Question PC stalls on DRAM check when booting ? (SOLVED)

[Oli_Pan]

Hi all,

Built a new PC a few months back. Took me a while to get it to boot when it was first built.

The issue I had was that it would boot just fine with one RAM stick plugged in, but not with both.

I tried swapping out the sticks and they would both work independently, in various slot combinations, but not together.

The workaround I found at the time was to boot to BIOS with 1 stick, enable XMP, then it would boot with both sticks.

However, more recently, my PC stalls when booting, cycling between the CPU and DRAM warning lights several times before finally posting.

I've tried updating the BIOS, reseating the CPU and RAM and cleaning the components.

I'm wondering if there is something else I should configure in the BIOS or if this sounds like a hardware issue?

I haven't overclocked anything or changed anything else in the BIOS except the aforementioned XMP.

Any suggestions would be greatly appreciated, thankyou!

Case: Corsair 4000D Airflow
Motherboard: Gigabyte Z790 Aorus Elite AX Rev:1.1
CPU: Intel i5 13600k
RAM: Corsair Vengeance 32GB (2x16GB) 5600MHz
PSU: Corsair RM750x 80PLUS Gold
GPU: Gigabyte RTX 4070 WINDFORCE OC 12GB
SSD: WD_BlACK SN850X 1TB
Heatsink: Noctua NH-D15 chromax.black

 

[SkyNetRising]

Turn off automatic overclocking (XMP) and overclock manually.
Can you show screenshots from CPU-Z - memory and spd sections?
(upload to imgur.com and post link)

Are ram modules installed in proper ram slots for dual channel operation - slots A2/B2 ?

 

[Oli_Pan]

Turn off automatic overclocking (XMP) and overclock manually.
Can you show screenshots from CPU-Z - memory and spd sections?
(upload to imgur.com and post link)

Are ram modules installed in proper ram slots for dual channel operation - slots A2/B2 ?

Thanks for getting back to me.

Sorry, but I've never overclocked RAM manually before so I will have to look into that.

These are the requested screenshots:

https://imgur.com/a/Ewfvw89

View: https://imgur.com/a/Ewfvw89


Yes, the RAM modules are installed in A2/B2 - I have also tried all other configurations.

 

[Misgar]

In theory your system should be more stable at the quoted (non overclocked) "SPD Speed" of 4800MT/s than when it's overclocked at an "XMP Speed" of 5600MT/s. However, your system seems to work in reverse, being more stable at the faster speed.
https://www.corsair.com/uk/en/p/mem...0mhz-c40-memory-kit-black-cmh32gx5m2b5600c40k

The DRAM Frequency in the CPUID plot shows your memory is running at 2800MHz, which equates to 5600MT/s, because the RAM is Double Data Rate (2 x 2800MHz = 5600MT/s). N.B. MT/s = Mega Transfers per second.

It might be an idea to test the DIMMs by booting up the computer from a USB memory stick configured with MemTest86+.
https://memtest.org/

It could take an hour or two to run a full pass on MemTest86+ and it doesn't hurt to wait until it completes several runs. If MemTest86+ returns zero errors, your RAM is probably OK. If you get even one error, at least one DIMM may be faulty and should be replaced.

If MemTest86+ shows any errors, remove one DIMM and run MemTest86+ again. Repeat with the second DIMM on its own. If either DIMM is faulty, you need to replace both sticks to ensure you have a "matched" pair. Dissimilar single DIMMs will have different characteristics and may not work well together as a pair.

I'm writing this on an ancient AMD FM2 system where only 2 out of the 4 memory slots work. I've never bothered to find out if it's a problem with the CPU, the DIMM sockets, or board warping. It's possible you have a similar problem with the Integrated Memory Controller in your CPU and/or a problem in the motherboard.

You have a Socket 1700 CPU and some motherboards are known to bend or warp under the strain of CPU cooler mounting systems. In extreme cases, this might affect the memory bus connections between the CPU and the socket, or the tracks between the CPU socket and the DIMM sockets.

If your board is susceptible bending, consider this adapter to stiffen up the board:
https://www.tomshardware.com/reviews/thermalright-lga1700-bcf-contact-frame

 

[Oli_Pan]

In theory your system should be more stable at the quoted (non overclocked) "SPD Speed" of 4800MT/s than when it's overclocked at an "XMP Speed" of 5600MT/s. However, your system seems to work in reverse, being more stable at the faster speed.
https://www.corsair.com/uk/en/p/mem...0mhz-c40-memory-kit-black-cmh32gx5m2b5600c40k

The DRAM Frequency in the CPUID plot shows your memory is running at 2800MHz, which equates to 5600MT/s, because the RAM is Double Data Rate (2 x 2800MHz = 5600MT/s). N.B. MT/s = Mega Transfers per second.

It might be an idea to test the DIMMs by booting up the computer from a USB memory stick configured with MemTest86+.
https://memtest.org/

It could take an hour or two to run a full pass on MemTest86+ and it doesn't hurt to wait until it completes several runs. If MemTest86+ returns zero errors, your RAM is probably OK. If you get even one error, at least one DIMM may be faulty and should be replaced.

If MemTest86+ shows any errors, remove one DIMM and run MemTest86+ again. Repeat with the second DIMM on its own. If either DIMM is faulty, you need to replace both sticks to ensure you have a "matched" pair. Dissimilar single DIMMs will have different characteristics and may not work well together as a pair.

I'm writing this on an ancient AMD FM2 system where only 2 out of the 4 memory slots work. I've never bothered to find out if it's a problem with the CPU, the DIMM sockets, or board warping. It's possible you have a similar problem with the Integrated Memory Controller in your CPU and/or a problem in the motherboard.

You have a Socket 1700 CPU and some motherboards are known to bend or warp under the strain of CPU cooler mounting systems. In extreme cases, this might affect the memory bus connections between the CPU and the socket, or the tracks between the CPU socket and the DIMM sockets.

If your board is susceptible bending, consider this adapter to stiffen up the board:
https://www.tomshardware.com/reviews/thermalright-lga1700-bcf-contact-frame

Hi thanks for your response!

I've just run memtest86+ off a usb drive and it's passed with no errors. If the issue was related to the CPU memory controller, is this something that would have been picked up or is there another way to check?

Likewise, how can I rule out a motherboard issue? As I said, the pc does eventually boot it just takes 30 seconds - a minute because it stalls on the DRAM and CPU warning lights.

I had come across the suggestion of heatsink pressure affecting the CPU or motherboard and this was something that I mentioned in my initial post. I basically disassembled the pc and reseated the CPU and all other components as well as unplugging and replugging all power cables. I've made sure that there isn't too much pressure applied.

As mentioned, I also tried each of the memory slots and different configurations. The only one that causes no issue is when only a single RAM stick is used.

Do you know of any other possible issues? Thanks for your help.

 

[Misgar]

The delay at boot up is probably due to "memory training" but I would have expected your system to stabilise after a few boots. Perhaps your board is different and needs to train each time.

I have one DDR5 system and it's the only one which pauses at startup with training. I just accept it can take a 20 to 30 seconds before it starts booting into Windows. I'm not running an XMP overclock.

My DDR4 systems seem less prone to training and the DDR3 systems just start immediately, (or as soon as they've initialised the SAS HBA controller).

 

[Oli_Pan]

The delay at boot up is probably due to "memory training" but I would have expected your system to stabilise after a few boots. Perhaps your board is different and needs to train each time.

I have one DDR5 system and it's the only one which pauses at startup with training. I just accept it can take a 20 to 30 seconds before it starts booting into Windows. I'm not running an XMP overclock.

My DDR4 systems seem less prone to training and the DDR3 systems just start immediately, (or as soon as they've initialised the SAS HBA controller).

I think this would be acceptable after changing components or BIOS settings. From what I can find online, these are the things that result in 'memory training' scenarios. It's not really acceptable for long boot times every time and the DRAM/CPU warning lights cycling leads me to believe the issue may be something else?

It's worth pointing out that 30sec-1min is a conservative estimate for the best case. Sometimes I'm waiting for 2-3 minutes, watching the lights blink back and forth while the fans spin and nothing shows on screen 🙁

Then out of nowhere the system posts and all the warning lights turn off. That's another point worth making - the DRAM and CPU lights are no longer on once booted to BIOS/Windows.

 

[Oli_Pan]

I've found a fix: Did some more fiddling around in BIOS and discovered there are BIOS presets for Hynix, Micron and Samsung at different speeds/voltages. By default, turning on XMP for my system was setting a 5600MHz frequency at 1.25V (with different SPD values), but there was also a few others up to 6400MHz for Samsung RAM (such as mine) that sets it to 1.35V.

Figured what the hell, tried them all. Turns out, the highest setting resolves my stalled boot issue!

That's right, my PC appears to be the most stable with the highest XMP overclock setting. My 5600MHz RAM is now set to 6400MHz and I can boot seamlessly in 5 seconds with 2 RAM Modules.

I was curious whether the increase in voltage was the deciding factor, so I actually went back to try running at a lower frequency and just up the voltage a bit manually. This didn't work so I'm going to leave it on the higher XMP setting for now since it seems to have solved the issue. Time will tell if this is actually stable for my hardware.