Revision History
|
Revision |
Date |
Description |
|---|---|---|
|
C2 |
September 15, 2025 |
Updated Firmware Files. |
|
C1 |
July 22, 2025 |
Updated Open Caveats in Release 1.1.0.250025. |
|
C0 |
July 03, 2025 |
Created release notes for 1.1.0.250025 for the following servers:
|
|
B1 |
June 19, 2025 |
Updated: |
|
B0 |
June 11, 2025 |
Created release notes for 1.1.0.250022 for the following servers:
|
|
A0 |
May 02, 2025 |
Created release notes for 1.1.0.250016 for the following servers:
|
Overview of the Cisco UCS C885A M8 Rack Server
The Cisco UCS C885A M8 Rack Server is a dense-GPU server designed to deliver massive, scalable accelerated compute capabilities to address the most demanding AI workloads, including deep learning/Large Language Model (LLM) training, model fine-tuning, large model inferencing, and Retrieval-Augmented Generation (RAG).
To deliver massive accelerated compute performance in a single server, the server offers a choice of eight GPUs of the following types:
-
NVIDIA® HGX H200 Server PCI Exrepess Module (SXM) or Nvidia HGX H100 SXM GPUs. SXM is a socket-based GPU interconnect method used by NVIDIA GPUs.
-
AMD MI300X OCP Accelerator Model (OAM) GPUs.
For north-south traffic, the server also supports up to two NVIDIA® BlueField B3220 DPUs or 2x200G ConnectX-7 NICs to scale AI model training across a cluster of dense-GPU servers. Up to eight NVIDIA® 1x400G ConnectX-7 or Bluefield-3 B3140H SuperNIC are supported for east-west traffic between GPUs.
Introduction
The Cisco Baseboard Management Controller (Cisco BMC) web GUI is HTML5 based and added security with SSL (HTTPS). It helps you manage the Cisco UCS C885A M8 Rack Server using the following options:
Hardware and Component Management
The Inventory feature enables administrators to record hardware devices and components on each server, such as central processing units (CPUs), memory modules, hard drives, network cards, and more.
Status and Checks
The Inventory feature also provides status and checks for hardware and software devices. This information can include device health status, temperature, voltage, connection status, and more.
What's New
Release 1.1.0.250016
-
Device Connector—Enabling the device connector establishes a bi-directional communication between Cisco Intersight (the cloud-hosted server management system) and Cisco Baseboard Management Controller. It is enabled by default. When enbled, you can manage the Cisco UCS C885A M8 Rack Server from Cisco Intersight.
-
Support for IPv6.
-
Support to add the configuration from an external server of virtual media feature.
-
Tech Support log files are now saved in .tgz format.
-
Support to configure BIOS tokens from
Supported Platforms
The following servers are supported in 1.1.0.250016 and later releases:
-
Cisco UCS C885A M8 Rack Server
Operating System and Browser Requirements
Cisco recommends the following browsers:
|
Recommended Browser |
Version Tested |
Minimum Recommended Operating System |
|---|---|---|
|
Mozilla Firefox |
132.0.2 (AArch64) |
macOS 15.1 (24B83) |
|
132.0 (64-bit) |
Ubuntu 20.04.3 LTS |
|
|
132.0.2 (64-bit) |
Microsoft Windows 11 Enterprise |
|
|
Apple Safari |
Version 18.1 (20619.2.8.11.10) |
macOS 15.1 (24B83) |
|
Google Chrome |
131.0.6778.71 (64-bit) |
Microsoft Windows 11 Enterprise |
|
Microsoft Edge |
131.0.2903.51 (64-bit) |
Microsoft Windows 11 Enterprise |
Default Ports
Following is a list of server ports and their default port numbers:
|
Port Name |
Port Number |
|---|---|
|
HTTP |
80 |
|
HTTPS |
443 |
|
SSH |
22 |
|
IPMI |
623 |
Upgrade and Downgrade Guidelines
Upgrade and Downgrade Guidelines for Release 1.1.0.250016 or any Later Release
-
Once you upgrade to release 1.1.0.250016 or any later release, you cannot downgrade to any release earlier than 1.0.1.250002.
-
After the firmware update completes, perform an AC power cycle to activate and complete the upgrade. See Cisco Baseboard Management Controller Configuration Guide for Cisco UCS C885A M8 Rack Server, Release 1.1 on how to perform an AC power cycle.
|
Upgrade from Release |
Recommended Upgrade Path |
|---|---|
|
Upgrade from release 1.0.1.250002 |
Direct upgrade to release 1.1.0.250016 or any later release. |
|
Upgrade from release 1.0.0.250003 |
|
|
Upgrade from release 1.0.0.240001 |
|
Firmware Files
Always upgrade all the components BIOS, BMC, and FPGA. Do not upgrade individual components (only BIOS or only BMC), since this could lead to unexpected behavior. If the BIOS and the BMC versions are from different container releases, it could result in unexpected behavior.
 Note |
The Host Upgrade Utility (HUU) firmware for the server is not included within the main firmware bundle. HUU firmware is a distinct component, managed under an independent release cycle, and is available for separate download. |
Firmware Files in Release 1.1.0.250025
The 1.1.0.250025 software release includes the following software files:
|
CCO Software Type |
File Name |
|---|---|
|
Firmware for OOB (BMC, BIOS, FPGA, and GPU). |
ucs-c885a-m8-1.1.0.250025.tar |
|
Host Upgrade Utility support firmware upgrade of ConnectX7, Bluefield, and OCP. HUU Version—1.0.1.250005 |
ucsc-885a-m8-huu-host-1.0.1.250005.iso |
|
Upgrade script |
ucs-c885a-m8-upgrade-v1.3.py |
|
GPU |
Minimum Firmware Version |
|---|---|
|
H100 / H200 Nvidia GPU |
HGX-22.10-1-rc80 |
|
AMD MI 300 GPU |
01.25.00.03 |
Note |
The versions provided here are the minimum supported versions that are compatible with the server with the current release. |
Firmware Files in Release 1.1.0.250022
The 1.1.0.250022 software release includes the following software files:
|
CCO Software Type |
File Name |
|---|---|
|
Firmware for OOB (BMC, BIOS, FPGA, and GPU) and In Band (ConnectX7, Bluefield, and OCP) components. |
ucs-c885a-m8-1.1.0.250022.tar |
|
Upgrade script |
ucs-c885a-m8-upgrade-v1.2.py |
|
GPU |
Minimum Firmware Version |
|---|---|
|
H100 / H200 Nvidia GPU |
HGX-22.10-1-rc80 |
|
AMD MI 300 GPU |
01.25.00.03 |
Note |
The versions provided here are the minimum supported versions that are compatible with the server with the current release. |
Firmware Files in Release 1.1.0.250016
The 1.1.0.250016 software release includes the following software files:
|
CCO Software Type |
File Name |
|---|---|
|
Firmware for OOB (BMC, BIOS, FPGA, and GPU) and In Band (ConnectX7, Bluefield, and OCP) components. |
ucs-c885a-m8-1.1.0.250016.tar |
|
Upgrade script |
ucs-c885a-m8-upgrade-v1.2.py |
|
GPU |
Minimum Firmware Version |
|---|---|
|
H100 / H200 Nvidia GPU |
HGX-22.10-1-rc80 |
|
AMD MI 300 GPU |
01.25.00.03 |
Note |
The versions provided here are the minimum supported versions that are compatible with the server with the current release. |
Bundle Information for Release 1.1.0.250025
The ucs-c885a-m8-1.1.0.250025.tar bundle contains the following:
|
Category |
Subcategory |
File Name |
|---|---|---|
|
Host |
bluefield-3 |
|
|
connectx7 |
|
|
|
x710-ocp |
|
|
|
m2-nvme-raid-ctrl |
|
|
|
OOB |
bios |
AMD_EPYC_5th_Gen
AMD_EPYC_4th_Gen
|
|
bmc |
AMD
NVIDIA
|
|
|
fpga |
dcscm
hib
hpm
|
|
|
gpu |
AMD
NVIDIA
|
Bundle Information for Release 1.1.0.250022
The ucs-c885a-m8-1.1.0.250022.tar bundle contains the following:
|
Category |
Subcategory |
File Name |
|---|---|---|
|
Host |
bluefield-3 |
|
|
connectx7 |
|
|
|
x710-ocp |
|
|
|
m2-nvme-raid-ctrl |
|
|
|
OOB |
bios |
AMD_EPYC_5th_Gen
AMD_EPYC_4th_Gen
|
|
bmc |
AMD
NVIDIA
|
|
|
fpga |
dcscm
hib
hpm
|
|
|
gpu |
AMD
NVIDIA
|
Bundle Information for Release 1.1.0.250016
The ucs-c885a-m8-1.1.0.250016.tar bundle contains the following:
|
Category |
Subcategory |
File Name |
|---|---|---|
|
Host |
bluefield-3 |
|
|
connectx7 |
|
|
|
x710-ocp |
|
|
|
m2-nvme-raid-ctrl |
|
|
|
OOB |
bios |
AMD_EPYC_5th_Gen
AMD_EPYC_4th_Gen
|
|
bmc |
AMD
NVIDIA
|
|
|
fpga |
dcscm
hib
hpm
|
|
|
gpu |
AMD
NVIDIA
|
Resolved Caveats
Resolved Caveats in Release 1.1.0.250025
The following defects are resolved in Release 1.1.0.250025:
|
Defect ID |
Symptom |
First Affected Release |
Resolved in Release |
|---|---|---|---|
|
CSCwo82679 |
A critical SEL error, Temperature sensor reading error assert, SensorName: TEMP_HIB_PEX1 is observed during a power cycle. Similar events are also noted for sensors PEX1 to PEX4, along with equivalent deassert events. This issue is resolved. |
1.1.0.250016 |
1.1.0.250025 |
|
CSCwo89927 |
Management IP is lost on the Shared_nic eth1 interface after a system warm reboot, impacting both DHCP and static environments. Symptoms include loss of BMC IP communication when the system is configured with the eth1 network interface. This issue is resolved. |
1.1.0.250025 | 1.1.0.250025 |
|
CSCwo76759 |
Cisco BMC is unable to dynamically re-learn the network gateway MAC address when the default gateway MAC is not set or has changed. As a result, while it can communicate within the same subnet, it cannot respond to traffic from a different subnet unless the default gateway MAC is manually configured. This may cause the Cisco BMC to become unreachable outside the subnet when switching between static environments with different subnets. This issue is resolved. |
1.0.0.240001 |
1.1.0.250025 |
|
CSCwp26271 |
The PCIe switches have Peer-to-Peer (P2P) disabled, which may affect GPU performance when using AMD MI300 GPU accelerators and Cisco CX7 PCIe switches in high-performance computing environments. This setting can influence data transfer efficiency between GPUs, leading to performance differences. This issue is resolved. |
1.0.0.240001 |
1.1.0.250025 |
Resolved Caveats in Release 1.1.0.250022
The following defects are resolved in Release 1.1.0.250022:
|
Defect ID |
Symptom |
First Affected Release |
Resolved in Release |
|---|---|---|---|
|
CSCwn85582 |
The FHHL adapter goes into disable state. This issue occurs when adapter is set as InfiniBand. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo68356 |
NVIDIA ConnectX CX7 (4x25) adapter details are not displayed correctly on the Inventory. Additionally, the overall server status and temperatures are incorrectly shown as Critical when the system is connected with the ConnectX CX7 (4x25) adapter. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo70872 |
The TEMP_FH13_Chip and TEMP_FH13_TRX sensors may show critical alerts for empty riser slots after upgrading to release 1.1.0.250016 and powering on the host. There is no impact on functionality. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo89930 |
In a setup equipped with MI300 GPU, the OS may hang during reboot or power cycle. This issue is resolved. |
1.1.0.250022 | 1.1.0.250022 |
|
CSCwo68052 |
Network Adapter and Processor details are not fully displayed in the Cisco BMC GUI when cross-launched from Cisco Intersight using Read-Only user accounts. You may experience incomplete data visibility in this scenario. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo68698 |
When initiating a logout action from the KVM cross-launched via Cisco Intersight, an exit confirmation icon labeled Leave repeatedly appears in the browser window. This behavior prevents you from closing the browser and disrupts the experience. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo71111 |
Warning SEL logs indicate that multiple services, including phosphor-ipmi-host.service and bmcweb.service, exited unsuccessfully during network configuration changes. This occurs when switching the static network configuration from eth0 to eth1. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo81255 |
The mapped vMedia is ejected upon Web GUI session timeout, causing OS installation on x86 CPUs to fail. This issue is resolved. |
1.1.0.250016 |
1.1.0.250022 |
|
CSCwo81247 |
The GUI session times out every 30 minutes, even during active use, causing interruptions to ongoing tasks. This behavior affects long-running activities, such as OS installations, as the session timeout leads to the disconnection of mounted vMedia. This issue is resolved. |
1.0.0.240001 |
1.1.0.250022 |
Resolved Caveats in Release 1.1.0.250016
The following defects are resolved in Release 1.1.0.250016:
|
Defect ID |
Symptom |
First Affected Release |
Resolved in Release |
|---|---|---|---|
|
CSCwn86588 |
Certain sensors (CPU, GPU, HIB, DIMM, OCP) show a critical state when the host is in an off state. This also causes the temperature to display as critical with a red indication in the UI dashboard. This issue is resolved. |
1.1.0.250016 |
1.1.0.250016 |
|
CSCwn86753 |
A standby power issue affects the OCP adapter on the eth1 interface, causing BMC network connectivity to go down when the host is powered off. This impacts systems using the shared-NIC concept for the eth1 interface. This issue is resolved. |
1.0.0.250003 |
1.1.0.250016 |
|
CSCwn89077 |
Due to a configuration issue, some servers may experience limited network accessibility restricted to the same subnet. To resolve this, a MAC address configuration update is required. This affects environments with static configuration settings. This issue is resolved. |
1.0.0.250003 |
1.1.0.250016 |
|
CSCwo02028 |
Disabling the domain name, DNS servers, or NTP servers from the network settings causes the BMC IP to reset to 0.0.0.0, making the system CIMC inaccessible over the network. Recovery requires manually setting a static IP from the BIOS F2 page. This issue is resolved. |
1.0.0.240001 |
1.1.0.250016 |
|
CSCwo71754 |
User-configured BIOS settings are not retained after a BIOS or AMD Generic Encapsulated Software Architecture (AGESA) firmware update. This issue is resolved. |
1.0.0.240001 |
1.1.0.250016 |
Open Caveats
Open Caveats in Release 1.1.0.250025
The following defects are open in release 1.1.0.250025:
|
Defect ID |
Symptom |
Workaround |
First Affected Release |
|---|---|---|---|
|
CSCwq31612 |
BIOS and BMC update fail when the server Provision state shows as unknown on the firmware page. |
Reboot the BMC and verify that the Provision State is set to Provision.. Once verified, restart the firmware update. |
1.1.0.250025 |
|
CSCwp87458 |
Critical SEL assert and deassert logs are observed back-to-back under conditions such as AC cycle, host reboot, or BMC reboot. |
This is a transient event that can be safely disregarded, as the assert and deassert happen within a very short time frame. |
1.1.0.250025 |
|
CSCwo78458 |
The Cisco BMC GUI does not show any BIOS tokens under . You cannot view or modify BIOS settings through the GUI, so BIOS configuration options are unavailable. This prevents you from changing BIOS settings using the Cisco BMC interface. |
Restart the system, access the BIOS settings during startup, and use the BIOS setup to configure. |
1.1.0.250025 |
Open Caveats in Release 1.1.0.250022
The following defects are open in release 1.1.0.250022:
|
Defect ID |
Symptom |
Workaround |
First Affected Release |
|---|---|---|---|
|
CSCwp07855 |
A critical SEL alert, SensorError Assert, EventData=21FF46, occurs after host reboot or AC cycle or BMC reboot. The issue causes the system health status to change to critical. |
There is no known workaround. This issue has no functionality impact. |
1.1.0.250022 |
|
CSCwp07720 |
A warning/critical SEL log, FAN Front/Rear warning low threshold assert, is observed during AC cycle, BMC reboot, or host power cycle FAN Speed Error asserts are reported with details such as PWM(49), Current Tach(6874), and EventData=11FF0B, affecting the fans. |
There is no known workaround. This issue has no functionality impact. |
1.1.0.250022 |
|
CSCwp23143 |
A critical SEL alert, AC_lost assert, AC lost, EventData=04FFFF, is observed during an AC cycle or when the system loses power abruptly. The log health status remains critical even after the condition is resolved, and the alert is deasserted. |
There is no known workaround. |
1.1.0.250022 |
Open Caveats in Release 1.1.0.250016
The following defects are open in Release 1.1.0.250016:
|
Defect ID |
Symptom |
Workaround |
First Affected Release |
|---|---|---|---|
|
CSCwn84681 |
The M.2 NVMe RAID Controller (UCS-M2-NVRAID) inventory is not accessible through BMC out-of-band communication. |
There is no known workaround. |
1.1.0.250016 |
|
CSCwn85582 |
The FHHL adapter goes into disable state. This issue occurs when adapter is set as InfiniBand. |
There is no known workaround. |
1.1.0.250016 |
|
CSCwn85620 |
The OCP card serial number is listed as NA in BMC out-of-band communication in the Rest API interface. |
There is no known workaround. This issue does not impact any functionality. |
1.1.0.250016 |
|
CSCwo14370 |
Rest API response does not include accurate and complete Product Identifier (PID) details for the GPU sled. |
Contact TAC for more assistance. |
1.1.0.250016 |
|
CSCwo53435 |
Sensor readings unexpectedly report 0°C in the Host ON - Post Completion state. This issue impacts accurate monitoring of system temperatures. |
Reset BMC to factory default settings. |
1.1.0.250016 |
|
CSCwo59380 |
In some servers, a PSU output LED occasionally shows amber. |
Remove and reseat the PSU. |
1.1.0.250016 |
|
CSCwo73008 |
Power on failure observed during a chassis power cycle through IPMI. The system fails to boot with the following error: Power supply power good failed to assert within 8000 milliseconds. This occurs when powering on immediately after a BMC reboot or AC power cycle. |
Wait for at least 10 minutes after the BMC is fully operational before proceeding. |
1.1.0.250016 |
|
CSCwo81255 |
The mapped vMedia is ejected upon Web GUI session timeout, causing OS installation on x86 CPUs to fail. |
Use remote vMedia mapping. From |
1.1.0.250016 |
|
CSCwo71111 |
Warning SEL logs indicate that multiple services, including phosphor-ipmi-host.service and bmcweb.service, exited unsuccessfully during network configuration changes. This occurs when switching the static network configuration from eth0 to eth1. |
There is no known workaround. This issue does not impact any functionality. |
1.1.0.250016 |
|
CSCwo78861 |
User creation takes longer than usual. This occurs when LDAP is enabled with an incorrect configuration, leading to slower user creation and session handling. |
Update LDAP configuration or wait for action to complete . |
1.1.0.250016 |
|
CSCwo68052 |
Network Adapter and Processor details are not fully displayed in the Cisco BMC GUI when cross-launched from Cisco Intersight using Read-Only user accounts. You may experience incomplete data visibility in this scenario. |
There is no known workaround for Intersight read-only users. All required data is displayed for Operator users created in the Cisco BMC GUI, as well as for Account Admin and Server Admin users from Cisco Intersight. |
1.1.0.250016 |
|
CSCwo68698 |
When initiating a logout action from the KVM cross-launched via Cisco Intersight, an exit confirmation icon labeled Leave repeatedly appears in the browser window. This behavior prevents you from closing the browser and disrupts the experience. |
Click the Close button in the top-right corner of the browser. |
1.1.0.250016 |
|
CSCwo68356 |
NVIDIA ConnectX CX7 (4x25) adapter details are not displayed correctly on the Inventory. Additionally, the overall server status and temperatures are incorrectly shown as Critical when the system is connected with the ConnectX CX7 (4x25) adapter. |
There is no functionality impact. |
1.1.0.250016 |
|
CSCwo85473 |
When the Host OS is turned off, PCIe is disconnected, which is expected to cause abnormal metric values. However, in GPU properties, the LanesInUse value is incorrect, and PCIeType is displayed as unknown under certain conditions when the Host OS is turned off for other reasons. |
When the Host OS is turned off, PCIe becomes disconnected, which can result in abnormal metric values. Intersight filters out these invalid values and displays accurate data in the Metrics Explorer. |
1.1.0.250016 |
|
CSCwo82679 |
A critical SEL error, Temperature sensor reading error assert, SensorName: TEMP_HIB_PEX1 is observed during a power cycle. Similar events are also noted for sensors PEX1 to PEX4, along with equivalent deassert events. |
This is a transient event that can be safely disregarded, as the assert and deassert happen within a very short time frame. |
1.1.0.250016 |
|
CSCwo76759 |
Cisco BMC is unable to dynamically re-learn the network gateway MAC address when the default gateway MAC is not set or has changed. As a result, while it can communicate within the same subnet, it cannot respond to traffic from a different subnet unless the default gateway MAC is manually configured. This may cause the Cisco BMC to become unreachable outside the subnet when switching between static environments with different subnets. |
Using any host in same subnet set the default getway MAC address: $ipmitool -H <BMC IP> -U root -P <password> -I lanplus lan set 1 defgw macaddr <MAC address of the getway> |
1.0.0.240001 |
|
CSCwo81247 |
The GUI session times out every 30 minutes, even during active use, causing interruptions to ongoing tasks. This behavior affects long-running activities, such as OS installations, as the session timeout leads to the disconnection of mounted vMedia. |
Refresh the session by logging out and logging back into the web UI. |
1.0.0.240001 |
Known Behaviors and Limitations
Known Behaviors and Limitations in Release 1.1.0.250025
The following caveats are known limitations in release 1.1.0.250025:
-
Defect ID—CSCwp31918
After upgrading the PCIe switch firmware on AMD Genoa-based servers, only 7 GPUs are detected instead of the expected 8. This issue occurs post-PCIe switch firmware update.
Workaround—The issue can be resolved by adding iommu=pt to the grub.conf file.
Known Behaviors and Limitations in Release 1.1.0.250022
The following caveats are known limitations in release 1.1.0.250022:
-
Defect ID—CSCwp91659
For Cisco UCS C885A M8 Rack Server, LPx/FHHLx adapters (CX7/BF3) appear in a critical state on the sensor page and GUI dashboard if Transceiver cables are not connected. This causes the temperature status in the GUI to display as critical.
Known Behaviors and Limitations in Release 1.1.0.250016
The following caveats are known limitations in release 1.1.0.250016:
-
Defect ID—CSCwo70892
Hot-plugged NVMe drives do not appear in the BMC inventory on the Storage page after reinsertion, though the host-based inventory works correctly.
Workaround—Reboot the host to retrieve the complete hardware inventory.
-
Defect ID—CSCwo70872
The TEMP_FH13_Chip and TEMP_FH13_TRX sensors may show critical alerts for empty riser slots after upgrading to release 1.1.0.250016 and powering on the host. There is no impact on functionality.
Workaround—While this issue has been resolved, if you encounter it, performing a one-time BMC reboot after the host is booted clears the alert.
Feedback