testing and precaution regarding system temps.
If your are not running on water, your system should be adequately vented, specially over CPU and RAM sticks.
A case with a good airflow is advisable.
If you look backwards in this thread, you'll find in recent pages some rigs that have 64gb ram sticks like yours.
 
How common is it for a Hynix A-die kit to not be able to hit 8000? I have two kits, Team 7000 and Gskill 8000, both 32GB SR A-die, tested with same CPU (7800X3D) and mobo (x870 tomahawk). The Gskill kit easily does 8000, no work with expo, even tightened a few timings, and passes 6+ hours of karhu and a few hours of VT3, no errors. The Team kit took me hours of tuning testing voltages/impedance with very loose timings to prevent it from quickly erroring in Karhu, and longest pass before error was around 3 hours with a few fails quicker than that in VT3. Temp isn't an issue, fan on rams. What conclusion could I draw other than the Team kit is a dud? I've been told it's a "skill issue" lol. This is a 2022 kit I got in 2023 when Newegg marked down a lot of their 32gb kits to $54. Maybe some poor bins on clearance? Was going crazy with this Team kit but the Gskill felt like easy mode.
 
Any clues on what I could change, kinda stuck atm. After 6 days and a lot of no posting and no back to back stability I've finally gotten this kit/board to post and test reliably back to back. Settings in the following SS (vddio at 1.43v set) keep failing at just under 8 hours.

I've tried:

  • VDD and or VDDQ +10mv steps, anything above and below 1.53/1.43 set in bios seems to be less stable or picks up an error around 1hr
  • VDDIO 1.4 to 1.45, doesn't seem to change anything
  • RTT and Termination settings are fairly locked in from what I can tell, post to post was wildly unstable until setting DramDqDs to 40 ohms, 48 ohms is back to large post to post variance.
  • Timings aren't overly tight yet as I didn't want to mess with that or GDM off until actually getting this stable.
  • VDDGs are set to auto atm, may be to high (board defaulting to 1050 mv); can this cause instability for memory clocks?
  • Temps peak around 49c and sit around 48c most of the test
  • Nitro currently at 2/3/1 8x8x and tphy sync'd with pointintv set to 2 (down from 3 default)

Number Font Software Technology Screenshot


So tired and I want to be done with this, pc has literally be running karhu for almost 6 days straight now:LOL:, my sanity is dwindling.
 
How common is it for a Hynix A-die kit to not be able to hit 8000? I have two kits, Team 7000 and Gskill 8000, both 32GB SR A-die, tested with same CPU (7800X3D) and mobo (x870 tomahawk). The Gskill kit easily does 8000, no work with expo, even tightened a few timings, and passes 6+ hours of karhu and a few hours of VT3, no errors. The Team kit took me hours of tuning testing voltages/impedance with very loose timings to prevent it from quickly erroring in Karhu, and longest pass before error was around 3 hours with a few fails quicker than that in VT3. Temp isn't an issue, fan on rams. What conclusion could I draw other than the Team kit is a dud? I've been told it's a "skill issue" lol. This is a 2022 kit I got in 2023 when Newegg marked down a lot of their 32gb kits to $54. Maybe some poor bins on clearance? Was going crazy with this Team kit but the Gskill felt like easy mode.
CPU could be the cause
 
Any clues on what I could change, kinda stuck atm. After 6 days and a lot of no posting and no back to back stability I've finally gotten this kit/board to post and test reliably back to back. Settings in the following SS (vddio at 1.43v set) keep failing at just under 8 hours.

I've tried:

  • VDD and or VDDQ +10mv steps, anything above and below 1.53/1.43 set in bios seems to be less stable or picks up an error around 1hr
  • VDDIO 1.4 to 1.45, doesn't seem to change anything
  • RTT and Termination settings are fairly locked in from what I can tell, post to post was wildly unstable until setting DramDqDs to 40 ohms, 48 ohms is back to large post to post variance.
  • Timings aren't overly tight yet as I didn't want to mess with that or GDM off until actually getting this stable.
  • VDDGs are set to auto atm, may be to high (board defaulting to 1050 mv); can this cause instability for memory clocks?
  • Temps peak around 49c and sit around 48c most of the test
  • Nitro currently at 2/3/1 8x8x and tphy sync'd with pointintv set to 2 (down from 3 default)

View attachment 2692017

So tired and I want to be done with this, pc has literally be running karhu for almost 6 days straight now:LOL:, my sanity is dwindling.
Welcome to Asus bioses (y)

For me it started with AGESA 1202B bioses, i lost my stable 6400 and had to settle for 6200
 
So tired and I want to be done with this, pc has literally be running karhu for almost 6 days straight now:LOL:, my sanity is dwindling.
Have you changed ProcOdt Pu to 40? I would set DramDqDs back to 34.3 and check. You have it in the submenu
 
Managed to improve 6400 stability (still unstable tho) by changing powering settings on my bad IMC 9800X3D, maybe my settings will help you if you're also using the Gene:
Font Number Screenshot

Notes:
VDDG defaults to 1050/1050, lowering it to 950 may help
ProcDqDs is actually 40ohm
Increasing VDDP did jack ****
VDDIO had very little effect, tested 1.25, 1.35, 1.4. 1.439. Best was 1.4
Playing with DramDqDs or RTTs just made things worse
Highly recommend leaving pointer interval last
 
Have you changed ProcOdt Pu to 40? I would set DramDqDs back to 34.3 and check. You have it in the submenu
At one point I had ProcOdt PU at 40, didnt seem to make a difference. DramDqDs at 34 or 48 is wildly unstable/post to post variance > could fail at one hour or 9 seconds in karhu.

I will trill ProcOdt at 40 again, but my gut is telling me its something else if it’s failing late into testing.
 
Any clues on what I could change, kinda stuck atm. After 6 days and a lot of no posting and no back to back stability I've finally gotten this kit/board to post and test reliably back to back. Settings in the following SS (vddio at 1.43v set) keep failing at just under 8 hours.

I've tried:

  • VDD and or VDDQ +10mv steps, anything above and below 1.53/1.43 set in bios seems to be less stable or picks up an error around 1hr
  • VDDIO 1.4 to 1.45, doesn't seem to change anything
  • RTT and Termination settings are fairly locked in from what I can tell, post to post was wildly unstable until setting DramDqDs to 40 ohms, 48 ohms is back to large post to post variance.
  • Timings aren't overly tight yet as I didn't want to mess with that or GDM off until actually getting this stable.
  • VDDGs are set to auto atm, may be to high (board defaulting to 1050 mv); can this cause instability for memory clocks?
  • Temps peak around 49c and sit around 48c most of the test
  • Nitro currently at 2/3/1 8x8x and tphy sync'd with pointintv set to 2 (down from 3 default)

View attachment 2692017

So tired and I want to be done with this, pc has literally be running karhu for almost 6 days straight now:LOL:, my sanity is dwindling.
Try Auto pointer and maybe lower VDDG. Use iGPU to speed up testing
 
I will trill ProcOdt at 40 again, but my gut is telling me its something else if it’s failing late into testing.
Did you reduce VSOC? For IF 2000 it's a bit high.
 
At one point I had ProcOdt PU at 40, didnt seem to make a difference. DramDqDs at 34 or 48 is wildly unstable/post to post variance > could fail at one hour or 9 seconds in karhu.

I will trill ProcOdt at 40 again, but my gut is telling me its something else if it’s failing late into testing.
Same as me at 6400 but i have a B650TUF, my record for faling in karhu was 8 sec and longest was 9 hours, at same settings but a reboot
Nothing fixed it and i tested everything, messing with RTTs, DQs and ODTs just made it worse.

and before 1202b this was tested 100% stable.
 
Did you reduce VSOC? For IF 2000 it's a bit high.
Haven’t reduced it again yet, been testing with 1.15 since attempting 8000 again, isolating only adjusting memory atm.

Same settings as previous ss were 12hr+ stable with flck vdic mode pref set to predictive in amd oc menu at 1.5/1.42 vdd/vddq, but that setting adds a massive latency penalty (4-5ns).

*Going to drop vsoc and vddgs and re-run.
 
Even if one kit passes and another fails?
You had it right. The TeamGroup memory kit is a dud and the G.SKILL memory kit is not. There are lots of crappy Hynix DDR5 kits from G.SKILL and TeamGroup and Corsair and others. That happens to be one of them. Silicon lottery applies to CPU, RAM, GPUs and motherboards. Sometimes certain samples are just trash.
 
Any clues on what I could change, kinda stuck atm. After 6 days and a lot of no posting and no back to back stability I've finally gotten this kit/board to post and test reliably back to back. Settings in the following SS (vddio at 1.43v set) keep failing at just under 8 hours.

I've tried:

  • VDD and or VDDQ +10mv steps, anything above and below 1.53/1.43 set in bios seems to be less stable or picks up an error around 1hr
  • VDDIO 1.4 to 1.45, doesn't seem to change anything
  • RTT and Termination settings are fairly locked in from what I can tell, post to post was wildly unstable until setting DramDqDs to 40 ohms, 48 ohms is back to large post to post variance.
  • Timings aren't overly tight yet as I didn't want to mess with that or GDM off until actually getting this stable.
  • VDDGs are set to auto atm, may be to high (board defaulting to 1050 mv); can this cause instability for memory clocks?
  • Temps peak around 49c and sit around 48c most of the test
  • Nitro currently at 2/3/1 8x8x and tphy sync'd with pointintv set to 2 (down from 3 default)

View attachment 2692017

So tired and I want to be done with this, pc has literally be running karhu for almost 6 days straight now:LOL:, my sanity is dwindling.
For Asus X870 4 dimm mobo, I also found if vdd over 1.55 is unstable too, I can stabilise in karhu 500K% 8000C34, vdd higher than 1.55 can cause error with same timing, so I am stucked
 
Discussion starter · #24,678 · (Edited)
You had it right. The TeamGroup memory kit is a dud and the G.SKILL memory kit is not. There are lots of crappy Hynix DDR5 kits from G.SKILL and TeamGroup and Corsair and others. That happens to be one of them. Silicon lottery applies to CPU, RAM, GPUs and motherboards. Sometimes certain samples are just trash.
My first TeamGroup XTreem 8000 2x16GB was a dud, but second same kit from Newegg quite good. :)

Number Screenshot Software
 
My first TeamGroup XTreem 8000 2x16GB was a dud, but second same kit from Newegg quite good. :)

View attachment 2692034
Yes that has happened to me several times. I have purchased probably 15 or 16 DDR5 memory kits and 4 were good enough to keep. The rest were returned for refund. A couple of times my first kit was great so I ordered the same part number and the second kit was trash.
 
  • Rep+
Reactions: Braegnok
Anyone had any experience with B850 tomahawk yet? Since I can get one mobo for free, I've been looking between live mixer, riptide, mortar and tomahawk. Tomahawk seems like a good choice since the rest have their own little annoyances. It also has my memory on qvl, but that won't help with my potato 7800x3d :p
 
24,661 - 24,680 of 31,499 Posts