Not sure if its Asus DOCP or my ram with reboot halting.

[middy500]

I asked this over at OCN but thought I would ask it here for help too. There seems to be about a 30% chance of my kit, Gskill F4-3600C16D-32GTZNC (on Asus and Gskill QVL for the X570 Tuf which I'm using), to halt at reboot. My Tuf has its DRAM LED light on and it just halts there until I hold down my power button. I never got to try any other bios below 1403 as the first thing I did when I got the board was flash 1403. Then 1404 when it came out and finally now 1405. All three bios versions have this issue.

The odd thing about it is that, so far, it has been exclusive to rebooting. Not shutdown and powering back on and not from a cold boot after pulling out the power cord. But just from rebooting. And when it happens, once I hold down my power button to shutdown, and boot it back up, it boots just fine like normal. No issues when gaming long sessions, did prime95 well when I did my initial benchmark stuff, etc. I've tried clearing cmos, reflashing my bios, manually setting FCLK to 1800 and SOC voltage to 1.1v (which is what Auto with DOCP enabled was defaulting to), etc.

I did see a guy on newegg for the X570 Tuf complain about something that sounds very similar to what I have been enduring (chance of halting at boot up with his 3200mhz kit since updating to 1404 after coming from 1203 that worked fine) so it lends me to believe its a Asus bios issue. The guy was hinting at it being a DOCP issue. I am running it with DOCP. So yesterday I decided to write down my timings that where being set with DOCP and manually set everything. But I ran into a two problems.

The first problem is the very last timing to set in the bios, tcke, was post to be set to 0 (as 0 was the value DOCP was setting on auto) but when I put in 0, it changes it to 1. So some how auto sets it to 0 but manually I can't. Ryzen Master also shows it as 0 as well when DOCP is enabled. The second problem was after setting all the timings, memory to 3600, FCLK to 1800, DRAM voltage to 1.35v's, and SoC voltage to 1.1v's, my computer can't post. It just halts 100% of the time on the DRAM led light. So I had to clear cmos to get back in. So either that very last timing is what's causing the manual setting to fail to post or DOCP is also enabling other stuff that I can't figure out what else. Even though everything else is on auto by default either with or without DOCP, I'm assuming DOCP changes what those auto settings default to.

I am using slots A2 and B2 as well and fast boot in windows is disabled.

So because of that it does have me a bit concerned that maybe those times where I fail to reboot are not due to an Asus bios bug, but maybe hardware? Since I still have a return window for both a refund and replacement I don't know if I should gamble it?

I've run for an hour and a half of Ramtest from Karhusoftware that I found on OCN and after 1575% coverage, no errors. After that I ran Memtest Pro 7.0 for a bit over an hour to reach a little over 100% coverage and no errors. Ran both yesterday.

What do you guys think? Anyone else with a similar problem? Do you think its just Asus bios bug or hardware?

 

[dejo]

mine will halt on reboot maybe once every 30 reboots. if it doesnt like a ram timing setting- it just sticks till I clear the bios, but that is only when I overreach my rams ability/voltage

 

[v3x0rg]

Hey guys, I am having the exact same issues!

DCOP settings do not work. Basically IF the computer boots, everything is fine. I also ran Memtest86 for ~3.5 hours and then for ~2 hours with zero errors. It messes up when I restart, and the only way to fix it is to CMOS reset. No amount of turning it off/on will fix the issue, must be cleared. Running 1405 BIOS on Asus TUF Gaming x570 Wifi Plus.

@middy500, I saw you subscribed to my thread as well. I have a couple and I am currently communicating with GSkill / Asus via email support. So far their recommendations haven't really solved the problem, synopses below. I am going to send them a link to this thread/my post here.

GSkill Support
- Use slot 3/4
- Try sticks individually
- Try lower RAM frequency

Asus Support (waiting to hear back)

I have two posts:
Overclocke.net: https://www.overclock.net/forum/11-...t-z-neo-2x16gb-3600-mhz-freq-timing-help.html
Overclockers.com: https://www.overclockers.com/forums...0-MHz-Freq-Timing-Help!?p=8131963#post8131963

I will keep you guys posted.

Currently up and running with the below.

https://imgur.com/a/G53IE5L

 

[Woomack]

It usually happens when something isn't fully stable or motherboard sets wrong sub timings or something else. Sometimes higher VDIMM or SOC voltages are helping, sometimes just have to set lower memory frequency.
As long as you run memory in the right slots (should be 2/4 counting from the CPU side, no matter how are they called) then all should run at least up to 3600.
There is also one more issue which in theory shouldn't appear but it seems more common that most users think. Some CPUs simply don't like memory clock above rated 3200. I had that issue with one 3700X and it was appearing only on one motherboard. Support (not the first helpdesk line) said it sometimes happens.

You may try manual settings like:
Manual 3600 18-20-20 1.40V. Disabled DOCP/XMP. All other timings at auto. If you make it work without issues then try tighter timings like 16-20-20 or 18-19-19. I doubt that higher memory voltage will help.

No other ideas right now.

 

[middy500]

It usually happens when something isn't fully stable or motherboard sets wrong sub timings or something else. Sometimes higher VDIMM or SOC voltages are helping, sometimes just have to set lower memory frequency.
As long as you run memory in the right slots (should be 2/4 counting from the CPU side, no matter how are they called) then all should run at least up to 3600.
There is also one more issue which in theory shouldn't appear but it seems more common that most users think. Some CPUs simply don't like memory clock above rated 3200. I had that issue with one 3700X and it was appearing only on one motherboard. Support (not the first helpdesk line) said it sometimes happens.

You may try manual settings like:
Manual 3600 18-20-20 1.40V. Disabled DOCP/XMP. All other timings at auto. If you make it work without issues then try tighter timings like 16-20-20 or 18-19-19. I doubt that higher memory voltage will help.

No other ideas right now.

Yeah I thought it could be my 3800x as well but the thing that puzzles me is that outside the like every 20th reboot, its stable. Even when it halts, if I power off, and turn back on, it boots fine. Passes memory test's, gaming 14 hours straight, etc. That one post I saw on Newegg about the same thing and he tried different kits and couldn't even do 3200 without a random halt at reboot. But never had that problem on 1203 and less bios. Though at the same time, ryzen itself could be playing a role.

Maybe ill pick up a 3200mhz kit to try.

Edit:
I went ahead and ordered this kit to test with: https://www.amazon.com/gp/product/B07TB47MN4
Its the Corsair Vengeance RGB Pro 32GB (2x16GB) DDR4 3200 - "AMD Optimized" (CMW32GX4M2Z3200C16) and on the QVL for the Tuf by Asus.

 

[middy500]

So want to update that the new kit so far has not had any DRAM reboot halts for the short amount of tests I have done. So this lends me to believe maybe my 3800x either doesn't like 3600mhz, or whatever timings Asus is setting for my gskill neo kit.

So in the meantime I'm going to continue to test it but I did order a MSI X570 Unify along with a 3900x. So I'm going to test my gskill kit with that setup as well.

 

[dejo]

my tuf x570 has decided it doesnt like to sleep, or wake from sleep. This isnt after a bios change and not sure what is causing it. I am getting ready to change cases and will throw in a different m.2 and see if its something that microslop has decided I needed.

 

[middy500]

So after I posted about no DRAM halt on reboot, I just got a DRAM halt on reboot with the corsair 3200mhz kit. Another kit on Asus QVL and Corsair as well for the Tuf. So either my 3800x doesn't like 3200mhz either OR there is a bug with Asus and DOCP.

My MSI X570 Unify came today and now just waiting for my 3900x to arrive today as well. I also ordered this https://www.amazon.com/Ballistix-Sport-PC4-25600-288-Pin-Memory/dp/B07M5RKH5Z/ kit as well to test. Its on QVL too. Gonna do a lot of testing this weekend.

edit:
For those wondering, my Corsair kit came with SK Hynix:

https://imgur.com/HBJ2x7G

Interesting as Asus in their QVL stated that kit came with Micron and MSI on their QVL list for the X570 Unify said that kit comes with "Spectek Z11B." The Gskill Neo is Hynix as well. Maybe that's a theme with Hynix? Either way, my Crucial kit coming should be Micron E-Die so that be interesting.

 

[dejo]

I have the patriot viper blackout that is rated 4000mhz 19-19-19-39 and it is samsung b die

 

[middy500]

Holy batman.

My 3900x with my MSI X570 Unify with my Crucial 32gb 3200mhz kit:

https://imgur.com/zv8nfXQ

My former 3800x with my former Asus X570 Tuf with my Gskill Neo 3600mhz kit:

https://imgur.com/yQXWosD

I really don't know what to say. Is Asus butchering the timings with DOCP?

 

[dejo]

mine is similarly horrid on the memory write speeds. This is my first AMD cpu in many years, so I thought it was just normal to have that slow of writes. I have a feeling that this is not the Asus of old- where they tried to make the best products

 

[PolRoger]

Holy batman.

My 3900x with my MSI X570 Unify with my Crucial 32gb 3200mhz kit:

https://imgur.com/zv8nfXQ

My former 3800x with my former Asus X570 Tuf with my Gskill Neo 3600mhz kit:

https://imgur.com/yQXWosD

I really don't know what to say. Is Asus butchering the timings with DOCP?

mine is similarly horrid on the memory write speeds. This is my first AMD cpu in many years, so I thought it was just normal to have that slow of writes. I have a feeling that this is not the Asus of old- where they tried to make the best products

Seem kind of normal to me?...

All single "chiplet" Ryzen 3000 series (3800X and below) have lower writes in AIDA memory benchmark compared to double "chiplet" Ryzen 9 series (3900X/3950X). Also I believe the AIDA memory benchmark is multi threaded... So more cores/threads equal improved benchmark scores.

See the snip below from the OCF Zen2 review:

 

[dejo]

Nice catch. Reading FTW

 

[middy500]

I already knew about the write. I was shocked about the reads and copy. And anyways, my neo kit is running faster than it did of the Tuf. With different timings too. Right now I've been running it for the day with my 3800x. I haven't tested it yet with my 3900x.

Asus is screwing up the timings with DOCP. After googling it some more with the neo kit and Asus, I saw some people posting their timings DOCP was defaulting to and they were the same that mine was defaulting too. These where with people on different Asus x570 boards like the prime and even Asus b450 boards. Which tells me Asus is using the same DOCP list on their entire product stack.

 

[Bensam123]

Just registered to reply to this thread. I also have this very same issue and it is incredibly hard to troubleshoot as figuring out what to google is very difficult.

I own a 3900X, Asus X570 Tuf Wifi, and Trident Z Neo F4-3600C16D-32GTZNC

My motherboard currently works fine with DOCP, however manually changing timings for memory results in the above. It will either refuses to make it past the post phase and stick on the 'memory' LED or even when rebooting it will sometimes cause it. Power cycling doesn't fix it and I almost always need to reset the bios, restoring from a profile will once again cause it to work perfectly fine until it doesn't. It has absolutely nothing to do with system stability, for either memory or CPU overclocking. Memory voltage seems to have no bearing on whether or not this happens. One of the artifacts I found while manually OCing my CPU is that it happens more often. Specifically when increasing vcore it eventually got to a point where it was happening all the time (which is usually a power delivery issue). This was before I manually OC'd my memory, just using DOCP. Going over 1.39 vcore would cause it to lock almost every reboot. I had no idea what was going on.

While I was trying to figure out what was causing this (I thought my CPU OC wasn't stable), I found that manually setting PPT/TDC/EDC to maximum values caused it to happen less. I have no idea why as manually OCing supposedly disregards these values. Best I can figure is something is triggering a safeguard somewhere when it tries to pull too much wattage. Another thing that caused it to happen less frequently is setting memory VDDP Voltage to 1.1. It's normally set to .9. However this still seems to randomly happen regardless of how stable my system is.


So something else I want to share on this. Earlier this year when I bought a 3900x, I bought a second motherboard a month after I originally bought my system back in Sept. The reasoning for this at the time was it felt like my CPU wasn't boost high or long enough comparatively to other models on the market with PBO. Despite the recommendations this motherboard got for it's power delivery subsystem something felt off. I bought a GIGABYTE X570 AORUS ELITE WIFI as it also has a very nice power delivery system for the price.

I made a thread on this on Reddit and had nothing to do with memory. Unfortunately I was running out on a return window for my Tuff was closing and it's a huge PITA swapping motherboards for testing. What I found out while using the Aorus was that it boosted higher for longer (25-50mhz while gaming) and gave more consistent .1% and 1% frametimes (2fps on .1 and 5 on 1% in HotS). However Gigabytes bios has a lot to be desired and it's a huge turn off, especially when it comes to fan curves. Everything in the bios was all over the place and it took me a lot of effort to get the same settings dialed in I had on the Asus. I thought any of the differences were just due to early AGESA coding and opted to return the Gigabyte as I thought incorrectly at the time that Asus used a T-Topology for memory instead of Daisy Chained, it doesn't both are Daisy Chain. I had four dimms of memory (once again at the time) and thought that the Asus board would be a better value in the long run and AGESA would eventually get cleared up in later bios revisions.

Months after I sent it back it still felt like something was wonky even after the 1.0.0.4b AGESA update and still felt as though something was wrong. I also figured out my 3900x is a bit of a turd, both the craplet and the 'good' chiplet aren't very good, they're both mediocre. However this thread and my recent problems with my Neo kit makes me think the board itself might be just poor compared to other options. Matching this with Asus starting to artificially segment their products and regardless of how good the power delivery looks, there might be something else going on under the surface here. This thread is even more affirmation that I should really try the Gigabyte board again and see if my overclocking as well as memory tweaking reach the same results as on the Asus board. I have a sneaky suspicion they wont.

Just to clarify here, when I'm talking about 3900x OCing I'm talking specifically about CCX OCing.

 

[Balulex]

I asked this over at OCN but thought I would ask it here for help too. There seems to be about a 30% chance of my kit, Gskill F4-3600C16D-32GTZNC (on Asus and Gskill QVL for the X570 Tuf which I'm using), to halt at reboot. My Tuf has its DRAM LED light on and it just halts there until I hold down my power button. I never got to try any other bios below 1403 as the first thing I did when I got the board was flash 1403. Then 1404 when it came out and finally now 1405. All three bios versions have this issue.

The odd thing about it is that, so far, it has been exclusive to rebooting. Not shutdown and powering back on and not from a cold boot after pulling out the power cord. But just from rebooting. And when it happens, once I hold down my power button to shutdown, and boot it back up, it boots just fine like normal. No issues when gaming long sessions, did prime95 well when I did my initial benchmark stuff, etc. I've tried clearing cmos, reflashing my bios, manually setting FCLK to 1800 and SOC voltage to 1.1v (which is what Auto with DOCP enabled was defaulting to), etc.

I did see a guy on newegg for the X570 Tuf complain about something that sounds very similar to what I have been enduring (chance of halting at boot up with his 3200mhz kit since updating to 1404 after coming from 1203 that worked fine) so it lends me to believe its a Asus bios issue. The guy was hinting at it being a DOCP issue. I am running it with DOCP. So yesterday I decided to write down my timings that where being set with DOCP and manually set everything. But I ran into a two problems.

The first problem is the very last timing to set in the bios, tcke, was post to be set to 0 (as 0 was the value DOCP was setting on auto) but when I put in 0, it changes it to 1. So some how auto sets it to 0 but manually I can't. Ryzen Master also shows it as 0 as well when DOCP is enabled. The second problem was after setting all the timings, memory to 3600, FCLK to 1800, DRAM voltage to 1.35v's, and SoC voltage to 1.1v's, my computer can't post. It just halts 100% of the time on the DRAM led light. So I had to clear cmos to get back in. So either that very last timing is what's causing the manual setting to fail to post or DOCP is also enabling other stuff that I can't figure out what else. Even though everything else is on auto by default either with or without DOCP, I'm assuming DOCP changes what those auto settings default to.

I am using slots A2 and B2 as well and fast boot in windows is disabled.

So because of that it does have me a bit concerned that maybe those times where I fail to reboot are not due to an Asus bios bug, but maybe hardware? Since I still have a return window for both a refund and replacement I don't know if I should gamble it?

I've run for an hour and a half of Ramtest from Karhusoftware that I found on OCN and after 1575% coverage, no errors. After that I ran Memtest Pro 7.0 for a bit over an hour to reach a little over 100% coverage and no errors. Ran both yesterday.

What do you guys think? Anyone else with a similar problem? Do you think its just Asus bios bug or hardware?

I have the same problem as you and updating the bios to 1405 didn't fixed it. Froze almost every time I try to reset the pc. Had to force power down and then power up to work again. Also if i shut the pc down and turn it on the next day I it hangs again with the orange DDR led on the motherboard lit. But once it starts it works perfectly. This till today when I started the pc and it wont boot anymore with both ram sticks, just with one and only with the docp disabled. The other ram stick is not detected anymore. I hope is still working because I already applied for the motherboard return. I just bought the motherboard last week.

I have an Asus tuf gaming 570x Plus motherboard with a ryzen 3700x cpu and a 2x8 corsair vengeance rgb pro 3200 DDR4.

 

[HaVa]

I registered just to offer yet more confusion in the mix here.

Let's see...

I have ASUS Prime X570-P motherboard with Ryzen 3900X going. In the original build I used 2x8Gb Kingston 3200Mhz memory, on DOCP. Everything was just fine. Last Friday I got myself some G.Skill Trident Z Neo ddr4-3600 cl16-19-19-39 1.35v 2x16gb. I put them in, enabled DOCP. Evrything's fine. Gamed all through the evening and long into the night. Shut down the computer. Saturday, I turn it on. And... doesn't post. All the lights are on, fans are running... but peripherals aren't working. No keyboard leds, monitor doesn't connect, Xbox controller is mute. I press the reset button on the case, it starts up normally. And functions normally, just like with others in this thread.

So I thought it was some weird hiccup and entered the BIOS. Returned default memory settings, restarted, entered DOCP again. And again, everything's fine. This is Saturday evening. Everything works nicely until Tuesday afternoon, when this same weird non-POST hang happens (I always turn off the computer when I leave for work or go to bed, so between Saturday and Tuesday I'd turned it on and off around 4 times without issue). And again, pressing restart makes it go away. In my case, restarting doesn't seem to make it happen, only powering down and powering back up (I restarted the system around 11-13 times on Sunday when I was trying to work around the annoying 0db fan thing in latest Radeon drivers, that's why). Shutting down the computer seems a bit odd now, though. Earlier it took long enough for me to turn off the monitor and speakers, then a second or two later the computer would shut down. Now though... I can barely reach behind the monitor to turn it off before the system shuts down. Windows doesn't report any abrupt, failed system shutdowns, though.

So far I haven't done much else about it than update my BIOS to the 1405 version. I had the earlier 1404 before. I'm not a 100% sure but I kind of remember my SOC voltage having been 1.2 on the earlier BIOS, and since my CPU is now running 4-5 degrees colder on idle, the new BIOS seems to have lowered SOC to 1.1. I only got it set up before leaving for work, so I'll need a couple of days max to determine whether the new BIOS actually does anything (I'm not optimistic). From all the threads I've read regarding this issue, people suggest ramping up the SOC voltage instead to boost memory control or something. There's also a setting in the BIOS under Digi+, called VDDCR CPU Power Phase Control. For one reason or another, I had this set to extreme before I loaded factory defaults back. So I put it back on extreme now, in case it'll do something useful. But this seems to be such a major case of silicon lottery that anything's possible.

If it still does this, I'm going to try setting the FCLK to 1800Mhz manually. If there's still an issue, manual timings. And finally, going back to my old memory kit (and now I fear putting them back in case this same **** continues with them, too...)

 

[Johan45]

If it still does this, I'm going to try setting the FCLK to 1800Mhz manually. If there's still an issue, manual timings. And finally, going back to my old memory kit (and now I fear putting them back in case this same **** continues with them, too...)

For all of you that are using 3600 on an ASUS board. I haven't tested the newer BIOS files so I can't say for sure but during review, 3600 MHz was where ASUS was dropping to 1:2 compared to MSI/GIGA which did it at 3733 MHz. I think HaVa might be on to something here with setting the IMC ratio manually. I do know that when it jumps from 1:1 to 1:2 there is a shutdown and longer boot to accommodate the switch. Maybe the ASUS boards are getting hung up here almost confused and the "switch" is causing the issue?

 

[HaVa]

Could be, yeah. What would also speak for that being the issue is the fact that when I used my old memory's DOCP profile at 3200Mhz, it just put it on like that. With this 3600Mhz one, switching to DOCP displays a message window about FLCK=MLCK and FLCK capping at 1800Mhz. Also a warning about not overdoing SoC voltage. So maybe the automatic setting thinks it needs to go over the 1800Mhz FLCK limit (maybe something to do with the DOCP profile having 3603Mhz instead of a solid 3600Mhz). Still, got to see how it goes. So far though, when I've set the DOCP there hasn't been a longer boot or an initial shutdown like you said, because of a switch.

 

[Johan45]

It's worth checking in Windows with CPUz memory tab. If the NB readout is 1000 ish then it has switched to 1:2 mode. If you're running 1:1 then NB should read 1800