Platform test configuration file#
2025-11-12
2 min read time
The platform host check scripts can validate various host-side configurations, such as model load balancing enablement, PCIe checks, kernel version, and ROCm version.
Here’s a code snippet of the host_config.json file for reference:
host_config.json
{
"host":
{
"os_version": "Ubuntu 24.04.1 LTS",
"kernel_version": "6.8.0-60-generic",
"rocm_version": "6.4.1",
"bios_version": "20171212",
"pci_realloc": "off",
"online_memory": "1.3T",
"gpu_count": "8",
"gpu_pcie_speed": "32",
"gpu_pcie_width": "16",
"fw_dict":
{
"CP_MEC1": "32945",
"CP_MEC2": "32945",
"RLC": "65",
"SDMA0": "24",
"SDMA1": "24",
"VCN": "09.11.70.09",
"RLC_RESTORE_LIST_GPM_MEM": "4",
"RLC_RESTORE_LIST_SRM_MEM": "4",
"RLC_RESTORE_LIST_CNTL": "4",
"PSP_SOSDRV": "00.36.02.56",
"TA_RAS": "1B.36.02.14",
"TA_XGMI": "20.00.00.14",
"PM": "07.85.11.01"
}
}
}
Parameters#
Here’s an exhaustive list of the available parameters in the Platform configuration file.
Configuration parameters |
Default values |
Description |
|---|---|---|
|
Ubuntu 24.04.1 LTS |
Version of OS |
|
|
Version of kernel |
|
7.0.2 |
ROCm version |
|
|
BIOS version |
|
Off |
PCI reallocation |
|
1.3T |
Available system RAM |
|
8 |
Number of GPUs |
|
32 |
PCIe speed |
|
16 |
Width of PCIe |
|
32945 |
Compute Pipeline MicroEngine Controller 1 firmware |
|
32945 |
Compute Pipeline MicroEngine Controller 2 firmware |
|
65 |
RunList Controller firmware |
|
24 |
System DMA Engine 0 firmware |
|
24 |
System DMA Engine 1 firmware |
|
|
Video Core Next firmware |
|
4 |
RunList Controller restore mechanisms for power state transitions |
|
4 |
RunList Controller restore mechanisms for power state transitions |
|
4 |
RunList Controller restore mechanisms for power state transitions |
|
|
Platform Security Processor SOS driver |
|
|
Trusted application for RAS (Reliability, Availability, and Serviceability) |
|
|
Trusted application for xGMI (External Global Memory Interconnect) |
|
|
Power Management firmware |