[cagrisezgen]

I think I found the reason. DFE Read Training = Auto/Enabled makes my ram unstable. I am still testing my 5800 Mhz but seems stable, i did force retraining couple times. Perhaps DFE Read does not work correctly on Asrock boards or specifically on mine, or the reason is in the memory itself
My settings update post above:
DFE Read Training = Disabled
AMD-->CBS-->DDR Options-->DDR Training Options
All Auto except:
(for faster training ~3:30 minutes)
RX Burst Lenght = 1x
TX Burst Lenght = 1x
other settings on the screen below
View attachment 2652739 View attachment 2652742

I'm trying to stabilize it with low voltage. It's not going bad for now, but do you think I'll have problems tightening the times?

 

[cagrisezgen]

I think I found the reason. DFE Read Training = Auto/Enabled makes my ram unstable. I am still testing my 5800 Mhz but seems stable, i did force retraining couple times. Perhaps DFE Read does not work correctly on Asrock boards or specifically on mine, or the reason is in the memory itself
My settings update post above:
DFE Read Training = Disabled
AMD-->CBS-->DDR Options-->DDR Training Options
All Auto except:
(for faster training ~3:30 minutes)
RX Burst Lenght = 1x
TX Burst Lenght = 1x
other settings on the screen below
View attachment 2652739 View attachment 2652742

I couldn't boot with this traninig options

edit....

I suppose it was a bug. I couldnt boot with default boot. After cmos reset, i boot with your new training options.

 

[DemonAk]

I'm trying to stabilize it with low voltage

y-cruncher FFT N63 VT3 pass? at least 2 hours? If it stable then you can try tight your timings of course, try to set first timings like 28-36-36, below that i'll think you get errors but I could be wrong. use calculator

 

[cagrisezgen]

y-cruncher FFT N63 VT3 pass? at least 2 hours? If it stable then you can try tight your timings of course, try to set first timings like 28-36-36, below that i'll think you get errors but I could be wrong. use calculator

I am using calculator. After y-cruncher, I will share with you.

 

[cagrisezgen]

I am completing "1usmus_v3" successfully. I just can't get past "SVT" on the Y-cruncher. Anyone have any ideas?

 

[lostking]

I am completing "1usmus_v3" successfully. I just can't get past "SVT" on the Y-cruncher. Anyone have any ideas?

View attachment 2652850

I would say tRCD is too tight. Keep it 36.

 

[cagrisezgen]

I am completing "1usmus_v3" successfully. I just can't get past "SVT" on the Y-cruncher. Anyone have any ideas?

View attachment 2652850

I handled with "VDDP"

 

[cagrisezgen]

I would say tRCD is too tight. Keep it 36.

I changed VDDP 0,95 to 1,01. I suppose it's ok. I will long test at night. If i take a problem, i will try tRCD.

 

[lostking]

I changed VDDP 0,95 to 1,01. I suppose it's ok. I will long test at night. If i take a problem, i will try tRCD.

Have you tested 5600 with JEDEC timings? I would verify if the data rate is working first.

 

[DemonAk]

I changed VDDP 0,95 to 1,01. I suppose it's ok. I will long test at night. If i take a problem, i will try tRCD.

Too low SOC or VDDIO, VDDQ, VDD if you follow the delta VDDP for 1.18 SOC = 1.03v (1.18v-0.15v)
For example, I have errors with VDDIO on 1.35v, only 1.4 works fine
For tCL 28 perhaps too low VDD

I am completing "1usmus_v3" successfully. I just can't get past "SVT" on the Y-cruncher. Anyone have any ideas?

You Pass y-cruncher with only first timings: 28-34-34-34-46-80?, set other timings to auto

 

[cagrisezgen]

Have you tested 5600 with JEDEC timings? I would verify if the data rate is working first.

No I didn't. I will try. Thank you.

 

[cagrisezgen]

Too low SOC or VDDIO, VDDQ, VDD if you follow the delta VDDP for 1.18 SOC = 1.03v (1.18v-0.15v)
For example, I have errors with VDDIO on 1.35v, only 1.4 works fine
For tCL 28 perhaps too low VDD

You Pass y-cruncher with only first timings: 28-34-34-34-46-80?, set other timings to auto

I used delta 0.17 and take a short test and it's ok. I will long test at night

 

[cagrisezgen]

I was able to go a little further, but I'm still getting errors. Could there be an RTT or resistance (ohm) problem other than voltage? If I don't get an error in TM5, does that mean that there is no error in the resistance or Rtt values? @DemonAk @lostking

 

[DemonAk]

I was able to go a little further, but I'm still getting errors. Could there be an RTT or resistance (ohm) problem other than voltage? If I don't get an error in TM5, does that mean that there is no error in the resistance or Rtt values?

Unfortunately, TM5 with config 1usmus does not give a 100% guarantee that everything will be stable (3 circles 1usmus is too little, you need at least 20-30, edit config line Cycles=3), you can also test with a heavier config extreme from Anta download link: bit.ly/2MUvl6n and even he won’t give 100% so you need to test using other programs such as y-cruncher, karhu, HCI MemTest, prime95 large fft and heavy games like Cyberpunk 2077, BF2042.
You can try Rtt like 5-5-2-5-5 or 5-5-3-5-5 and DramDqDs 40, 48 but I think the reason is the lack of voltage, increase SOC to 1.2 or 1.22, set equal VDDIO=VDD=VDDQ=1.36 and test again if gets errors then increase again to 1,37v etc

 

[cagrisezgen]

Unfortunately, TM5 with config 1usmus does not give a 100% guarantee that everything will be stable (3 circles 1usmus is too little, you need at least 20-30, edit config line Cycles=3), you can also test with a heavier config extreme from Anta download link: bit.ly/2MUvl6n and even he won’t give 100% so you need to test using other programs such as y-cruncher, karhu, HCI MemTest, prime95 large fft and heavy games like Cyberpunk 2077, BF2042.
You can try Rtt like 5-5-2-5-5 or 5-5-3-5-5 and DramDqDs 40, 48 but I think the reason is the lack of voltage, increase SOC to 1.2 or 1.22, set equal VDDIO=VDD=VDDQ=1.36 and test again if gets errors then increase again to 1,37v etc

I think I'll take a break from this because it's starting to get tiring. There is @5600, 1.4V that you shared, which I am currently using stably. I will continue to use it. The reason why I did not use the voltages equally is that I calculated them using "Delta VDD & VDDQ in mVolt" in "DDR5 Timing Calculator for AMD".

 

[cagrisezgen]

Hello @DemonAk and @lostking .@5600 I brought my times JEDEC value. I started testing starting from low VDDIO, VDD, VDDQ and VSOC, Vddp values. I managed to surpass VT3 even at very small values. As I increased my voltage values, I started to exceed VT3. Finally, I increased it to VDDIO=VDD=VDDQ=1.4 and VSoc=1.22(delta=0.15), but I can only pass 7 tests. Will some changes I make on DDR Training Options and Nitro mode have an effect on VT3? Or is there anything you can suggest me to pass VT3? Thank you.

When I turn on Nitro mode, VT3 fails in the first test. Can we say that it cannot be used below @6000?

DFE Read Training = Disabled
AMD-->CBS-->DDR Options-->DDR Training Options
All Auto except:
RX Burst Lenght = 1x
TX Burst Lenght = 1x
Nitro mode = Disabled

 

[lostking]

Hello @DemonAk and @lostking .@5600 I brought my times JEDEC value. I started testing starting from low VDDIO, VDD, VDDQ and VSOC, Vddp values. I managed to surpass VT3 even at very small values. As I increased my voltage values, I started to exceed VT3. Finally, I increased it to VDDIO=VDD=VDDQ=1.4 and VSoc=1.22(delta=0.15), but I can only pass 7 tests. Will some changes I make on DDR Training Options and Nitro mode have an effect on VT3? Or is there anything you can suggest me to pass VT3? Thank you.

DFE Read Training = Disabled
AMD-->CBS-->DDR Options-->DDR Training Options
All Auto except:
(for faster training ~3:30 minutes)
RX Burst Lenght = 1x
TX Burst Lenght = 1x
Nitro mode = Disabled

View attachment 2654223

These settings below work for me. For your reference
Nitro: disabled, it is somewhat more stable than enabled.
IMC Tweaks: Enable Read DFE, 1 step, 1 step, Enable Write DFE, 1 step, 1 step, 1 tap, 1 tap.

In the past, 128GB loves low VDDQ, the sweet spot was around 1.27v. I am not sure if it is still the case with the latest bios. You can leave VDDQ=VDDIO=1.27v, and VDD=1.4v for tight timings.
VDDP. I use 0.95v. Higher value seems to have negative effect on the stability.

Is your kit Hynix or Micron.
Hynix can do 32-36-36 1.2VDD at 5600; or 30-38-38 1.35VDD at 6000.

 

[cagrisezgen]

These settings below work for me. For your reference
Nitro: disabled, it is somewhat more stable than enabled.
IMC Tweaks: Enable Read DFE, 1 step, 1 step, Enable Write DFE, 1 step, 1 step, 1 tap, 1 tap.

In the past, 128GB loves low VDDQ, the sweet spot was around 1.27v. I am not sure if it is still the case with the latest bios. You can leave VDDQ=VDDIO=1.27v, and VDD=1.4v for tight timings.
VDDP. I use 0.95v. Higher value seems to have negative effect on the stability.

Is your kit Hynix or Micron.
Hynix can do 32-36-36 1.2VDD at 5600; or 30-38-38 1.35VDD at 6000.

I will try your voltage values and training values. I did the training values as follows and for the first time I can go above the 10th test in VT3 and the test is still running.

DFE Read Training = Enabled
TX DFE Taps = 2 taps
RX2D voltage step size (2^n) = 2
TX2D voltage step size (2^n) = 2
RX DFE Taps = 1 taps
RX Burst Lenght = 1x
TX Burst Lenght = 1x

 

[cagrisezgen]

These settings below work for me. For your reference
Nitro: disabled, it is somewhat more stable than enabled.
IMC Tweaks: Enable Read DFE, 1 step, 1 step, Enable Write DFE, 1 step, 1 step, 1 tap, 1 tap.

In the past, 128GB loves low VDDQ, the sweet spot was around 1.27v. I am not sure if it is still the case with the latest bios. You can leave VDDQ=VDDIO=1.27v, and VDD=1.4v for tight timings.
VDDP. I use 0.95v. Higher value seems to have negative effect on the stability.

Is your kit Hynix or Micron.
Hynix can do 32-36-36 1.2VDD at 5600; or 30-38-38 1.35VDD at 6000.

I couldn't boot with VDDQ=VDDIO=1.27v

 

[DemonAk]

DFE Read Training = Disabled
AMD-->CBS-->DDR Options-->DDR Training Options
All Auto except:
RX Burst Lenght = 1x
TX Burst Lenght = 1x
Nitro mode = Disabled

It looks like I was wrong about disabling DFE because if you disable it even at stock settings at 3600, the memory will become unstable. also it looks like we need a full memory training, i.e. about 9 minutes, so try these values
DFE Read Training = auto

AMD-->CBS-->DDR Options-->DDR Training Options

All Auto or tx rx dfe taps 4 or you can try 2 or 3
you need to look where there will be no errors or there will be few


Nitro mode = enabled by default