Firmware icon indicating copy to clipboard operation
Firmware copied to clipboard
verified publisher icon loongson

[XA612A0] With XFX RX 580 the BIOS stucks at [AmdgpuGop]

Open RESBI opened this issue 1 year ago • 16 comments

With my XFX RX 580 plugged on the XA612A0 board, UART reads it stuck at

...
debug: pcie_init_time 2 ret 0x43 data 0x83
dev 16 init begin...
syn0 Mac is invalid, Please update spi mac address
 ------ this is ls132 -------
ls132 version: V1.1.1-R231018.
ls132 init done
Publish Memory Initialize done.
Find AMD GPU device:0x67DF
Install GopDevicePath Handle FC1F1718, Success
AMD GPU Frame Buffer Found.
add ip block number 0 <vi_common>
add ip block number 1 <dce_v11_0>
ATOMBIOS detected
PCI I/O BAR is not found. Using MMIO to access ATOM BIOS
ATOM BIOS:113-P20_XTX_D58GB_HY5GQ_180203_W8
ASSERT [AmdgpuGop] /data/gaoqihang/uefi-loongarch-cleanscreen/MdePkg/Library/BaseLib/SafeString.c(1839): (DestMax > SourceLen)

No display signal, no disk reading activities, and had no further outputs.

I've tried all the firmwares for both XA612A0 and XA61200, got the same result. Changing PCIe settings got no effects.

The GPU itself, the board, and the PCIe slot have all been tested, functioning normally.

Here's UART readouts when using the latest firmwares of XA612A0 and XA61200 on XA612A0

XA612A0-with-XA612A0BIOS_prestable2405rel.txt

XA612A0-with-XA61200BIOS_prestable2405_0523rel.txt

RESBI avatar Mar 10 '25 16:03 RESBI

Sorry for this bug. This root cause of this problem is the atombios name is longer(113-P20_XTX_D58GB_HY5GQ_180203_W8, 33 bytes) than I assumed(32 bytes). I'll give a provisional version later, please test with it again.

EricGao2015 avatar Mar 11 '25 06:03 EricGao2015

Thanks for your work! That's really out of mind...

BTW, would we have overclocking options on XA612A0? I use the 00's firmware on A0 for overclocking, which now causes GbE port onboard not working.

RESBI avatar Mar 11 '25 08:03 RESBI

Thanks for your work! That's really out of mind...

BTW, would we have overclocking options on XA612A0? I use the 00's firmware on A0 for overclocking, which now causes GbE port onboard not working.

UEFI-Firmware-fix-amd-20250313.tar.gz

Please test with the two files. 4c1909d1685a2523ce0f18645ed4aaa9 UEFI_3A6000x1-7A2000x1-XA61200.fd 091b375a15cc74b00f7a40ed289d85fb UEFI_3A6000x1-7A2000x1-XA612A0.fd

About overclocking, you can ask @MarsDoge for help.

EricGao2015 avatar Mar 13 '25 08:03 EricGao2015

Since the voltage regulation method of A0 does not comply with the AVS standard compared to 00, there is no plan to support A0 overclocking options for the time being.

MarsDoge avatar Mar 13 '25 10:03 MarsDoge

Thanks for your work! That's really out of mind... BTW, would we have overclocking options on XA612A0? I use the 00's firmware on A0 for overclocking, which now causes GbE port onboard not working.

UEFI-Firmware-fix-amd-20250313.tar.gz

Please test with the two files. 4c1909d1685a2523ce0f18645ed4aaa9 UEFI_3A6000x1-7A2000x1-XA61200.fd 091b375a15cc74b00f7a40ed289d85fb UEFI_3A6000x1-7A2000x1-XA612A0.fd

About overclocking, you can ask @MarsDoge for help.

Thank you! I'll test it once I touched my machine!

RESBI avatar Mar 14 '25 10:03 RESBI

Since the voltage regulation method of A0 does not comply with the AVS standard compared to 00, there is no plan to support A0 overclocking options for the time being.

Thanks! I see there're errors on reading Vddn and Vddp value, which seems did not happens on 00 boards.

My A0 is running at 2.65 GHz now with my modified A0 firmware, but it will get bus error when running xhpl tests, so I wanna raise it's Vcore a little bit. Is there any way to config it? I didn't find the corresponding codes...

RESBI avatar Mar 14 '25 10:03 RESBI

Thanks for your work! That's really out of mind... BTW, would we have overclocking options on XA612A0? I use the 00's firmware on A0 for overclocking, which now causes GbE port onboard not working.

UEFI-Firmware-fix-amd-20250313.tar.gz

Please test with the two files. 4c1909d1685a2523ce0f18645ed4aaa9 UEFI_3A6000x1-7A2000x1-XA61200.fd 091b375a15cc74b00f7a40ed289d85fb UEFI_3A6000x1-7A2000x1-XA612A0.fd

About overclocking, you can ask @MarsDoge for help.

Thanks for your works! I had touched my machine for a while, and tested the A0's firmware roughly, which is running well now.

But I meet some problem on 6.13.3-aosc kernel, which seems trying to launch LG110 driver (I didn't confirm this), then stuck at before the sddm launched.

Image Image

RESBI avatar Mar 15 '25 11:03 RESBI

But I meet some problem on 6.13.3-aosc kernel, which seems trying to launch LG110 driver (I didn't confirm this), then stuck at before the sddm launched.

You can add modprobe.blacklist=loonggpu in the kernel parameters to avoid starting LG110 for verification.

MarsDoge avatar Mar 15 '25 14:03 MarsDoge

According to our design, if there is an independent graphics card RX580 (03:00.*), I think loonggpu (00:06.*) should not be seen. Please check whether the AMD model does not match our scan list when you have time? @EricGao2015

MarsDoge avatar Mar 15 '25 14:03 MarsDoge

According to our design, if there is an independent graphics card RX580 (03:00.), I think loonggpu (00:06.) should not be seen. Please check whether the AMD model does not match our scan list when you have time? @EricGao2015

Yes with the 6.13.2-aosc-main kernel it would not list 00:06.* when booted, I will check 6.13.3-aosc-main next time. Image

RESBI avatar Mar 15 '25 14:03 RESBI

According to our design, if there is an independent graphics card RX580 (03:00.), I think loonggpu (00:06.) should not be seen. Please check whether the AMD model does not match our scan list when you have time? @EricGao2015

@RESBI This issue have been resolved in stable202402, please check your verison.

EricGao2015 avatar Mar 17 '25 06:03 EricGao2015

According to our design, if there is an independent graphics card RX580 (03:00.), I think loonggpu (00:06.) should not be seen. Please check whether the AMD model does not match our scan list when you have time? @EricGao2015

@RESBI This issue have been resolved in stable202402, please check your verison.

Thank you! But I didn't find stable202402 under the XA612A0 folder, do you mean the one for XA61200?

RESBI avatar Mar 17 '25 08:03 RESBI

According to our design, if there is an independent graphics card RX580 (03:00.), I think loonggpu (00:06.) should not be seen. Please check whether the AMD model does not match our scan list when you have time? @EricGao2015

@RESBI This issue have been resolved in stable202402, please check your verison.

Thank you! But I didn't find stable202402 under the XA612A0 folder, do you mean the one for XA61200?

stable202402 is not provided in XA612A0 folder.

EricGao2015 avatar Mar 17 '25 09:03 EricGao2015

According to our design, if there is an independent graphics card RX580 (03:00.), I think loonggpu (00:06.) should not be seen. Please check whether the AMD model does not match our scan list when you have time? @EricGao2015

@RESBI This issue have been resolved in stable202402, please check your verison.

Thank you! But I didn't find stable202402 under the XA612A0 folder, do you mean the one for XA61200?

stable202402 is not provided in XA612A0 folder.

I now got what you mean, thank you!

Now I meet a serious system crushing problem, I'll suddenly lose all network connections to it when I'm using it normally through VNC and have to do a hard reboot, which doesn't happens before. I can't sure what caused this problem (Firmware unstable, GPU reset, Linux kernel unstable, or simply network cable not plugged well...), I'll try to figure out.

RESBI avatar Mar 17 '25 09:03 RESBI

Thanks! There is no problem now.

It's the low VDDN causing the CPU unstable at higher frequency and higher Node Clock, everything is fine now.

RESBI avatar Mar 19 '25 05:03 RESBI

XA612A0 stucks at the long BEEP when using this 00's firmware, and no uart outputs.

RESBI avatar Mar 24 '25 06:03 RESBI