MSI X670E Carbon won't reboot reliably - DDR5 issue?

Zinn

Zinn

2[H]4U
Joined
Jan 31, 2010
Messages
3,441
I have an MSI X670E Carbon on the latest BIOS with 2x32gb Vengeance DDR5-6000 (AMD Expo version) that won't reboot reliably. If I cold cold boot it, it usually boots to Windows after about 5 minutes of black screen. If I reboot it never comes back up. I have to turn it off and on repeatedly before it will boot again.

I have a feeling it's gotta be a memory issue. When I try it with my Dominator 4x16gb DDR5-6000 (Intel XMP version) it NEVER passes POST, just blackscreens forever. And again with the Vengeance 2x32gb it's hit or miss. Is it a known issue that X670E kind of sucks at memory support or is my motherboard suspect?

Once I get into Windows everything is fine and stable. I have an Intel Z790 CPU and motherboard lying around that I could rebuild with, and it was rock solid with this memory when it was built, but I really like the power efficiency of the Ryzen 7900. But these booting issues are driving me nuts lately.

I was eyeing this G.SKILL Trident 96gb DDR5 Expo kit but don't know if that would just be asking for more trouble. I run a lot of Virtual Machines so the extra capacity would be nice.
 
Last edited:
Are you running bios default and still having issues posting? It’s quite possible your particular cpu and or mobo and those kits are just not playing nice for whatever reason. As far as timings and speed some memory controllers on some CPUs are just ass. Silicon lottery and all that.

My experience:
My AM5 system is super pissy regarding ram. X670e Gene 7700x and GSkill x5 c30 32g expo kit. It will not run 6000 stable but setting the expo profile and dropping the speed to 5600 manually it’s solid as a rock. This is with the latest bios with all the latest AGESA , latest chipset drivers from AMD, all windows 11 updates etc.

Edit: I’ve not tried any other ram kits but since I have an AMD expo rated kit I suspect my 7700x has a weak memory controller and I’m not going to push more voltage and so on to make 6000 when 5600 is fine. Eventually I’ll swap to a 2x32 64g kit (2slots on the mobo) and see how that goes.
 
  • Like
Reactions: Zinn
like this
TheHig said:
Are you running bios default and still having issues posting? It’s quite possible your particular cpu and or mobo and those kits are just not playing nice for whatever reason. As far as timings and speed some memory controllers on some CPUs are just ass. Silicon lottery and all that.

My experience:
My AM5 system is super pissy regarding ram. X670e Gene 7700x and GSkill x5 c30 32g expo kit. It will not run 6000 stable but setting the expo profile and dropping the speed to 5600 manually it’s solid as a rock. This is with the latest bios with all the latest AGESA , latest chipset drivers from AMD, all windows 11 updates etc.

Edit: I’ve not tried any other ram kits but since I have an AMD expo rated kit I suspect my 7700x has a weak memory controller and I’m not going to push more voltage and so on to make 6000 when 5600 is fine. Eventually I’ll swap to a 2x32 64g kit (2slots on the mobo) and see how that goes.
Click to expand...
That's a good idea - I'll try setting 5600 in bios next time I have to reboot and see how that goes. I wouldn't mind giving up a little bit of speed if it's more reliable.
 
Zinn said:
I have an MSI X670E Carbon on the latest BIOS with 2x32gb Vengeance DDR5-6000 (AMD Expo version) that won't reboot reliably. If I cold cold boot it, it usually boots to Windows after about 5 minutes of black screen. If I reboot it never comes back up. I have to turn it off and on repeatedly before it will boot again.

I have a feeling it's gotta be a memory issue. When I try it with my Dominator 4x16gb DDR5-6000 (Intel XMP version) it NEVER passes POST, just blackscreens forever. And again with the Vengeance 2x32gb it's hit or miss. Is it a known issue that X670E kind of sucks at memory support or is my motherboard suspect?

Once I get into Windows everything is fine and stable. I have an Intel Z790 CPU and motherboard lying around that I could rebuild with, and it was rock solid with this memory when it was built, but I really like the power efficiency of the Ryzen 7900. But these booting issues are driving me nuts lately.

I was eyeing this G.SKILL Trident 96gb DDR5 Expo kit but don't know if that would just be asking for more trouble. I run a lot of Virtual Machines so the extra capacity would be nice.
Click to expand...

This is my current build (well the new parts of it anyway):
https://pcpartpicker.com/list/VNN29c
Everything was open box except the CPU. I'm using the same motherboard and same capacity RAM, on the same BIOS. It's been working fine at its rated speeds. I just picked the XMP profile MSI had listed in the bios and haven't really had any instability or anything. As far as booting, it does take a bit (1-2 mins or so?) every boot because AM5 has to do a memory check (and afaik you can't disable that memory check; otherwise most boards won't even post with XMP ram slotted), but I've never had it just not boot during a restart. That being said most of the time I just leave it on anyway. While troubleshooting my (non-)issue, though, I did many restarts; it never failed to come back up. You might have some weird issue with your board. Did you read the post code readout during these boots? From my experience, my 1-2 minutes every boot is usually (iirc) a yellow LED light. I didn't keep track of the post code, but I think it was like 15 or 16 or something and indicated memory training or memory check. Afterwards it would just boot entirely like normal. The period is around 1-2 minutes (or maybe less, I didn't really time it). The very first boot I did on this board, though, did take a very long time. Like 3-5 minutes. I think that was a more thorough memory check or something. It seems like yours might be doing that repeatedly.

Dumb question but are you sure you're populating the memory slots properly? This is what mine looks like, snagged from my other topic (where I thought I was having some issues but in reality was not lol):

1694540378801.png


To put that into physical context, that's these slots:
1694540862994.jpeg


I remember being a little confused while reading the manual about which slots to populate, so I ended up watching the youtube video and remember many people complaining because it was counterintuitive or something.

Edit: Oh, wait, I remember having to restart multiple times when playing Divinity Original Sin 2 with my friend as well, recently. It had some weird network issues that could only be resolved by restarting... again yeah no issues.
 
Last edited:
I have not had an AM5 system with RAM over 5600 work with memory context restore on yet (and by work I mean every day - plenty of people claim its ok and then suddenly a week later it keeps failing). Any 6000+ kit always needed to retrain the controller every boot which adds 15-20 seconds. I stopped caring about the extra boot time and just let it train now and it works fine.

That being said, if you're just loading XMP/EXPO its luck of the draw with dual rank kits like yours. You may need to do some serious tuning to get it fully stable. This is simply the reality with DDR5, its the same on Intel. The profiles do not store enough data for the subtimings which make or break your ability to POST.
 
TheHig said:
Are you running bios default and still having issues posting? It’s quite possible your particular cpu and or mobo and those kits are just not playing nice for whatever reason. As far as timings and speed some memory controllers on some CPUs are just ass. Silicon lottery and all that.

My experience:
My AM5 system is super pissy regarding ram. X670e Gene 7700x and GSkill x5 c30 32g expo kit. It will not run 6000 stable but setting the expo profile and dropping the speed to 5600 manually it’s solid as a rock. This is with the latest bios with all the latest AGESA , latest chipset drivers from AMD, all windows 11 updates etc.

Edit: I’ve not tried any other ram kits but since I have an AMD expo rated kit I suspect my 7700x has a weak memory controller and I’m not going to push more voltage and so on to make 6000 when 5600 is fine. Eventually I’ll swap to a 2x32 64g kit (2slots on the mobo) and see how that goes.
Click to expand...
I tried this and now the system is booting / rebooting reliably -- thanks!

Still dealing with 5 min+ POST times due to memory training but I'm going to cut my losses here. Tweaking the memory context restore setting only leads to misery.
 
Last edited:
Glad to hear its up and running at least. I just did a full shut down and full cold boot and timed it. 54 seconds to the Windows log in screen from cold boot. After updating to the most recent bios I have only set:
-15 per core curve optimizer
1.35v memory and 1.35v CPU VDDIO/MC. Expo primary timings and rest on auto.
Ram ends up like this:

1694555878127.png

SOC is 1.25ish on auto which is the bios corrected setting since earlier versions had it over 1.36

Disabled the IGPU on the CPU
Disabled Thunderbolt via USB
Disabled Asus Armory Crate and some other Asus BS that actually has a bios setting to keep it from downloading bloatware --- enabled by default of course! :facepalm:

Fast Startup is disabled in Windows.
 
Zinn said:
Voltage is spec'ed at 1.35v at 6000 (Expo), I don't know what specific chips are in my Corsair Vengeance kit, so is it even safe to push it higher?

It's this kit fwiw: https://www.newegg.com/corsair-64gb/p/N82E16820236939?Item=N82E16820236939
Click to expand...
wouldnt suggest it if it wasnt. every kit of corsair ive ever used needed just a bit more that what the said they wanted. although i havent used ddr5 corsair... bump it to like 1.38, itll be fine. makers say 1.4v max for daily use is safe.
 
pendragon1 said:
wouldnt suggest it if it wasnt. every kit of corsair ive ever used needed just a bit more that what the said they wanted. although i havent used ddr5 corsair... bump it to like 1.38, itll be fine. makers say 1.4v max for daily use is safe.
Click to expand...
Alright, thanks. I bumped it up to 1.39v, set the speed back to the Expo 6000mhz and now I'm able to reboot 100% reliably! It's kind of stupid they spec'ed these kits so close to the wire, but I don't care as long as it's working now. Also it boots in about 45 seconds now instead of 5 minutes so that's ... better I guess.
 
TheHig said:
Glad to hear its up and running at least. I just did a full shut down and full cold boot and timed it. 54 seconds to the Windows log in screen from cold boot. After updating to the most recent bios I have only set:
-15 per core curve optimizer
1.35v memory and 1.35v CPU VDDIO/MC. Expo primary timings and rest on auto.
Ram ends up like this:

View attachment 598366
SOC is 1.25ish on auto which is the bios corrected setting since earlier versions had it over 1.36

Disabled the IGPU on the CPU
Disabled Thunderbolt via USB
Disabled Asus Armory Crate and some other Asus BS that actually has a bios setting to keep it from downloading bloatware --- enabled by default of course! :facepalm:

Fast Startup is disabled in Windows.
Click to expand...

I think MSI AM4 and AM5 boards have longer boot times by default, too. I know my MSI X570S Carbon had pretty long boot times, comparatively, too. No idea why.

My default XMP profile for the RAM sticks I have set it to 1.4V, so I'm assuming that should be fine... since it was default. My SOC as dictated by the motherboard seems to be set at exactly 1.3. I've been wondering for quite a while as to whether that's unsafe or not, but the system has been stable. I've asked this before, but is 1.3 too high?

Edit: Just went ahead and set SoC to 1.25 manually, seems to be working fine. Shrug. I might set my own RAM voltage down to 1.39 or so if I can, too, since pendragon says 1.4 is the upper limit of what you should be running. Informative topic, since we have the same board, thanks.
 
Last edited:
StoleMyOwnCar said:
I think MSI AM4 and AM5 boards have longer boot times by default, too. I know my MSI X570S Carbon had pretty long boot times, comparatively, too. No idea why.

My default XMP profile for the RAM sticks I have set it to 1.4V, so I'm assuming that should be fine... since it was default. My SOC as dictated by the motherboard seems to be set at exactly 1.3. I've been wondering for quite a while as to whether that's unsafe or not, but the system has been stable. I've asked this before, but is 1.3 too high?

Edit: Just went ahead and set SoC to 1.25 manually, seems to be working fine. Shrug. I might set my own RAM voltage down to 1.39 or so if I can, too, since pendragon says 1.4 is the upper limit of what you should be running. Informative topic, since we have the same board, thanks.
Click to expand...
For whatever it's worth the DIMM voltage setting in BIOS turns red when I set 1.4v or higher. It's the normal color at 1.39v or below. I'm keeping mine at 1.39v since that seems to be the upper limit of what MSI thinks of as "safe."
 
Zinn said:
For whatever it's worth the DIMM voltage setting in BIOS turns red when I set 1.4v or higher. It's the normal color at 1.39v or below. I'm keeping mine at 1.39v since that seems to be the upper limit of what MSI thinks of as "safe."
Click to expand...

Did you have to dial all of the settings in manually, or could you work off of the existing XMP/EXPO profile and just tweak the voltage alone? It's kind of funny that GSkill just decided to set the default voltage on these to 1.4, for EXPO profile... it's kind of just a handwave "we don't really feel like tweaking it" sort of thing. It's been working fine, but I'll probably try to adjust it tomorrow. While I'm at it, I might try to dial these settings in for the subtimings:
https://www.patreon.com/posts/low-effort-rank-77403831

I'll probably just set the voltage to like 1.39 like you did, though, since he was working with single rank memory while mine is dual rank... Hopefully it will be alright since it's Hynix M-die. Did you check who manufactured yours?
 
StoleMyOwnCar said:
Did you have to dial all of the settings in manually, or could you work off of the existing XMP/EXPO profile and just tweak the voltage alone? It's kind of funny that GSkill just decided to set the default voltage on these to 1.4, for EXPO profile... it's kind of just a handwave "we don't really feel like tweaking it" sort of thing. It's been working fine, but I'll probably try to adjust it tomorrow. While I'm at it, I might try to dial these settings in for the subtimings:
https://www.patreon.com/posts/low-effort-rank-77403831

I'll probably just set the voltage to like 1.39 like you did, though, since he was working with single rank memory while mine is dual rank... Hopefully it will be alright since it's Hynix M-die. Did you check who manufactured yours?
Click to expand...
Setting the Expo profile populates the memory overclocking settings screen, so you can hand tune settings from the profile without keying everything in manually.

I don't know how to check the chip manufacturer besides taking the heat spreaders off, CPU-Z doesn't list it anyway.

Just out of curiosity for AM5 - does cooking your memory with too much voltage risk damaging the CPU / chipset, or would it just mean swapping memory modules when they die an early death?
 
Does it have an ASMedia SATA controller on it for the "extra" controller? If so, make sure no drives are connected to it and disable that thing in the bios. I had similar boot issues on my Z690 carbon wifi and disabling that controller fixed everything.
EDIT: oh never mind .. looks like that board doesn't have that controller on it.
 
Last edited:
Zinn said:
Setting the Expo profile populates the memory overclocking settings screen, so you can hand tune settings from the profile without keying everything in manually.

I don't know how to check the chip manufacturer besides taking the heat spreaders off, CPU-Z doesn't list it anyway.

Just out of curiosity for AM5 - does cooking your memory with too much voltage risk damaging the CPU / chipset, or would it just mean swapping memory modules when they die an early death?
Click to expand...

I have no idea about the last question, but your CPU-Z doesn't have anything here?
1694864755222.png


You can Google around for your RAM model and maker, sometimes there are random forum threads in the know where you can basically scan the model number and get which die you have.

For instance, here is where I found mine: https://forums.overclockers.com.au/threads/hynix-a-die.1329881/

And then if I scan my tag sticker on my ram stick (excuse the garbage quality, getting a picture of this with my phone was hard enough with the computer already running):
1694866223883.png


That's how I know it's M-die.

Here's a thread I found for Corsair, but it might be too old:
https://linustechtips.com/topic/1412104-finding-my-dram-die-type/?do=findComment&comment=15263824
I have no idea where the heck they get this info at.

LigTasm Very slight derail, but... does anything here look weird? I tweaked my settings to the "low effort Hynix" timings from Buildzoid. I thought it wouldn't work since mine are dual rank, but... system POSTed with no issues (actually posted faster for some reason). Running prime95, but am seeing nothing but passes. Some of the stuff was named differently (his document said "tRFC4" but the only other RFC I saw on the list was RFCSB, so I just changed that one...).

1694866479836.png


Should I still do a memtest86 loop sometime?
 
Last edited:
Actually someone in a thread gave the tipoff that you can look at motherboard RAM support lists and it would give you the type of die you're working with. I looked up DDR5 32GB Corsair on here:
https://www.gigabyte.com/Motherboard/Z790-AORUS-MASTER-rev-10/support#support-memsup

1694867614349.png


Is this what you have? Looks like Samsung B die if so.

I tried looking it up on here, too, and didn't even see a 6000 speed RAM that was 32GB capacity (I think it's per-stick basis) from Corsair:
https://www.gigabyte.com/Motherboard/X670E-AORUS-MASTER-rev-1x/support#support-memsup

There's also MSI's page, but it's kind of harder to browse:
https://www.msi.com/Motherboard/MPG-X670E-CARBON-WIFI/support#mem
But I think I did find your sticks on there?

1694868094321.png



Doesn't seem like it's always accurate, though. The only G.Skill sticks on either of Gigabyte's pages at 6000 and 1.4v say Samsung, but my sticks certainly aren't Samsung at all...
 
Last edited:
StoleMyOwnCar said:
I have no idea about the last question, but your CPU-Z doesn't have anything here?
View attachment 599062

You can Google around for your RAM model and maker, sometimes there are random forum threads in the know where you can basically scan the model number and get which die you have.

For instance, here is where I found mine: https://forums.overclockers.com.au/threads/hynix-a-die.1329881/

And then if I scan my tag sticker on my ram stick (excuse the garbage quality, getting a picture of this with my phone was hard enough with the computer already running):
View attachment 599064

That's how I know it's M-die.

Here's a thread I found for Corsair, but it might be too old:
https://linustechtips.com/topic/1412104-finding-my-dram-die-type/?do=findComment&comment=15263824
I have no idea where the heck they get this info at.

LigTasm Very slight derail, but... does anything here look weird? I tweaked my settings to the "low effort Hynix" timings from Buildzoid. I thought it wouldn't work since mine are dual rank, but... system POSTed with no issues (actually posted faster for some reason). Running prime95, but am seeing nothing but passes. Some of the stuff was named differently (his document said "tRFC4" but the only other RFC I saw on the list was RFCSB, so I just changed that one...).

View attachment 599065

Should I still do a memtest86 loop sometime?
Click to expand...

That looks good, but DDR5 is a strange animal. I’ve done 24 hours testing marathons with all sorts of programs and passed but then crashes the first time i loaded a game, or had weird anomalies in browsers like tabs refusing to load or crashing. I’d say just use it normally after you do an initial stress test and see what happens.
 
LigTasm said:
That looks good, but DDR5 is a strange animal. I’ve done 24 hours testing marathons with all sorts of programs and passed but then crashes the first time i loaded a game, or had weird anomalies in browsers like tabs refusing to load or crashing. I’d say just use it normally after you do an initial stress test and see what happens.
Click to expand...
Knock on wood, but I think it's pretty stable. The only "instability" I've had is some Starfield crashes when loading saves. But they were just clean crashes to desktop and nothing else on the system had any issues; there wasn't anything in the event viewer, either. They seemed to be very consistently only in a certain areas, and they only happened in the first place because I was spam reloading while save scumming weapon drops. I've been doing normal play since yesterday or even before yesterday and have had no crashes (including spam reloading, but in another area...). So I think that's completely unrelated to RAM. All my browsers and other programs have worked perfectly fine and have had no issues for around... 8 days now, I guess?

So yeah these Buildzoid timings are pretty legit if you have Hynix ram. The system definitely feels much snappier. I left mine on 1.39v because it's dual rank, seems to be doing fine.
 
You must log in or register to reply here.
Back
Top