New 3950x build, random BSODs. RAM or HBA incompatible with board?

[DAOWAce]

EDIT: The RAM isn't supported on this motherboard at its rated timings. The stock BIOS was the original cause of instability in my OP; bad support -> some support after updating.

It runs at 2800MHz on auto timings, but setting any timing manually causes the system to fail to POST, no matter what other values I've tried to tweak to stabilize it (voltage, ohm, gear, clock rates, etc)

It's coming to near a year since I upgraded to Ryzen from Intel-E and ASUS have not supported this memory, even giving them SPD dumps, official bug reports through support filings, and contacting G.Skill to talk to ASUS directly.

The last time this happened with a motherboard, it was an MSI one. They gave me a BETA BIOS within a week to fix it.

Zen 3 is almost out. If the new BIOS still doesn't support my memory, I can't recommend any ASUS motherboard to anyone.

---
OP:

Built a new Ryzen system a few days ago with a 3950x and the VIII Hero. All other components were transferred from old system (1080 Ti, 32GB 2800MHz CL15 8Gx4 memory kit, X-Fi Titanium, Dell HBA330, multitude of SSDs/HDDs)

Everything on auto settings as the first thing I did after build was enable DOCP and it lead to a POST failure. I also have no reason to manually overclock the CPU right now (and based on research, probably ever).

On day 2 of the new system running, I suddenly BSOD'd the instant I started a Prime95 blend test.

I BSOD'd a few minutes after getting back into Windows.
I BSOD'd again right after the Windows boot screen on separate a fresh install (multi-boot system).
Removed 2 RAM sticks, everything seemed fine.

Next day, random BSOD. Removed RAM stick and HBA. Seemed fine again.

I'm right now running with 2 sticks in at its rated memory speed, but auto timings (2800MHz, CL19). It's coming around to 24 hours since the last BSOD.

The BSOD's I've gotten so far are:

ATTEMPTED_WRITE_TO_READONLY_MEMORY ; ntoskrnl.exe
ATTEMPTED_WRITE_TO_READONLY_MEMORY ; hal.dll
UNEXPECTED_KERNEL_MODE_TRAP ; peauth.sys
MEMORY_MANAGEMENT
IRQL_NOT_LESS_OR_EQUAL ; NTFS.sys
PAGE_FAULT_IN_NONPAGED_AREA ; clipsp.sys


It seems to be happening at complete random and only when booted or booting to an OS. I've yet to have anything happen while in BIOS.

I've run the Windows memory diagnostics in a few configs which came back negative. Ran some Windows tools (AIDA's system test, Prime95, Memtest by HCI), but nothing has reported a single error in the last few days.

Initially I thought my old RAM kit incompatible with Ryzen, but after researching for a while, it doesn't seem like Zen2 has anywhere near the same problems as first gen Ryzen, so memory compatibility should be a thing of the past. It reads the XMP profile as DOCP, but like stated above, the system fails to POST with it enabled, with various error codes (0E, 15, 33, 51). Manually setting it to 2800MHz is fine, but I cannot seem to lower the timings to rated specs without POST failure. Interestingly, if I set the RAM speed to above 2800MHz (overclocking), the board keeps it at 2800MHz. I can't for the life of me figure out why.


Either way, the common link between all the times I've had my BSODs is when my HBA has been installed.

I have 12 drives and this board only has 8 SATA ports (old X99 board had 10). Needless to say, I need my HBA now more than ever.

I'm still in the troubleshooting process (aka, waiting for a BSOD to happen), but from initial impressions, it seems like the HBA is somehow incompatible with the motherboard. With a BSOD at 4 DIMMs, 2DIMMs, 1 DIMM, and different modules every time, it does not seem like a memory issue.

Attempting to search/google about this is extremely difficult as people with HBAs in consumer systems are a tiny minority. I'm posting here hoping someone can give me any insight into this issue, or a way to ask someone at ASUS (more than entry-level techs) about this issue. Between this and AMD's onboard RAID not working at all with SATA drives (only sees PCI-E when BIOS is set to RAID), it's making me want to rebuild my old system and return everything.

[sean_s_adkins]

DAOWAce wrote:
Built a new Ryzen system a few days ago with a 3950x and the VIII Hero. All other components were transferred from old system (1080 Ti, 32GB 2800MHz CL15 8Gx4 memory kit, X-Fi Titanium, Dell HBA330, multitude of SSDs/HDDs)

Everything on auto settings as the first thing I did after build was enable DOCP and it lead to a POST failure. I also have no reason to manually overclock the CPU right now (and based on research, probably ever).

On day 2 of the new system running, I suddenly BSOD'd the instant I started a Prime95 blend test.

I BSOD'd a few minutes after getting back into Windows.
I BSOD'd again right after the Windows boot screen on separate a fresh install (multi-boot system).
Removed 2 RAM sticks, everything seemed fine.

Next day, random BSOD. Removed RAM stick and HBA. Seemed fine again.

I'm right now running with 2 sticks in at its rated memory speed, but auto timings (2800MHz, CL19). It's coming around to 24 hours since the last BSOD.

The BSOD's I've gotten so far are:

ATTEMPTED_WRITE_TO_READONLY_MEMORY ; ntoskrnl.exe
ATTEMPTED_WRITE_TO_READONLY_MEMORY ; hal.dll
UNEXPECTED_KERNEL_MODE_TRAP ; peauth.sys
MEMORY_MANAGEMENT
IRQL_NOT_LESS_OR_EQUAL ; NTFS.sys
PAGE_FAULT_IN_NONPAGED_AREA ; clipsp.sys


It seems to be happening at complete random and only when booted or booting to an OS. I've yet to have anything happen while in BIOS.

I've run the Windows memory diagnostics in a few configs which came back negative. Ran some Windows tools (AIDA's system test, Prime95, Memtest by HCI), but nothing has reported a single error in the last few days.

Initially I thought my old RAM kit incompatible with Ryzen, but after researching for a while, it doesn't seem like Zen2 has anywhere near the same problems as first gen Ryzen, so memory compatibility should be a thing of the past. It reads the XMP profile as DOCP, but like stated above, the system fails to POST with it enabled, with various error codes (0E, 15, 33, 51). Manually setting it to 2800MHz is fine, but I cannot seem to lower the timings to rated specs without POST failure. Interestingly, if I set the RAM speed to above 2800MHz (overclocking), the board keeps it at 2800MHz. I can't for the life of me figure out why.


Either way, the common link between all the times I've had my BSODs is when my HBA has been installed.

I have 12 drives and this board only has 8 SATA ports (old X99 board had 10). Needless to say, I need my HBA now more than ever.

I'm still in the troubleshooting process (aka, waiting for a BSOD to happen), but from initial impressions, it seems like the HBA is somehow incompatible with the motherboard. With a BSOD at 4 DIMMs, 2DIMMs, 1 DIMM, and different modules every time, it does not seem like a memory issue.

Attempting to search/google about this is extremely difficult as people with HBAs in consumer systems are a tiny minority. I'm posting here hoping someone can give me any insight into this issue, or a way to ask someone at ASUS (more than entry-level techs) about this issue. Between this and AMD's onboard RAID not working at all with SATA drives (only sees PCI-E when BIOS is set to RAID), it's making me want to rebuild my old system and return everything.

looks like your memory is unstable as hell with the current bios settings...relax the timings and see what effect it has. DOCP is unfortunately no good on mine either.....but its real easy to fix by tweaking/relaxing the ram timings (with docp selected)

[bluestang]

Maybe try using Thaiphoon Burner to see what chips those sticks use and then DRAM Calculator for Ryzen to get some "safe" settings to plug into your BIOS?

[DAOWAce]

Wow the site is horribly slow right now and multiquote isn't even working.

sean.s.adkins wrote:
looks like your memory is unstable as hell with the current bios settings...relax the timings and see what effect it has. DOCP is unfortunately no good on mine either.....but its real easy to fix by tweaking/relaxing the ram timings (with docp selected)

My memory is relaxed. I can't run it at its rated timings else it POST fails. I'm forced to keep it at auto to even use my system (which is 19-19-19-45 instead of 15-16-16-31).

I'd rather not keep resetting my BIOS and reconfiguring all my fans over and over and over again, so I haven't tested further.

bluestang wrote:
Maybe try using Thaiphoon Burner to see what chips those sticks use and then DRAM Calculator for Ryzen to get some "safe" settings to plug into your BIOS?

I already know they're Hynix chips. From further research, it seems Samsung chips are what Ryzen is most compatible with. There's quite a significant performance difference from 2800MHz to a tuned 3800MHz kit, but I'm definitely not letting my 32GB just go to waste in storage.

I suppose I'll give the DRAM calculator a try, as running these loose timings does hurt performance in some applications and games.

Still no explanation as to why the BIOS isn't increasing my memory above 2800MHz when I manually set it higher.


All this said, I still haven't BSOD'd since my post, and I re-added my HBA and 2 RAM sticks.

But, I also updated the BIOS, which updated the AGESA version to 1.0.0.4. Originally I didn't because of people reporting performance regressions and I wanted to see if there was truth to that, but it seems ASUS fixed something in the later BIOS which was causing me instability despite the changelog not mentioning anything relevant.

Only unfortunate thing I noticed with this board is that the PCI-E 16x slots are wired 16x, 8x, 4x. My HBA is 8x, and slotting it into the 8x slot would starve one of the GPU fans, which is unacceptable. I have no clue if setting the slots to operate at PCI-E 4.0 would remedy this (4.0 4x = 3.0 8x) or cause compatibility issues or do absolutely nothing.

Overall, I feel like while the 3900/3950x are great being on mainstream platforms to be more accessible.. they're really held back by the non-HEDT chipset/boards, so anyone coming over from an HEDT platform like myself is in for a very rough time, losing features we're used to, even on these enthusiast class boards (which don't even have PS/2 ports, while my HEDT did.. and I need them to even use Windows 7 again thanks to Ryzen having zero USB drivers in Win7; thanks AMD and ASUS!). Just, a sub-par experience, but upgrading to threadripper is not only much too expensive in comparison, it's also pointless for what I do.