Did a couple more tests.
First same 5900XT and 2x32 ECC DDR4 3200 but on a x570 motherboard - all OK. So the CPU can work with the memory on another motherboard.
Then again with B450D4U-V1LQ5:
Added extra fan blowing directly to VRM heatsink. Heatsink now noticeably cooler. Still crash.
Same...
Before crashing memtest shows the CPU max temperature I think around 78C.
VRM temperature during test I have not checked, will try to do that with a thermocamera.
Otherwise the system is in a closed 2U case with Dynatron A47 cooler and 2x 80mm case fans. Fan settings are on auto in BIOS and...
When the crash occurs, fan speeds go down, power usage sinks to 20W (from around 33W when idle). And the power button will not shut down the machine. One has to pull the power cord out to boot again. Googling around I got the impression this points to IMC crash. Lowering the ram speed from 1600...
Two bits of information:
- I managed to read out fan RPMs in Linux on this board. All info is here:
https://github.com/AndrewLinden/nct6779_ahb
- I made a new version m3 of modified BIOS that now has hopefully all CBS submenus available (to for example set the SOC voltage) Find it here...
After bios mod I did a quick compare of SCELNX output, and to me it seems that the settings were the same between original and modded bios. But only with modded bios ECC would be activated upon boot.
Yes I saw your dmidecode data with UUID similar to what I had.
Regarding to UUID I found this...
Small update from me:
Unsuppressed "Platform first error reporting" setting. This should be disabled for ECC reporting to work.
I do not have Memtest paid version to test error injection.
When I have some time I plan to test other methods to induce errors (memory overclock, pin shortening...
Well it was not enough for my test board. If you are brave try reset to defaults in bios and see if you can enable ECC again with just setup_var.efi ;)
I will share of course!
I am at the moment checking and testing some things:
1. Do these motherboards have anything unique on the bios flash...
Since the chip is socketed I use a socket on the programmer.
See here [Guide] Flash BIOS with CH341A programmer
Section "SOP8 Socket details".
But you can connect a loose chip to the clip adapter too.
If you get a dump of you chip, theoretically this would enable to clone the ECC activation to...
@luckylinux Do you have the possibility to put the flash chip from your motherboard with active ECC to a SPI flash programmer and dump the contents of the chip to a file?
/root/.cargo/bin/uefisettings hii show-ifr -f extract-db.db > show-ifr-from-extracted-db.txt
Will produce output in unique sorting order every run. So that if you run it twice without changing any settings you get different files.
So out of the box it is not possible to compare settings of two...
So you are using files generated with this method?
When I took a look at some files extracted that way that also had "ECC Enable" in them, I saw that at least these two were not correct. Read through these two text files that I quoted below. The content does not make any sense to me.
Edit 1...
Here are my files.
Also lets compare our CPUs. Here is beginning of /proc/cpuinfo of mine:
vendor_id : AuthenticAMD
cpu family : 25
model : 33
model name : AMD Ryzen 7 5700X 8-Core Processor
stepping : 2
microcode : 0xa201210
I ran your script.
Basically most that differs from your result is boot entries that depend what drives are connected.
Seems your script does not read anything from Setup or PBS.
If we want to compare settings, I think you have to send me SCELNX_64 output.
1. disconnect network and any nvme/ssd/hdd
2. boot up any live linux usb stick
3. if that live linux usb has a r/w partition put SCELNX_64 there (no need for second usb stick), if it does not have r/w partition put SCELNX_64 on another usb stick
What exactly are you afraid of when running it on a system without network and without harddrives only with a usb stick with a bootable linux and another usb stick with a r/w partition with SCELNX_64 ?
That SCELNX kinda does that.
Download it from here: tools/SCELNX64 at master · doublechiang/tools
Then run it like this:
./SCELNX_64 /o /s nvram.txt /a /v
Then upload nvram.txt somewhere.
I will try to compare the settings you currently have with active ECC to mine without ECC active
Just for clarity: I did the reset to default only to verify that our thinking what is needed for ECC activation is correct.
And now I can't get ECC active again. And yes I have set 0xD4
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.