SPEC CPU20017 Platform Settings for Epsylon systems based on AMD Solutions
- drop_caches:
-
sysctl -w vm.drop_caches=3 - is used to clear filesystem caches at run-time.
- SMT Control:
-
This option can be used to managing Simultaneous Multi-Threading.
Available settings are [Disabled], [Enabled] and [Auto].
- [Disabled]: Disables simultaneous multithreading and only one thread or CPU instruction stream is run on a physical CPU core.
- [Enabled]: Enables simultaneous multithreading.
- [Auto]: Automatic activation of simultaneous multithreading when it is needed (Default).
- Determinism Control:
-
This option allows user can set customized determinism slider mode to control performance.
Available settings are: [Auto] and [Manual]
- [Auto] - Use the fused determinism slider mode. (Default)
- [Manulal] - Let user specifies customized determinism slider mode.
- Determinism Enable:
-
This option allows for AGESA determinism to control performance.
Available settings are: [Performane], [Power] and [Auto]
- [Performance] - Provides predictable performance across all processors of the same type.
- [Power] - Maximizes performance within the power limits defined by TDP and PPT.
- [Auto] -Use AGESA default value for deterministic performance control. (Default)
- TDP Control:
-
This option is for "Configurable TDP (cTDP)", it allows user set customized value for TDP.
Available settings are: [Auto] and [Manual]
- [Auto] - Use the fused TDP value. (Default)
- [Manual] - Let user specifies customized TDP value.
- TDP:
-
TDP is the recommended target for power used when designing the cooling capacity for a server.
EPYC processors are able to control this target power consumption within certain limits. This capability is referred to as "configurable TDP" or "cTDP."
TDP can be used to reduce power consumption for greater efficiency, or in some cases, increase power consumption above the default value to provide additional performance.
TDP is controlled using a BIOS option.
The default EPYC TDP value corresponds with the microprocessor's nominal TDP.
The default TDP value is set at a good balance between performance and energy efficiency.
Decrese the EPYC CPU TDP, which will minimize the power consumption for the processor under load, but at the expense of peak performance.
Increasing the EPYC CPU TDP will maximize peak performance by allowing the CPU to maintain higher dynamic clock speeds, but will make the microprocessor less energy efficient.
Note that at maximum TDP, the CPU thermal solution must be capable of dissipating at least highest energy peak or the EPYC processor might engage in thermal throttling under load.
The available TDP ranges for each EPYC model are in the table below:
Model | Minimum TDP | Maximum TDP |
EPYC 9654 | 320 | 400 |
EPYC 9654P | 320 | 400 |
EPYC 9634 | 240 | 300 |
EPYC 9554 | 320 | 400 |
EPYC 9554P | 320 | 400 |
EPYC 9534 | 240 | 300 |
EPYC 9474F | 320 | 400 |
EPYC 9454 | 240 | 300 |
EPYC 9454P | 240 | 300 |
EPYC 9374F | 320 | 400 |
EPYC 9354 | 240 | 300 |
EPYC 9354P | 240 | 300 |
EPYC 9334 | 200 | 240 |
EPYC 9274F | 320 | 400 |
EPYC 9254 | 200 | 240 |
EPYC 9224 | 200 | 240 |
EPYC 9174F | 320 | 400 |
EPYC 9124 | 200 | 240 |
EPYC 9754 | 320 | 400 |
EPYC 9754S | 320 | 400 |
EPYC 9734 | 320 | 400 |
EPYC 9684X | 320 | 400 |
EPYC 9384X | 320 | 400 |
EPYC 9184X | 320 | 400 |
* TDP must remain below the thermal solution design parameters or thermal throttling could be frequently encountered.
- SVM Mode:
-
This option stand for CPU virtualization function. What it allow user to install the virtual machnine.
Available settings are: [Enable] and [Disable]
- [Enable] - User will be able to install a virtual machine on system. (Default)
- [Disable] - User will not be able to install a virtual machine on system.
- SR-IOV Support:
-
This option stand for sharing the PCIe. Single Root Input/Output Virtualization (SR-IOV) this option allows isolate the PCI Express resources for manageability and performance reasons for Virtualization.
A single physical PCI Express can be shared on a virtual environment using the SR-IOV specification.
If system has SR-IOV capable PCIe Devices, this option Enables or Disables Single Root IO Virtualization Support.
Available settings are: [Enable] and [Disable]
- [Enable] - User will be able to share a PCIe. (Default)
- [Disable] - Turn off capability for sharing PCIe.
- DRAM Scrub time:
-
This option is a mechanism for the memory controller to periodically read all memory locations and write back corrected data.
Available settings are:[Disable], [1 hour], [4 hours], [8 hours], [16 hours], [24 hours], [48 hours] and [Auto]
- [Auto] - Auto use AGESA default values. (Default)
- L1 Stream HW Prefetcher:
-
L1 Stream HW Prefetcher uses history of memory access patterns to fetch additional sequential lines in ascending or descending order into the L1 Cache.
Avalible settins are: [Enable], [Disable], [Auto]
- [Enable]: Enable the L1 Stream HW
- [Disable]: Disable the L1 Stream HW
- [Auto]: Use AGESA default value. (Default)
- L2 Stream HW Prefetcher:
-
L2 Stream HW Prefetcher uses history of memory access patterns to fetch additional sequential lines in ascending or descending order into the L2 Cache.
Avalible settins are: [Enable], [Disable], [Auto]
- [Enable]: Enable the L2 Stream HW
- [Disable]: Disable the LS Stream HW
- [Auto]: Use AGESA default value. (Default)
- NUMA nodes per socket:
-
Specifies the number of desired NUMA nodes per populated socket in the system:
Avalible settins are: [NPS1], [NPS2], [NPS4], [Auto]
- [NPS1]: Each physical processor is a NUMA node, and memory accesses are interleaved across all memory channels directly connected to the physical processor.
- [NPS2]: Each physical processor is two NUMA nodes, and memory accesses are interleaved across 4 memory channels.
- [NPS4]: Each physical processor is four NUMA nodes, and memory accesses are interleaved across 2 memory channels.
- [Auto]: Use AGESA default value. (Default)
- APBDIS:
-
Application Power Management (APM) allows the processor to provide maximum performance while remaining within
the specified power delivery and removal envelope. APM dynamically monitors processor activity and generates an
approximation of power consumption. If power consumption exceeds a defined power limit, a P-state limit is applied by
APM hardware to reduce power consumption. APM ensures that average power consumption over a thermally significant
time period remains at or below the defined power limit. Set APBDIS=1 will disable Data Fabric APM and the SOC P-state will be fixed.
Available settings are: [0], [1], [Auto]
- [0]: Disable APBDIS.
- [1]: Enable APBDIS.
- [Auto] - Use default value for APBDIS. The current value is 0. (Default)
- DfPstate:
-
To minimize variance or trade-off memory latency versus bandwidth, algorithm performance boost
(APBDIS) can be set and specific hard-fused Data Fabric (SoC) P-states forced for optimized
workloads sensitive to latency or throughput.
Available settings are: [P0], [P1], [P2], [P3] and [Auto]
- [P0] - The highest SOC P-state.
- [P1] - The middle-high SOC P-state.
- [P2] - The middle-low SOC P-state.
- [P3] - The lowest SOC P-state.
- [Auto] - Use AGESA default value. (Default)
- ACPI SRAT L3 Cache as NUMA Domain:
-
Each L3 Cache will be exposed as a NUMA node when enabling ACPI SRAT L3 Cache as a NUMA node.
On a dual processor system, with up to 8 L3 Caches per processor, this setting will expose 16 NUMA domains.
Available settings are: [Auto] and [Enable]
- [Auto] - Disable this function. (Default)
- [Enabled] - Enable this function.
- Package Power Limit Control:
-
This option allows user can set customized value for processor package power limit (PPT).
Available settings are: [Auto] and [Manual]
- [Auto] - Use the fused PPT. (Default)
- [Manual] - User can set customized PPT.
- DLWM Support:
-
Dynamic Link Width Management(DLWM) reduces xGMI lane width from x16 to x8 or x2 if xGMI links have limited traffic. DLWM feature is optimized to
trade power between CPU core intensive workloads and I/O bandwidth intensive workloads. When link activity is above a threshold, DLWM will
increase lane width from x8 to x16 at the cost of some delay, because the I/O die must disconnect the links, retrain them at the new speed
and release the system back to functionality.
Avalible settings are: [Enable], [Disable] and [Auto]
- [Enable] - Enable this function.
- [Disable] - Dusable this function.
- [Auto] - Use AGESA default value. Current is Enable. (Default)
- Engine Boost:
-
Platform feature with the power acceleration design to increase CPU over-all performance.
Available settings are: [Disabled], [Normal], and [Aggressive]
- [Disable] - Disable this function. (Default)
- [Normal] - Increases CPU performance.
- [Aggressive] - Aggressive mode improve CPU performance more aggressively than Normal mode, but come with more power consumption and higher CPU temperature.
- PPT Control:
-
Avalibe settings are: [Auto] (Default) and [Manual]
- [Auto] - Use the fused PPT.
- [Manual] - User can set customized PPT [W] in decimal.
- PPT:
-
Set customize processor PPT (Package Power Limit) value to be used on all populated processors in the system.
- [0] - Use the fused PPT value (Default).
- [other value] - Use the customized PPT value [W] in decimal.
***PPT will be used as the ASIC power limit***
- IOMMU:
-
The Input-Output Memory Management Unit(IOMMU) provides several benefits and is required when using x2APIC. IOMMU allows devices
(such as the EPYC integrated SATA controller) to present separate IRQs for each attached device instead of one IRQ for the subsystem.
The IOMMU also allows operating systems to provide additional protection for DMA capable I/O devices.
Avalibe settings are: [Enable], [Disable] and [Auto]
- [Enable] - Allow devices to present separate IRQs for each device instead of one IRQ for subsystem.
- [Disable] - Disable this function.
- [Auto] - Use AGESA default value. Current is Enable. (Default)
- Memory Interleaving:
-
Memory interleaving is a technique that CPUs use to increase the memory bandwidth available for an
application. Without interleaving, consecutive memory blocks, often cache lines, are read from the
same memory bank. Because of this, software that reads consecutive memory must wait for a
memory transfer to complete before starting the next memory access, reducing throughput and
increasing latency.
By enabling memory interleaving, consecutive memory blocks are in different
banks and can all contribute to the overall memory bandwidth, thus increasing throughput and lowering latency.
Avalibe settings are: [Auto] and [Disabled]
- [Auto] - Use AGESA default value. Current is Enable. (Default)
- [Disable] - Disable this function.
- Fan mode:
-
To Select system Fan policy on BMC WebUI.
Available settings are: [Generic mode] and [Full speed mode]
- [Generic mode]: makes fan speed self-adjust actively according to the reading of the system temperature sensors. (Default)
- [Full speed mode]: makes fan run at full speed to provide best cooling performance .