cancel
Showing results for 
Search instead for 
Did you mean: 

x670E and 128GB of ram

sbloom1967
Level 8

Hi All,

I am running a x670 Hero , 7950XRD processor and 4 sticks of 32GB 5600 Dominator ram and bios 925.  I currently have XMP disabled and have the AI tuner to auto but I set the frequency to ddr5-3600.  With this set is somewhat ok, it still reboots once or twice a day which is really not that good.  Is there someway I can turn up my ram to the proper speed and make everything even more stable or how can i just get everything stable using the parts I have,  Thanks Steven

5,017 Views
1 ACCEPTED SOLUTION

Accepted Solutions

SkyBeam
Level 11

I am afraid there might be no stable solution. You should always check the memory QVL before. The one for your X670E Hero can be found here. Note that there is NO configuration with 4x32GB certified as of now. Yes it's a pity. There are also still some memory controller issues with AMD platforms with multiple modules per channel. The only 4-stick configurations which contain some certified modules is 4x8GB.

So having said that you chose to go for an unsupported memory configuration - which is fine, I have chosen that route too btw.
But just to be clear here you will be on your own and might have to figure out your specific configurations yourself. I did describe some success with my 4x16GB DDR5-6000 configuration (on a different board) here. This might give some hints.

However 4x32GB is putting even more load on the memory controller and VRM than my 4x16GB configuration. Some recommendations I can give:

  • Try first to get JEDEC timings (DDR5-3600 or DDR5-4800) stable
  • Try loading a custom preset matching you memory chips from AI Tweaker / DRAM Timing Control menu
  • Crank up your MC voltage to 1.4V and increase by 0.025 increments until you might reach a stable state
  • Run a full memtest86+ run after each modification checking for reported errors - keep it running for at least 1 hour or a complete run (all 10 tests pass)
  • If any errors reported by memtest86+ try adjusting timings (higher numbers) and tweak voltages

Exceeding safe voltage limits might cause permanent hardware damage. As said you have to do this on your own risk.

The fact you are reporting occasional reboots is a clear indicator your memory configuration currently is not stable. I am sure if you run memtest86+ in your current setup you will find a large amount of issues reported. Your system might still be able to boot into Windows just fine and you might experience occasional application crashes "only". That does not mean your system is stable! Sure you can blame browser tab crashes or occasional bluescreens on Windows or application faults but the truth is potentially garbled data in your memory. So absolutely make sure your memory configuration is stable by TESTING it (memtest86+).

And no, I do not believe you can go the full DDR5-5600 configuration with 4 sticks of this size. So you might be lucky already getting this configuration to run stable at DDR5-4800 JEDEC timings. Sometimes more memory is more important than fast memory. That's why I am running 4x16GB too, just that I was lucky to get it working at DDR5-6000 with some tweaks. But this is borderline situation and I might be forced to lower to reduced speeds at some point in time. Be aware that everything beyond DDR5-4800 is OVERCLOCKING territory and not covered by JEDEC standards. If you run an important production workload (non-gaming) it is still recommended to stick to JEDEC timings for stability reasons. Unless you know about the potential tradeoffs in terms of data corruption, crashes and potential hardware wear/damage.

View solution in original post

7 REPLIES 7

SkyBeam
Level 11

I am afraid there might be no stable solution. You should always check the memory QVL before. The one for your X670E Hero can be found here. Note that there is NO configuration with 4x32GB certified as of now. Yes it's a pity. There are also still some memory controller issues with AMD platforms with multiple modules per channel. The only 4-stick configurations which contain some certified modules is 4x8GB.

So having said that you chose to go for an unsupported memory configuration - which is fine, I have chosen that route too btw.
But just to be clear here you will be on your own and might have to figure out your specific configurations yourself. I did describe some success with my 4x16GB DDR5-6000 configuration (on a different board) here. This might give some hints.

However 4x32GB is putting even more load on the memory controller and VRM than my 4x16GB configuration. Some recommendations I can give:

  • Try first to get JEDEC timings (DDR5-3600 or DDR5-4800) stable
  • Try loading a custom preset matching you memory chips from AI Tweaker / DRAM Timing Control menu
  • Crank up your MC voltage to 1.4V and increase by 0.025 increments until you might reach a stable state
  • Run a full memtest86+ run after each modification checking for reported errors - keep it running for at least 1 hour or a complete run (all 10 tests pass)
  • If any errors reported by memtest86+ try adjusting timings (higher numbers) and tweak voltages

Exceeding safe voltage limits might cause permanent hardware damage. As said you have to do this on your own risk.

The fact you are reporting occasional reboots is a clear indicator your memory configuration currently is not stable. I am sure if you run memtest86+ in your current setup you will find a large amount of issues reported. Your system might still be able to boot into Windows just fine and you might experience occasional application crashes "only". That does not mean your system is stable! Sure you can blame browser tab crashes or occasional bluescreens on Windows or application faults but the truth is potentially garbled data in your memory. So absolutely make sure your memory configuration is stable by TESTING it (memtest86+).

And no, I do not believe you can go the full DDR5-5600 configuration with 4 sticks of this size. So you might be lucky already getting this configuration to run stable at DDR5-4800 JEDEC timings. Sometimes more memory is more important than fast memory. That's why I am running 4x16GB too, just that I was lucky to get it working at DDR5-6000 with some tweaks. But this is borderline situation and I might be forced to lower to reduced speeds at some point in time. Be aware that everything beyond DDR5-4800 is OVERCLOCKING territory and not covered by JEDEC standards. If you run an important production workload (non-gaming) it is still recommended to stick to JEDEC timings for stability reasons. Unless you know about the potential tradeoffs in terms of data corruption, crashes and potential hardware wear/damage.

Thank u so much !!!!!!  I tried what u said and the memory checker had 1000's of errors.  I wound up increasing the voltage to 1.275V and ran the memory checker to find zero issues.  No more reboots.  thanks for all your help !!!!!!

Hi, did you succeed in solving the problem with x4 32gb ddr5-5600 on the last update 1516?

I just spoke with the support and they told me to remove x2 32 rams in order for my PC to be stable, I did it, and it works but the problem is that I want 128gb ram

You will need to buy a matched kit. The reason support is telling you to do this is because combining memory kits is not recommended or supported.

13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

I would love to, but the problem is that asus x670e-e is advertising that support 128gb ram, but on the drivers section there aren't any stick of 128gb ram that are supported 1,2,4 slots, there are only 1,2 slots. It didn't find any ram x2 64gb ram. 

In this case what would be the solution? 

I would personally settle for 2x48GB. Four modules is tough on AM5.

Otherwise, you'll need to tune the system manually which may be beyond some individuals scope when combining kits. This includes: 

1). Evaluate if the kit is able to run at or even close to the rated frequency anymore. This depends on the amount of guardband the memory vendor has put in, the motherboard, platform and CPU. *

2). Evaluate if the kit is able to run at or close to the primary timings the memory vendor has put in. This also depends on the amount of guardband the memory vendor has put in.

3). Evaluate how much voltage the memory needs for the above two points.

4). Evaluate the motherboard sub timings to see if the kits are able to run at or close to the auto ruling assigned by the board vendor.

5). As with any memory kit, evaluate what voltages are needed on the CPU side (VSOC).

 

13900KS / 8000 CAS36 / ROG APEX Z790 / ROG TUF RTX 4090

I understand what you mean, I found this forum post with a similar problem: https://rog-forum.asus.com/t5/amd-600-series/x670e-and-128gb-of-ram/td-p/920938