Possibility of SR-IOV working with a Flex 170 cross-flash on an A770 16GB?

[arch_btw]

Hey all, I've seen claims of SR-IOV working on the A770 16GB with a Flex 170 cross-flash, but I haven't found any tutorial nor evidence of such claim. Could there be any truth to this and if so, how would one go about cross-flashing the datacenter Flex 170 firmware to an A770 16GB? All I know is both the datacenter Flex 170 gpu and the A770 use Intel's DG2 architecture.

I have the Flex 170 Firmware but from what I understand, it only has the main firmware payload without the OPROM, and I don't know if cross-flashing only the main firmware payload without the matching OPROM would break things.

I've attached the Flex 170 firmware below, any help is much appreciated.

 

[abrfilho]

I flashed once an older firmware with the newest OPROM, there were no problem with that, I did the opposite too, but it was with the regular firmware file.

 

[arch_btw]

I flashed once an older firmware with the newest OPROM, there were no problem with that, I did the opposite too, but it was with the regular firmware file.

That's great news! Now I'm just concerned about the cross-flash part.. Do you think I would be able to restore the GPU from a firmware backup if the cross-flash ended in a failure? I have a backup GPU I could use if that would work?

 

[abrfilho]

That's great news! Now I'm just concerned about the cross-flash part.. Do you think I would be able to restore the GPU from a firmware backup if the cross-flash ended in a failure? I have a backup GPU I could use if that would work?

I can't say for sure...
@Solaris17 have a thread with all firmwares for the card and other thread with the tool that he developed to flash the firmware, if I'm not mistaken, it's not possible to flash a file that it's not compatible, I think he mentions this fact in the thread.

 

[Solaris17]

I did there are safeguards to prevent cross flashing. The bios’s are also packaged and unpacked during flash. So you can’t just take a vbios provided in the driver and flash it directly to the chip with an eeprom reader. Like wise; you cannot take a dump with an eeprom reader and flash it with the tool.

I haven’t had time to look at this. Where was this vbios from?

 

[arch_btw]

I did there are safeguards to prevent cross flashing. The bios’s are also packaged and unpacked during flash. So you can’t just take a vbios provided in the driver and flash it directly to the chip with an eeprom reader. Like wise; you cannot take a dump with an eeprom reader and flash it with the tool.

I haven’t had time to look at this. Where was this vbios from?

I didn't obtain the firmware myself, someone on level1tech forums claimed to have downloaded it from the following link:

https://software.cisco.com/download/home/286331842/type/283850974/release/4.3(2.230270)

I can't download it myself from that link though, I guess it requires a cisco account or something?

Anyways, would I be able to use Intel's CLI tool to cross-flash instead?

 

[Solaris17]

Anyways, would I be able to use Intel's CLI tool to cross-flash instead?

No I added a file checker to the tool for just this purpose. Checking unknown files.

The IGSC doesnt understand this firmware at all. Any of them.

Spoiler

Checking selected file...

Checking file:
 "D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin"

Checking if we can identify this image type...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\oprom_parser.c:image_oprom_parse():565) Illegal oprom image pci header
Error: Unknown image type: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if FW...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
Error: Cannot retrieve firmware version from image: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if Oprom-Data...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\oprom_parser.c:image_oprom_parse():565) Illegal oprom image pci header
Error: Invalid image format: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if Oprom-Code...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\oprom_parser.c:image_oprom_parse():565) Illegal oprom image pci header
Error: Invalid image format: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if FW-Data...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
Error: Invalid image format: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Finished checking file!

It wont pass the checks for flashing.

 

[arch_btw]

No I added a file checker to the tool for just this purpose. Checking unknown files.

View attachment 329843

The IGSC doesnt understand this firmware at all. Any of them.

Spoiler

Checking selected file...

Checking file:
"D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin"

Checking if we can identify this image type...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\oprom_parser.c:image_oprom_parse():565) Illegal oprom image pci header
Error: Unknown image type: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if FW...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
Error: Cannot retrieve firmware version from image: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if Oprom-Data...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\oprom_parser.c:image_oprom_parse():565) Illegal oprom image pci header
Error: Invalid image format: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if Oprom-Code...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\oprom_parser.c:image_oprom_parse():565) Illegal oprom image pci header
Error: Invalid image format: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Checking if FW-Data...

IGSC: (D:\OneDrive\Documents\Projects\repos\igsc\lib\igsc_lib.c:gsc_fwu_img_layout_parse():343) Invalid FPT header marker (0xb17eb974)
Error: Invalid image format: D:\OneDrive\Downloads\Intel\Intel Flex 170\DG02_1.3260\XPUM_ATS_M150_512_C0_ES_106_gfx_DG02_1.3260_fwupdate.bin


Finished checking file!

It wont pass the checks for flashing.

Welp, it was worth trying I guess.. Thanks anyway.

 

[abrfilho]

@arch_btw if you want to risk, try this file: Intel® Data Center GPU Flex Series - Windows*
You'll extract the driver, navigate to the Graphics folder, inside there'll be two folders, fw (containing the firmware file) and oprom (containing only the oprom code).
FW Version: DG02_1.3253
OPROM CODE Version: 14 00 28 04 00 00 00 00

 

[Solaris17]

@arch_btw if you want to risk, try this file: Intel® Data Center GPU Flex Series - Windows*
You'll extract the driver, navigate to the Graphics folder, inside there'll be two folders, fw (containing the firmware file) and oprom (containing only the oprom code).
FW Version: DG02_1.3253
OPROM CODE Version: 14 00 28 04 00 00 00 00

For what its worth I tested this and it works just fine.

I cant see anything specifying that it is for flex only. Its just an older fw version. I also cant speak to if it enabled srv-io. Which is more going to be the magic inside the ATS or AMC firmware. It did however seem to disable the audio device, which seems like it has been giving proxmox users greif. Though honestly its likely because the firmware was just flashed and changed. I have seen this before.

Before:

https://teamdotexe.org/fwtool/ARC-Flash-log-3ccmbNcfLV2A.txt

After:

https://teamdotexe.org/fwtool/ARC-Flash-log-yxSAYRQmbUpN.txt

I will see if I can assist in this further, but I haven't even physically seen a DC gpu and I am working on other aspects of the tool additionally my test station for flashing is my AMD system which does not have SRV-IO so I am not certain how I would prove any kind of functionality.

If anyone comes across this and has a method to test other than api/smi (which appears to fail on amd systems??) I am all ears.

I didn’t even know this was a thing people were trying to solve.