[solved] [AM5, DDR5] Stuck on cpu+dram led unless I use a single specific slot with one specific stick of ram

Alright so Im currently pulling my hair out over this

I have yesterday build a new system comprised of

Asrock B650E PG Riptide WiFi (new)

AMD Ryzen 7600x (new)

2x Samsung DIMM 16GB, DDR5-4800, CL40-40-40-77, on-die ECC [M323R2GA3BB0-CQK] (new, lets call the two sticks alice and bob )

NH-D15 (from previous build, known good and working)

Seasonic Prime TX-750 750W PSU (from previous build, known good and working)

GTX 770 (from previous build, known good and working)

I checked beforehand that the samsung ram is on the QVL of the motherboard

Putting it all together and starting it for the first time resulted in… all fans spinning(probably at max speed, they are pretty loud), motherboard rgb kicking on aaand solid red lights from the cpu and dram debug leds from the second I push the power button and it stays like that until I longpress the power button to shut it down. It never does anything else, no blinking, no other leds come on even for an instant, its just a solid light on both cpu+dram.

The mother board has 4 ram slots and the manual advises you to fill the furthest slot from the socket first, leave one free then populate the next.

Trying around a bit I’ve gotten it to post if and only if I use Alice in the slot furthest from the cpu socket and nothing else:

  • alice in the furthest slot+gpu and everything else connected: Success! I get to boot the OS with graphics output after a minute

  • alice in the furthest slot with everything but Mobo and cpu fan disconnected: Success! I get to bios just fine after a minute

  • alice in the second furthest slot with everything but Mobo and cpu fan disconnected: Fail! cpu+dram lights indefinitely(waited 20 minutes)

  • alice in the third furthest slot with everything but Mobo and cpu fan disconnected: Fail! cpu+dram lights indefinitely(waited 20 minutes)

  • alice in the fourth furthest slot with everything but Mobo and cpu fan disconnected: Fail! cpu+dram lights indefinitely(waited 20 minutes)

  • bob in the furthest slot+gpu and everything else connected: Fail! cpu+dram lights indefinitely(waited 20 minutes)

  • bob in the furthest slot with everything but Mobo and cpu fan disconnected: Fail! cpu+dram lights indefinitely(waited 20 minutes)

  • alice in the furthest slot and bob in the second furthest slot with everything but Mobo and cpu fan disconnected: Fail! cpu+dram lights indefinitely(waited 20 minutes)

Yes, I did actually wait 20 minutes for each try because I read that the first boot times on AM5 can be ludicrous.

With most of those I did disconnect the power cable and jump the clear cmos pins with the noctua screw driver thing before trying to run them.

I dont have another cpu or ram to test with. I also dont have friends( ._.) I could borrow those from to test.

Honestly Im not entirely sure what to make of this. This feels like bob is dead and 3 out of 4 ram slots in my mobo are dead, which seems wildly unlikely.

I have not tried reseating the cpu yet because thats kind of a hassle(and for fear of getting thermalpaste everywhere and possibly bending the mobo pins while messing with that). Has anyone seen anything like that before or any idea on what that might be?

Also none of my components had any visual defects… that I could pick up on anyways.

Im on bios version 1.11(I used the bios flashback to put it there myself) which is the most recent not beta bios asrock lists.

QVL:

Perhaps a BETA BIOS is released to resolve this ?

My ASUS X670E mobo didn’t require any BIOS update to properly support the Corsair Vengence I installed into it, it ran at default speeds.

I was then able to update the BIOS to the latest build which does support better memory timings et.al., however I’ve not enabled EXPO yet.

Just flashed the most recent beta bios from asrocks website. Everything is exactly as it was before.

Not sure then, suggest checking out the ASrock forums too, maybe somebody has some insight there that they posted ?

Huh, I wasnt aware that they had a forum. I’ll ask there, fingers crossed I guess

Reseat the processor? The memory controller is on there and could be causing part of the problem. Also, I think on a recent Gamer’s Nexus video they mentioned the DDR5 slots are different from in the past, and it isn’t advisable to only push down on the RAM, but to push the retaining clips in to get it to snap in to place.

good timing, just got done reseating the cpu, tested a few ram configurations aaand everything appears exactly as before, alice working if and only if its alone and in the furthest slot. And bob not working in any slot.
Maybe the DDR5 thing is a brand specific thing because I can confirm that on my board you can just open the clips, push the ram down and they snap into where they need to be(I also saw the video and septuple checked that they are actually engaged and in there).

So what do I do now? I assume at this point I’ll just have to rma the board?

If Alice works in slot 4, but Bob doesn’t, then it seems the most likely problem is that Bob is defective.

The fact that Alice works in the fourth slot means that the motherboard works, the slot works, and the CPU works (and is installed correctly).

You should test with another memory stick, or ideally a pair in slots 4 & 2, before RMAing the motherboard.

I dont really have another stick to test with(nor the money to buy new ones, really)
I did test alice in all slots and it only works with alice in slot 4. in slot 1-3 it just gets stuck at training(I think? dram and cpu debug lights go on and stay on) indefinitely.

Yeah - it’s hard to resolve issues like this when you’re on a budget (as we all are). I’m still thinking that the memory is the culprit, so you may want to pursue a memory RMA. They (the store or Samsung) may suggest other tests for you to do, but that’s the most likely resolution IMO.

I have no experience with AM5/DDR5 (yet - ref: budgets :slight_smile: ), but I’m surprised that the single stick doesn’t work in slot 2. I know (from designing hardware with DDR3 and scanning the DDR5 specs) that signal termination is necessary at the end of the PCB traces, which would be the furthest memory slot. When you populate only the first slot, there’s no way to control reflections from the second slot. In theory, slots 2&4 should be at the ends of the traces for their respective channels, so either should work in a single-DIMM configuration.

There could be an issue with slot 2 of your motherboard, but you have to solve the problems sequentially. Make sure you have two working memory sticks, then see if you can get both to work at the same time.

yeah, I cant think of any other reason what else could cause the pattern Im seeing.
Perhaps my first hutch was right after all and its ram and motherboard.

This feels like bob is dead and 3 out of 4 ram slots in my mobo are dead, which seems wildly unlikely.

Guess I’ll start talking to samsung >.>

It just occurred to me, the ram I bought was bulk. Where do I turn to if one of those is dead, samsung or the seller where I got the ram from? Does that have implications for the rma process? I never really thought about that.

(as far as I understand it bulk means that the seller gets a pallet of 1000 or something and sells them individually to customers)

I’d go to the reseller. You might want to look at the product page - it may have warranty/returns info that’s specific to this memory. Samsung isn’t the place to go, especially because it’s bulk.

Yeah I figured.
So another update: I contacted the reseller and got the reply that the bulk order they placed was for an internal server upgrade and management just decided to flip the left over sticks on their store.
So now they cant really replace it because the bulk order left overs were completely sold through and they can only offer to give me a refund or keep it.
Makes sense in hindsight because those sticks were really cheap and the seller was reputable, but boy does that suck now.
I guess I’ll refund the broken stick and pay like 20 bucks more to get another one from a different seller(which sucks because shipping alone is half of the 20 more bucks I have to spent now).
Wow this is turning into my personal vent thread at an alarming rate…

return all the ram mate. and get a full refund.
theres no guarentee that when you buy another stick it will be the same.
even if its branded the same from the same manufacturer.

so yeah send it all back then buy a matched pair.

1 Like

yeah, I ended up doing that.

General update, I guess:
Ordered new ram, about the rma both of the samsung sticks.
The new ram is also on the QVL and also 2 sticks of 16GB each.
Because we are very near christmas the order is going to take a while to get here, if the board is still making trouble after that Im rma’ing the board.

Im feeling pretty smart about this because I can wait to actually send samsung rma until the new sticks get here so I have 4 sticks in total to test with a little bit before the samsungs have to go out the door ; )

1 Like

alright next and final update:

It works. I got the new sticks of ram in the mail(shout out to the delivery guy who delivered them on Christmas. I sincerely I never have to work for or with your employer ever), poped them into the furthest and third furthest slots(like the manual instructs) and bam, it booted within seconds 32 GBs detected and working.

Turns out the one samsung stick was just dead and my motherboard is apparently very particular about which slots you populate with ram sticks the working stick in any other slot but the first one also doesnt post. I wonder if thats an AM5 specific thing or an asrock specific thing or something specific to this board or this bios version, but if you are buying an AM5 board thats something you should keep in mind! Well at least thats what I learned now.

And thus this hardware adventure ends here. Thanks for all the help!

This topic was automatically closed 273 days after the last reply. New replies are no longer allowed.