ASUS ROG X570 Crosshair VIII Overclocking & Discussion Thread

BambusBlaster said:

Hi Guys,

I am having the suspicion that a high DRAM voltage 1.5v may be responsible for killing the IMC on the CPU. My understanding with DDR4 is, that you can crank up the voltage but you need to have proper air- or watercooling for the RAM. I have a 80mm fan and even tinkered a small casing, so that my RAM is properly cooled. Memory temperature never went above 32°C with stock settings for 16-16-16-36, 3600MHz @ 1.35V.

My components:

• Asus Crosshair VIII Hero (BIOS Version 3402)
• Ryzen 7 5800X
• 2x G.Skill Trident Z Royal 16 GB DDR4-3600 DIMM CL16
• Custom Water Loop (5x 140mm)

Spoiler: rest of post

Since the system was stable for 3 month I wanted to start overclocking the RAM. All the settings are on AUTO except for:

• AI Overclock Tuner: Manual
• Memory Frequency: DDR4-3600MHz
• FCLK Frequency: 1800MHz

• DRAM Voltage: 1.35
• VTTDDR Voltage: 0.675

I changed the Memory Frequency to DDR-4000 and lowered the FCLK to 1600MHz to make sure it is just the RAM that is being pushed to the limits. I upped the DRAM Voltage to "1.4" and VTTDDR to "0.7". The changes have been saved, computer attempts to reboot and BAM goes straight to RAM training, resets itself three times and halts after POST "your computer has been started in safe mode, press F1 to go to BIOS". OK, maybe the timings are way to tight, I will loosen them.

I changed them from "16-16-16-16-36-52" to "18-18-18-18-42-60"

Again, computer restarts three times in a row and stops right after POST. "Press F1 to go to BIOS".

Ok, maybe more voltage needed for 4000Mhz. So I upped the DRAM voltage to 1.45 and VTTDDR to 0.725.
Still no success. Loosening the timings even more?! Ok, lets give it a try
20-20-20-48-70, just to see whats happening.
Stuck at POST...
Maybe 4000MHz requires much more Voltage than I thought. I changed the DRAM Voltage to 1.5 and VTTDDR to 0.75
Still nothing. RAM training, computer started in safe mode once again and asks to press F1 to amend the settings.
Ok I will change everything back to what was working fine: FCLK 1800MHz, DRAM 3600MHz, 16-16-16-16-36-52, DRAM Voltage 1.35 and VTTDDR to 0.675.

COMPUTER IS GETTING STUCK IN RAM TRAINING AND REBOOTS THREE TIMES IN A ROW! Oh no! OH NO! Why? what?! how?!?

I cannot run the computer with the settings that have been stable for three month. I booted into Windows and AIDA stress test almost instantly shows an error message that the memory is unstable.
Now I have to run FCLK 1600 and DRAM at 3200, everything above that no matter what, will not run an AIDA Memory Stress Test. Error message within two minutes...

I thought I fried my memory. So I tried another couple of RAM and switched the 2x 16 GB against 4x 8GB (G.Skill Trident Z 16 GB DDR4-3600 DIMM CL15 Dual Kit (F4-3600C15D-16GTZ))
The problem remains the same. The computer wont do a proper POST when I set the Memory Frequency above 3200 and the FCLK Frequency above 1600.

Even a BIOS update to the most recent beta bios 3501 did not help or change anything.

I remember from the x58 platform, that there is a problem with the QPI/VTT and DRAM Voltage. It has to be within 0.5 range, otherwise it floods the gates of the CPU and kills it instantly.

Does something similar apply to a Ryzen CPU as well?
Was I wrong about the temperature and high DRAM voltage kills the IMC instantly, no matter how good the memory/CPU is being cooled?
Is there something I did not know when upping the voltage that much?
Have you ever come across this problem?

I appreciate any feedback or sugestions what to do, what to try. In case you need more info, just let me know and I can run some more tests or post screenshots.

Many thanks in advance

Click to expand...

rbys said:

So the last BIOS update is over a month old. Not sure what is taking so long.

Click to expand...

BambusBlaster said:

Hi Guys,

I am having the suspicion that a high DRAM voltage 1.5v may be responsible for killing the IMC on the CPU. My understanding with DDR4 is, that you can crank up the voltage but you need to have proper air- or watercooling for the RAM. I have a 80mm fan and even tinkered a small casing, so that my RAM is properly cooled. Memory temperature never went above 32°C with stock settings for 16-16-16-36, 3600MHz @ 1.35V.

My components:

• Asus Crosshair VIII Hero (BIOS Version 3402)
• Ryzen 7 5800X
• 2x G.Skill Trident Z Royal 16 GB DDR4-3600 DIMM CL16
• Custom Water Loop (5x 140mm)

Since the system was stable for 3 month I wanted to start overclocking the RAM. All the settings are on AUTO except for:

• AI Overclock Tuner: Manual
• Memory Frequency: DDR4-3600MHz
• FCLK Frequency: 1800MHz

• DRAM Voltage: 1.35
• VTTDDR Voltage: 0.675

I changed the Memory Frequency to DDR-4000 and lowered the FCLK to 1600MHz to make sure it is just the RAM that is being pushed to the limits. I upped the DRAM Voltage to "1.4" and VTTDDR to "0.7". The changes have been saved, computer attempts to reboot and BAM goes straight to RAM training, resets itself three times and halts after POST "your computer has been started in safe mode, press F1 to go to BIOS". OK, maybe the timings are way to tight, I will loosen them.

I changed them from "16-16-16-16-36-52" to "18-18-18-18-42-60"

Again, computer restarts three times in a row and stops right after POST. "Press F1 to go to BIOS".

Ok, maybe more voltage needed for 4000Mhz. So I upped the DRAM voltage to 1.45 and VTTDDR to 0.725.
Still no success. Loosening the timings even more?! Ok, lets give it a try
20-20-20-48-70, just to see whats happening.
Stuck at POST...
Maybe 4000MHz requires much more Voltage than I thought. I changed the DRAM Voltage to 1.5 and VTTDDR to 0.75
Still nothing. RAM training, computer started in safe mode once again and asks to press F1 to amend the settings.
Ok I will change everything back to what was working fine: FCLK 1800MHz, DRAM 3600MHz, 16-16-16-16-36-52, DRAM Voltage 1.35 and VTTDDR to 0.675.

COMPUTER IS GETTING STUCK IN RAM TRAINING AND REBOOTS THREE TIMES IN A ROW! Oh no! OH NO! Why? what?! how?!?

I cannot run the computer with the settings that have been stable for three month. I booted into Windows and AIDA stress test almost instantly shows an error message that the memory is unstable.
Now I have to run FCLK 1600 and DRAM at 3200, everything above that no matter what, will not run an AIDA Memory Stress Test. Error message within two minutes...

I thought I fried my memory. So I tried another couple of RAM and switched the 2x 16 GB against 4x 8GB (G.Skill Trident Z 16 GB DDR4-3600 DIMM CL15 Dual Kit (F4-3600C15D-16GTZ))
The problem remains the same. The computer wont do a proper POST when I set the Memory Frequency above 3200 and the FCLK Frequency above 1600.

Even a BIOS update to the most recent beta bios 3501 did not help or change anything.

I remember from the x58 platform, that there is a problem with the QPI/VTT and DRAM Voltage. It has to be within 0.5 range, otherwise it floods the gates of the CPU and kills it instantly.

Does something similar apply to a Ryzen CPU as well?
Was I wrong about the temperature and high DRAM voltage kills the IMC instantly, no matter how good the memory/CPU is being cooled?
Is there something I did not know when upping the voltage that much?
Have you ever come across this problem?

I appreciate any feedback or sugestions what to do, what to try. In case you need more info, just let me know and I can run some more tests or post screenshots.

Many thanks in advance

Click to expand...

stimpy88 said:

Do any of you guys have a problem with Windows saying a critical file is corrupted at boot? I keep getting it, and cant figure out why it happens. I run every memory test I can find, it passes overnight testing with no errors. If I simply reboot the computer, it will boot Windows fine, but I might have to do it a few times.

It started a few days ago, and I've made no changes to the computer. I have just done a clean install of Windows, and still the problem randomly shows up on boot. The computer is totally fine once it's in Windows, and runs for days without issue, this only happens if I have to boot/reboot.

Any ideas?

Click to expand...

rbys said:

1.2.0.2 beta

Click to expand...

xeizo said:

If I have VDIMM above 1.35V I need to have VTT over 0.7V to boot and be stable. Been running 1.38/0.725V since Zen 2 was new. Haven't noticed any regression of performance. I don't think X58 wisdom fully applies to Ryzen.

Click to expand...

Kokin said:

It can't have been related to temperature. I'm on a custom watercooled setup and have no active cooling on my RAM. Zero issues with my RAM at 3600MHz 16-15-15 @ 1.4v or 3733 16-15-15 @1.47V. It idles around 35-40C and will reach 55-60C during load. Similar G.Skill TridentZ b-die RAM but 2x16GB and 3200MHZ CL14 overclocked higher.

If you haven't already, try setting optimized defaults in BIOS, boot up, and then manually tweak everything again. Also try leaving VTTDDR to "Auto" for now.

Click to expand...

Sleepycat said:

VTTDDR is not the only voltage you need to adjust when overclocking your memory. VTT SOC, CLDO VDDP, VDDG CCD and VDDG IOD should not be left on Auto if you are overclocking and adjusting other voltages. If you can get into bios, set
VTT SOC: 1.09V
CLDO VDDP: 0.98V
VDDG CCD 1.05V
VDDG IOD 1.05V

Hopefully this gets you back to your stable settings again.

Click to expand...

Sleepycat said:

Can you post a Zentimings screenshot here? Also, try setting tRCDRD to 15.

Click to expand...

BambusBlaster said:

just changed it...

here you go

View attachment 2490542

I read a lot here and I remember that the Ryzen CPU seems to favour even before odd timings. Or is this tRCDRD 15 just to give it some "freedom"?

Click to expand...

CyrIng said:

Although I have a 3950X, it's very odd b/c it looks like the same troubles I had with my dual G.Skill F4-3600C16D-32GTZN

Whenever I had tried any BIOS release of versions of 3XXX, system encountered lots of Linux Machine Checks or Windows WHEA. DRAM OC was only stable only between 3200 and 3533 MHz, despite any Voltages tunings.

Back to 2206. So far very stable, even with a slight DDR OC to 3666 MHz at recommended 1.35V

View attachment 2490550

Can it be that some DRAM compatibility tables not being ported among BIOS versions ?

Click to expand...

jomama22 said:

So wanted to do some testing with new memory timings and PBO settings on the 5950x. Prior, PBO would lag slightly behind a strong all core oc in sottr "cpu game" results but I have figured out a way to boost them quite a bit.

All core 4.825 all-core oc:
View attachment 2490573

New PBO settings:
View attachment 2490574

New memory settings used for both:
View attachment 2490575

New PBO setting also improve things like r20 and such scores but havnt gotten around to it atm.

Click to expand...

J7SC said:

...nice ! Do you know by any chance what the CPU package peak power consumption was for both all-c 4.825 as well as updated PBO ? Temps ?

Click to expand...

jomama22 said:

Couldn't have been all that high as I have dynamic oc switcher set at 75A crossover and it doesn't trigger. EDC was around only 30-40 amps or so. Temps are very low because of that, ccd1 @ 50 or so and ccd2 @ 35 or so.

I genuinely have never really paid much attention to total package power but it's surly over 250w when going full bore through an all-core r20 run or somthing when doing 4.825 @ 1.33v. Temps for that are right around 80* for ccd1 and 75* for ccd2. Ambiant around 21*.

4.8 @ 1.3v nets around 77* for ccd1 and 72* for ccd2.

Click to expand...