feat(inputs.nvidia_smi): Add additional memory related fields (#13133)

This commit is contained in:
Alex Zherebtsov 2023-04-26 22:00:14 +02:00 committed by GitHub
parent bdfbe67cb0
commit cd535165d0
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
5 changed files with 856 additions and 18 deletions

View File

@ -57,6 +57,16 @@ Files\\NVIDIA Corporation\\NVSMI\\nvidia-smi.exe`
- `memory_free` (integer, MiB)
- `memory_used` (integer, MiB)
- `memory_total` (integer, MiB)
- `memory_reserved` (integer, MiB)
- `retired_pages_multiple_single_bit` (integer)
- `retired_pages_double_bit` (integer)
- `retired_pages_blacklist` (string)
- `retired_pages_pending` (string)
- `remapped_rows_correctable` (int)
- `remapped_rows_uncorrectable` (int)
- `remapped_rows_pending` (string)
- `remapped_rows_pending` (string)
- `remapped_rows_failure` (string)
- `power_draw` (float, W)
- `temperature_gpu` (integer, degrees C)
- `utilization_gpu` (integer, percentage)

View File

@ -118,6 +118,15 @@ func (s *SMI) genTagsFields() []metric {
setIfUsed("int", fields, "memory_total", gpu.Memory.Total)
setIfUsed("int", fields, "memory_used", gpu.Memory.Used)
setIfUsed("int", fields, "memory_free", gpu.Memory.Free)
setIfUsed("int", fields, "memory_reserved", gpu.Memory.Reserved)
setIfUsed("int", fields, "retired_pages_multiple_single_bit", gpu.RetiredPages.MultipleSingleBit.Count)
setIfUsed("int", fields, "retired_pages_double_bit", gpu.RetiredPages.DoubleBit.Count)
setIfUsed("str", fields, "retired_pages_blacklist", gpu.RetiredPages.PendingBlacklist)
setIfUsed("str", fields, "retired_pages_pending", gpu.RetiredPages.PendingRetirement)
setIfUsed("int", fields, "remapped_rows_correctable", gpu.RemappedRows.Correctable)
setIfUsed("int", fields, "remapped_rows_uncorrectable", gpu.RemappedRows.Uncorrectable)
setIfUsed("str", fields, "remapped_rows_pending", gpu.RemappedRows.Pending)
setIfUsed("str", fields, "remapped_rows_failure", gpu.RemappedRows.Failure)
setIfUsed("int", fields, "temperature_gpu", gpu.Temp.GPUTemp)
setIfUsed("int", fields, "utilization_gpu", gpu.Utilization.GPU)
setIfUsed("int", fields, "utilization_memory", gpu.Utilization.Memory)
@ -168,14 +177,14 @@ func setIfUsed(t string, m map[string]interface{}, k, v string) {
}
}
case "int":
if val != "" {
if val != "" && val != "N/A" {
i, err := strconv.Atoi(val)
if err == nil {
m[k] = i
}
}
case "str":
if val != "" {
if val != "" && val != "N/A" {
m[k] = val
}
}
@ -190,26 +199,49 @@ type SMI struct {
// GPU defines the structure of the GPU portion of the smi output.
type GPU []struct {
FanSpeed string `xml:"fan_speed"` // int
Memory MemoryStats `xml:"fb_memory_usage"`
PState string `xml:"performance_state"`
Temp TempStats `xml:"temperature"`
ProdName string `xml:"product_name"`
UUID string `xml:"uuid"`
ComputeMode string `xml:"compute_mode"`
Utilization UtilizationStats `xml:"utilization"`
Power PowerReadings `xml:"power_readings"`
PCI PCI `xml:"pci"`
Encoder EncoderStats `xml:"encoder_stats"`
FBC FBCStats `xml:"fbc_stats"`
Clocks ClockStats `xml:"clocks"`
FanSpeed string `xml:"fan_speed"` // int
Memory MemoryStats `xml:"fb_memory_usage"`
RetiredPages MemoryRetiredPages `xml:"retired_pages"`
RemappedRows MemoryRemappedRows `xml:"remapped_rows"`
PState string `xml:"performance_state"`
Temp TempStats `xml:"temperature"`
ProdName string `xml:"product_name"`
UUID string `xml:"uuid"`
ComputeMode string `xml:"compute_mode"`
Utilization UtilizationStats `xml:"utilization"`
Power PowerReadings `xml:"power_readings"`
PCI PCI `xml:"pci"`
Encoder EncoderStats `xml:"encoder_stats"`
FBC FBCStats `xml:"fbc_stats"`
Clocks ClockStats `xml:"clocks"`
}
// MemoryStats defines the structure of the memory portions in the smi output.
type MemoryStats struct {
Total string `xml:"total"` // int
Used string `xml:"used"` // int
Free string `xml:"free"` // int
Total string `xml:"total"` // int
Used string `xml:"used"` // int
Free string `xml:"free"` // int
Reserved string `xml:"reserved"` // int
}
// MemoryRetiredPages defines the structure of the retired pages portions in the smi output.
type MemoryRetiredPages struct {
MultipleSingleBit struct {
Count string `xml:"retired_count"` // int
} `xml:"multiple_single_bit_retirement"`
DoubleBit struct {
Count string `xml:"retired_count"` // int
} `xml:"double_bit_retirement"`
PendingBlacklist string `xml:"pending_blacklist"` // Yes/No
PendingRetirement string `xml:"pending_retirement"` // Yes/No
}
// MemoryRemappedRows defines the structure of the remapped rows portions in the smi output.
type MemoryRemappedRows struct {
Correctable string `xml:"remapped_row_corr"` // int
Uncorrectable string `xml:"remapped_row_unc"` // int
Pending string `xml:"remapped_row_pending"` // Yes/No
Failure string `xml:"remapped_row_failure"` // Yes/No
}
// TempStats defines the structure of the temperature portion of the smi output.

View File

@ -134,6 +134,99 @@ func TestGatherValidXML(t *testing.T) {
time.Unix(0, 0)),
},
},
{
name: "Tesla T4",
filename: "tesla-t4.xml",
expected: []telegraf.Metric{
testutil.MustMetric(
"nvidia_smi",
map[string]string{
"compute_mode": "Default",
"index": "0",
"name": "Tesla T4",
"pstate": "P0",
"uuid": "GPU-d37e67a5-91dd-3774-a5cb-99096249601a",
},
map[string]interface{}{
"clocks_current_graphics": 585,
"clocks_current_memory": 5000,
"clocks_current_sm": 585,
"clocks_current_video": 810,
"cuda_version": "11.7",
"driver_version": "515.105.01",
"encoder_stats_average_fps": 0,
"encoder_stats_average_latency": 0,
"encoder_stats_session_count": 0,
"fbc_stats_average_fps": 0,
"fbc_stats_average_latency": 0,
"fbc_stats_session_count": 0,
"power_draw": 26.78,
"memory_free": 13939,
"memory_total": 15360,
"memory_used": 1032,
"memory_reserved": 388,
"retired_pages_multiple_single_bit": 0,
"retired_pages_double_bit": 0,
"retired_pages_blacklist": "No",
"retired_pages_pending": "No",
"pcie_link_gen_current": 3,
"pcie_link_width_current": 8,
"temperature_gpu": 40,
"utilization_gpu": 0,
"utilization_memory": 0,
"utilization_encoder": 0,
"utilization_decoder": 0,
},
time.Unix(0, 0)),
},
},
{
name: "A10G",
filename: "a10g.xml",
expected: []telegraf.Metric{
testutil.MustMetric(
"nvidia_smi",
map[string]string{
"compute_mode": "Default",
"index": "0",
"name": "NVIDIA A10G",
"pstate": "P8",
"uuid": "GPU-9a9a6c50-2a47-2f51-a902-b82c3b127e94",
},
map[string]interface{}{
"clocks_current_graphics": 210,
"clocks_current_memory": 405,
"clocks_current_sm": 210,
"clocks_current_video": 555,
"cuda_version": "11.7",
"driver_version": "515.105.01",
"encoder_stats_average_fps": 0,
"encoder_stats_average_latency": 0,
"encoder_stats_session_count": 0,
"fbc_stats_average_fps": 0,
"fbc_stats_average_latency": 0,
"fbc_stats_session_count": 0,
"fan_speed": 0,
"power_draw": 25.58,
"memory_free": 22569,
"memory_total": 23028,
"memory_used": 22,
"memory_reserved": 435,
"remapped_rows_correctable": 0,
"remapped_rows_uncorrectable": 0,
"remapped_rows_pending": "No",
"remapped_rows_failure": "No",
"pcie_link_gen_current": 1,
"pcie_link_width_current": 8,
"temperature_gpu": 17,
"utilization_gpu": 0,
"utilization_memory": 0,
"utilization_encoder": 0,
"utilization_decoder": 0,
},
time.Unix(0, 0)),
},
},
}
for _, tt := range tests {
t.Run(tt.name, func(t *testing.T) {

View File

@ -0,0 +1,355 @@
<?xml version="1.0" ?>
<!DOCTYPE nvidia_smi_log SYSTEM "nvsmi_device_v11.dtd">
<nvidia_smi_log>
<timestamp>Mon Apr 24 16:11:51 2023</timestamp>
<driver_version>515.105.01</driver_version>
<cuda_version>11.7</cuda_version>
<attached_gpus>1</attached_gpus>
<gpu id="00000000:00:1E.0">
<product_name>NVIDIA A10G</product_name>
<product_brand>NVIDIA</product_brand>
<product_architecture>Ampere</product_architecture>
<display_mode>Disabled</display_mode>
<display_active>Disabled</display_active>
<persistence_mode>Disabled</persistence_mode>
<mig_mode>
<current_mig>N/A</current_mig>
<pending_mig>N/A</pending_mig>
</mig_mode>
<mig_devices>
None
</mig_devices>
<accounting_mode>Disabled</accounting_mode>
<accounting_mode_buffer_size>4000</accounting_mode_buffer_size>
<driver_model>
<current_dm>N/A</current_dm>
<pending_dm>N/A</pending_dm>
</driver_model>
<serial>0000000000000</serial>
<uuid>GPU-9a9a6c50-2a47-2f51-a902-b82c3b127e94</uuid>
<minor_number>0</minor_number>
<vbios_version>94.02.75.00.01</vbios_version>
<multigpu_board>No</multigpu_board>
<board_id>0x1e</board_id>
<gpu_part_number>000-00000-0000-000</gpu_part_number>
<gpu_module_id>0</gpu_module_id>
<inforom_version>
<img_version>G133.0210.00.04</img_version>
<oem_object>2.0</oem_object>
<ecc_object>6.16</ecc_object>
<pwr_object>N/A</pwr_object>
</inforom_version>
<gpu_operation_mode>
<current_gom>N/A</current_gom>
<pending_gom>N/A</pending_gom>
</gpu_operation_mode>
<gsp_firmware_version>515.105.01</gsp_firmware_version>
<gpu_virtualization_mode>
<virtualization_mode>Pass-Through</virtualization_mode>
<host_vgpu_mode>N/A</host_vgpu_mode>
</gpu_virtualization_mode>
<ibmnpu>
<relaxed_ordering_mode>N/A</relaxed_ordering_mode>
</ibmnpu>
<pci>
<pci_bus>00</pci_bus>
<pci_device>1E</pci_device>
<pci_domain>0000</pci_domain>
<pci_device_id>000000DE</pci_device_id>
<pci_bus_id>00000000:00:1E.0</pci_bus_id>
<pci_sub_system_id>000000DE</pci_sub_system_id>
<pci_gpu_link_info>
<pcie_gen>
<max_link_gen>4</max_link_gen>
<current_link_gen>1</current_link_gen>
</pcie_gen>
<link_widths>
<max_link_width>16x</max_link_width>
<current_link_width>8x</current_link_width>
</link_widths>
</pci_gpu_link_info>
<pci_bridge_chip>
<bridge_chip_type>N/A</bridge_chip_type>
<bridge_chip_fw>N/A</bridge_chip_fw>
</pci_bridge_chip>
<replay_counter>0</replay_counter>
<replay_rollover_counter>0</replay_rollover_counter>
<tx_util>0 KB/s</tx_util>
<rx_util>0 KB/s</rx_util>
</pci>
<fan_speed>0 %</fan_speed>
<performance_state>P8</performance_state>
<clocks_throttle_reasons>
<clocks_throttle_reason_gpu_idle>Active</clocks_throttle_reason_gpu_idle>
<clocks_throttle_reason_applications_clocks_setting>Not Active</clocks_throttle_reason_applications_clocks_setting>
<clocks_throttle_reason_sw_power_cap>Not Active</clocks_throttle_reason_sw_power_cap>
<clocks_throttle_reason_hw_slowdown>Not Active</clocks_throttle_reason_hw_slowdown>
<clocks_throttle_reason_hw_thermal_slowdown>Not Active</clocks_throttle_reason_hw_thermal_slowdown>
<clocks_throttle_reason_hw_power_brake_slowdown>Not Active</clocks_throttle_reason_hw_power_brake_slowdown>
<clocks_throttle_reason_sync_boost>Not Active</clocks_throttle_reason_sync_boost>
<clocks_throttle_reason_sw_thermal_slowdown>Not Active</clocks_throttle_reason_sw_thermal_slowdown>
<clocks_throttle_reason_display_clocks_setting>Not Active</clocks_throttle_reason_display_clocks_setting>
</clocks_throttle_reasons>
<fb_memory_usage>
<total>23028 MiB</total>
<reserved>435 MiB</reserved>
<used>22 MiB</used>
<free>22569 MiB</free>
</fb_memory_usage>
<bar1_memory_usage>
<total>32768 MiB</total>
<used>1 MiB</used>
<free>32767 MiB</free>
</bar1_memory_usage>
<compute_mode>Default</compute_mode>
<utilization>
<gpu_util>0 %</gpu_util>
<memory_util>0 %</memory_util>
<encoder_util>0 %</encoder_util>
<decoder_util>0 %</decoder_util>
</utilization>
<encoder_stats>
<session_count>0</session_count>
<average_fps>0</average_fps>
<average_latency>0</average_latency>
</encoder_stats>
<fbc_stats>
<session_count>0</session_count>
<average_fps>0</average_fps>
<average_latency>0</average_latency>
</fbc_stats>
<ecc_mode>
<current_ecc>Enabled</current_ecc>
<pending_ecc>Enabled</pending_ecc>
</ecc_mode>
<ecc_errors>
<volatile>
<sram_correctable>0</sram_correctable>
<sram_uncorrectable>0</sram_uncorrectable>
<dram_correctable>0</dram_correctable>
<dram_uncorrectable>0</dram_uncorrectable>
</volatile>
<aggregate>
<sram_correctable>0</sram_correctable>
<sram_uncorrectable>0</sram_uncorrectable>
<dram_correctable>0</dram_correctable>
<dram_uncorrectable>0</dram_uncorrectable>
</aggregate>
</ecc_errors>
<retired_pages>
<multiple_single_bit_retirement>
<retired_count>N/A</retired_count>
<retired_pagelist>N/A</retired_pagelist>
</multiple_single_bit_retirement>
<double_bit_retirement>
<retired_count>N/A</retired_count>
<retired_pagelist>N/A</retired_pagelist>
</double_bit_retirement>
<pending_blacklist>N/A</pending_blacklist>
<pending_retirement>N/A</pending_retirement>
</retired_pages>
<remapped_rows>
<remapped_row_corr>0</remapped_row_corr>
<remapped_row_unc>0</remapped_row_unc>
<remapped_row_pending>No</remapped_row_pending>
<remapped_row_failure>No</remapped_row_failure>
<row_remapper_histogram>
<row_remapper_histogram_max>192 bank(s)</row_remapper_histogram_max>
<row_remapper_histogram_high>0 bank(s)</row_remapper_histogram_high>
<row_remapper_histogram_partial>0 bank(s)</row_remapper_histogram_partial>
<row_remapper_histogram_low>0 bank(s)</row_remapper_histogram_low>
<row_remapper_histogram_none>0 bank(s)</row_remapper_histogram_none>
</row_remapper_histogram>
</remapped_rows>
<temperature>
<gpu_temp>17 C</gpu_temp>
<gpu_temp_max_threshold>98 C</gpu_temp_max_threshold>
<gpu_temp_slow_threshold>95 C</gpu_temp_slow_threshold>
<gpu_temp_max_gpu_threshold>88 C</gpu_temp_max_gpu_threshold>
<gpu_target_temperature>N/A</gpu_target_temperature>
<memory_temp>N/A</memory_temp>
<gpu_temp_max_mem_threshold>N/A</gpu_temp_max_mem_threshold>
</temperature>
<supported_gpu_target_temp>
<gpu_target_temp_min>N/A</gpu_target_temp_min>
<gpu_target_temp_max>N/A</gpu_target_temp_max>
</supported_gpu_target_temp>
<power_readings>
<power_state>P8</power_state>
<power_management>Supported</power_management>
<power_draw>25.58 W</power_draw>
<power_limit>300.00 W</power_limit>
<default_power_limit>300.00 W</default_power_limit>
<enforced_power_limit>300.00 W</enforced_power_limit>
<min_power_limit>100.00 W</min_power_limit>
<max_power_limit>300.00 W</max_power_limit>
</power_readings>
<clocks>
<graphics_clock>210 MHz</graphics_clock>
<sm_clock>210 MHz</sm_clock>
<mem_clock>405 MHz</mem_clock>
<video_clock>555 MHz</video_clock>
</clocks>
<applications_clocks>
<graphics_clock>1710 MHz</graphics_clock>
<mem_clock>6251 MHz</mem_clock>
</applications_clocks>
<default_applications_clocks>
<graphics_clock>1710 MHz</graphics_clock>
<mem_clock>6251 MHz</mem_clock>
</default_applications_clocks>
<max_clocks>
<graphics_clock>1710 MHz</graphics_clock>
<sm_clock>1710 MHz</sm_clock>
<mem_clock>6251 MHz</mem_clock>
<video_clock>1500 MHz</video_clock>
</max_clocks>
<max_customer_boost_clocks>
<graphics_clock>1710 MHz</graphics_clock>
</max_customer_boost_clocks>
<clock_policy>
<auto_boost>N/A</auto_boost>
<auto_boost_default>N/A</auto_boost_default>
</clock_policy>
<voltage>
<graphics_volt>693.750 mV</graphics_volt>
</voltage>
<supported_clocks>
<supported_mem_clock>
<value>6251 MHz</value>
<supported_graphics_clock>1710 MHz</supported_graphics_clock>
<supported_graphics_clock>1695 MHz</supported_graphics_clock>
<supported_graphics_clock>1680 MHz</supported_graphics_clock>
<supported_graphics_clock>1665 MHz</supported_graphics_clock>
<supported_graphics_clock>1650 MHz</supported_graphics_clock>
<supported_graphics_clock>1635 MHz</supported_graphics_clock>
<supported_graphics_clock>1620 MHz</supported_graphics_clock>
<supported_graphics_clock>1605 MHz</supported_graphics_clock>
<supported_graphics_clock>1590 MHz</supported_graphics_clock>
<supported_graphics_clock>1575 MHz</supported_graphics_clock>
<supported_graphics_clock>1560 MHz</supported_graphics_clock>
<supported_graphics_clock>1545 MHz</supported_graphics_clock>
<supported_graphics_clock>1530 MHz</supported_graphics_clock>
<supported_graphics_clock>1515 MHz</supported_graphics_clock>
<supported_graphics_clock>1500 MHz</supported_graphics_clock>
<supported_graphics_clock>1485 MHz</supported_graphics_clock>
<supported_graphics_clock>1470 MHz</supported_graphics_clock>
<supported_graphics_clock>1455 MHz</supported_graphics_clock>
<supported_graphics_clock>1440 MHz</supported_graphics_clock>
<supported_graphics_clock>1425 MHz</supported_graphics_clock>
<supported_graphics_clock>1410 MHz</supported_graphics_clock>
<supported_graphics_clock>1395 MHz</supported_graphics_clock>
<supported_graphics_clock>1380 MHz</supported_graphics_clock>
<supported_graphics_clock>1365 MHz</supported_graphics_clock>
<supported_graphics_clock>1350 MHz</supported_graphics_clock>
<supported_graphics_clock>1335 MHz</supported_graphics_clock>
<supported_graphics_clock>1320 MHz</supported_graphics_clock>
<supported_graphics_clock>1305 MHz</supported_graphics_clock>
<supported_graphics_clock>1290 MHz</supported_graphics_clock>
<supported_graphics_clock>1275 MHz</supported_graphics_clock>
<supported_graphics_clock>1260 MHz</supported_graphics_clock>
<supported_graphics_clock>1245 MHz</supported_graphics_clock>
<supported_graphics_clock>1230 MHz</supported_graphics_clock>
<supported_graphics_clock>1215 MHz</supported_graphics_clock>
<supported_graphics_clock>1200 MHz</supported_graphics_clock>
<supported_graphics_clock>1185 MHz</supported_graphics_clock>
<supported_graphics_clock>1170 MHz</supported_graphics_clock>
<supported_graphics_clock>1155 MHz</supported_graphics_clock>
<supported_graphics_clock>1140 MHz</supported_graphics_clock>
<supported_graphics_clock>1125 MHz</supported_graphics_clock>
<supported_graphics_clock>1110 MHz</supported_graphics_clock>
<supported_graphics_clock>1095 MHz</supported_graphics_clock>
<supported_graphics_clock>1080 MHz</supported_graphics_clock>
<supported_graphics_clock>1065 MHz</supported_graphics_clock>
<supported_graphics_clock>1050 MHz</supported_graphics_clock>
<supported_graphics_clock>1035 MHz</supported_graphics_clock>
<supported_graphics_clock>1020 MHz</supported_graphics_clock>
<supported_graphics_clock>1005 MHz</supported_graphics_clock>
<supported_graphics_clock>990 MHz</supported_graphics_clock>
<supported_graphics_clock>975 MHz</supported_graphics_clock>
<supported_graphics_clock>960 MHz</supported_graphics_clock>
<supported_graphics_clock>945 MHz</supported_graphics_clock>
<supported_graphics_clock>930 MHz</supported_graphics_clock>
<supported_graphics_clock>915 MHz</supported_graphics_clock>
<supported_graphics_clock>900 MHz</supported_graphics_clock>
<supported_graphics_clock>885 MHz</supported_graphics_clock>
<supported_graphics_clock>870 MHz</supported_graphics_clock>
<supported_graphics_clock>855 MHz</supported_graphics_clock>
<supported_graphics_clock>840 MHz</supported_graphics_clock>
<supported_graphics_clock>825 MHz</supported_graphics_clock>
<supported_graphics_clock>810 MHz</supported_graphics_clock>
<supported_graphics_clock>795 MHz</supported_graphics_clock>
<supported_graphics_clock>780 MHz</supported_graphics_clock>
<supported_graphics_clock>765 MHz</supported_graphics_clock>
<supported_graphics_clock>750 MHz</supported_graphics_clock>
<supported_graphics_clock>735 MHz</supported_graphics_clock>
<supported_graphics_clock>720 MHz</supported_graphics_clock>
<supported_graphics_clock>705 MHz</supported_graphics_clock>
<supported_graphics_clock>690 MHz</supported_graphics_clock>
<supported_graphics_clock>675 MHz</supported_graphics_clock>
<supported_graphics_clock>660 MHz</supported_graphics_clock>
<supported_graphics_clock>645 MHz</supported_graphics_clock>
<supported_graphics_clock>630 MHz</supported_graphics_clock>
<supported_graphics_clock>615 MHz</supported_graphics_clock>
<supported_graphics_clock>600 MHz</supported_graphics_clock>
<supported_graphics_clock>585 MHz</supported_graphics_clock>
<supported_graphics_clock>570 MHz</supported_graphics_clock>
<supported_graphics_clock>555 MHz</supported_graphics_clock>
<supported_graphics_clock>540 MHz</supported_graphics_clock>
<supported_graphics_clock>525 MHz</supported_graphics_clock>
<supported_graphics_clock>510 MHz</supported_graphics_clock>
<supported_graphics_clock>495 MHz</supported_graphics_clock>
<supported_graphics_clock>480 MHz</supported_graphics_clock>
<supported_graphics_clock>465 MHz</supported_graphics_clock>
<supported_graphics_clock>450 MHz</supported_graphics_clock>
<supported_graphics_clock>435 MHz</supported_graphics_clock>
<supported_graphics_clock>420 MHz</supported_graphics_clock>
<supported_graphics_clock>405 MHz</supported_graphics_clock>
<supported_graphics_clock>390 MHz</supported_graphics_clock>
<supported_graphics_clock>375 MHz</supported_graphics_clock>
<supported_graphics_clock>360 MHz</supported_graphics_clock>
<supported_graphics_clock>345 MHz</supported_graphics_clock>
<supported_graphics_clock>330 MHz</supported_graphics_clock>
<supported_graphics_clock>315 MHz</supported_graphics_clock>
<supported_graphics_clock>300 MHz</supported_graphics_clock>
<supported_graphics_clock>285 MHz</supported_graphics_clock>
<supported_graphics_clock>270 MHz</supported_graphics_clock>
<supported_graphics_clock>255 MHz</supported_graphics_clock>
<supported_graphics_clock>240 MHz</supported_graphics_clock>
<supported_graphics_clock>225 MHz</supported_graphics_clock>
<supported_graphics_clock>210 MHz</supported_graphics_clock>
</supported_mem_clock>
<supported_mem_clock>
<value>405 MHz</value>
<supported_graphics_clock>420 MHz</supported_graphics_clock>
<supported_graphics_clock>405 MHz</supported_graphics_clock>
<supported_graphics_clock>390 MHz</supported_graphics_clock>
<supported_graphics_clock>375 MHz</supported_graphics_clock>
<supported_graphics_clock>360 MHz</supported_graphics_clock>
<supported_graphics_clock>345 MHz</supported_graphics_clock>
<supported_graphics_clock>330 MHz</supported_graphics_clock>
<supported_graphics_clock>315 MHz</supported_graphics_clock>
<supported_graphics_clock>300 MHz</supported_graphics_clock>
<supported_graphics_clock>285 MHz</supported_graphics_clock>
<supported_graphics_clock>270 MHz</supported_graphics_clock>
<supported_graphics_clock>255 MHz</supported_graphics_clock>
<supported_graphics_clock>240 MHz</supported_graphics_clock>
<supported_graphics_clock>225 MHz</supported_graphics_clock>
<supported_graphics_clock>210 MHz</supported_graphics_clock>
</supported_mem_clock>
</supported_clocks>
<processes>
<process_info>
<gpu_instance_id>N/A</gpu_instance_id>
<compute_instance_id>N/A</compute_instance_id>
<pid>725</pid>
<type>G</type>
<process_name>/usr/lib/xorg/Xorg</process_name>
<used_memory>22 MiB</used_memory>
</process_info>
</processes>
<accounted_processes>
</accounted_processes>
</gpu>
</nvidia_smi_log>

View File

@ -0,0 +1,348 @@
<?xml version="1.0" ?>
<!DOCTYPE nvidia_smi_log SYSTEM "nvsmi_device_v11.dtd">
<nvidia_smi_log>
<timestamp>Mon Apr 24 16:22:39 2023</timestamp>
<driver_version>515.105.01</driver_version>
<cuda_version>11.7</cuda_version>
<attached_gpus>1</attached_gpus>
<gpu id="00000000:00:1E.0">
<product_name>Tesla T4</product_name>
<product_brand>NVIDIA</product_brand>
<product_architecture>Turing</product_architecture>
<display_mode>Disabled</display_mode>
<display_active>Disabled</display_active>
<persistence_mode>Disabled</persistence_mode>
<mig_mode>
<current_mig>N/A</current_mig>
<pending_mig>N/A</pending_mig>
</mig_mode>
<mig_devices>
None
</mig_devices>
<accounting_mode>Disabled</accounting_mode>
<accounting_mode_buffer_size>4000</accounting_mode_buffer_size>
<driver_model>
<current_dm>N/A</current_dm>
<pending_dm>N/A</pending_dm>
</driver_model>
<serial>0000000000000</serial>
<uuid>GPU-d37e67a5-91dd-3774-a5cb-99096249601a</uuid>
<minor_number>0</minor_number>
<vbios_version>90.04.84.00.06</vbios_version>
<multigpu_board>No</multigpu_board>
<board_id>0x1e</board_id>
<gpu_part_number>900-2G183-0000-001</gpu_part_number>
<gpu_module_id>0</gpu_module_id>
<inforom_version>
<img_version>G183.0200.00.02</img_version>
<oem_object>1.1</oem_object>
<ecc_object>5.0</ecc_object>
<pwr_object>N/A</pwr_object>
</inforom_version>
<gpu_operation_mode>
<current_gom>N/A</current_gom>
<pending_gom>N/A</pending_gom>
</gpu_operation_mode>
<gsp_firmware_version>515.105.01</gsp_firmware_version>
<gpu_virtualization_mode>
<virtualization_mode>Pass-Through</virtualization_mode>
<host_vgpu_mode>N/A</host_vgpu_mode>
</gpu_virtualization_mode>
<ibmnpu>
<relaxed_ordering_mode>N/A</relaxed_ordering_mode>
</ibmnpu>
<pci>
<pci_bus>00</pci_bus>
<pci_device>1E</pci_device>
<pci_domain>0000</pci_domain>
<pci_device_id>1EB810DE</pci_device_id>
<pci_bus_id>00000000:00:1E.0</pci_bus_id>
<pci_sub_system_id>12A210DE</pci_sub_system_id>
<pci_gpu_link_info>
<pcie_gen>
<max_link_gen>3</max_link_gen>
<current_link_gen>3</current_link_gen>
</pcie_gen>
<link_widths>
<max_link_width>16x</max_link_width>
<current_link_width>8x</current_link_width>
</link_widths>
</pci_gpu_link_info>
<pci_bridge_chip>
<bridge_chip_type>N/A</bridge_chip_type>
<bridge_chip_fw>N/A</bridge_chip_fw>
</pci_bridge_chip>
<replay_counter>0</replay_counter>
<replay_rollover_counter>0</replay_rollover_counter>
<tx_util>0 KB/s</tx_util>
<rx_util>0 KB/s</rx_util>
</pci>
<fan_speed>N/A</fan_speed>
<performance_state>P0</performance_state>
<clocks_throttle_reasons>
<clocks_throttle_reason_gpu_idle>Not Active</clocks_throttle_reason_gpu_idle>
<clocks_throttle_reason_applications_clocks_setting>Not Active</clocks_throttle_reason_applications_clocks_setting>
<clocks_throttle_reason_sw_power_cap>Not Active</clocks_throttle_reason_sw_power_cap>
<clocks_throttle_reason_hw_slowdown>Not Active</clocks_throttle_reason_hw_slowdown>
<clocks_throttle_reason_hw_thermal_slowdown>Not Active</clocks_throttle_reason_hw_thermal_slowdown>
<clocks_throttle_reason_hw_power_brake_slowdown>Not Active</clocks_throttle_reason_hw_power_brake_slowdown>
<clocks_throttle_reason_sync_boost>Not Active</clocks_throttle_reason_sync_boost>
<clocks_throttle_reason_sw_thermal_slowdown>Not Active</clocks_throttle_reason_sw_thermal_slowdown>
<clocks_throttle_reason_display_clocks_setting>Not Active</clocks_throttle_reason_display_clocks_setting>
</clocks_throttle_reasons>
<fb_memory_usage>
<total>15360 MiB</total>
<reserved>388 MiB</reserved>
<used>1032 MiB</used>
<free>13939 MiB</free>
</fb_memory_usage>
<bar1_memory_usage>
<total>256 MiB</total>
<used>5 MiB</used>
<free>251 MiB</free>
</bar1_memory_usage>
<compute_mode>Default</compute_mode>
<utilization>
<gpu_util>0 %</gpu_util>
<memory_util>0 %</memory_util>
<encoder_util>0 %</encoder_util>
<decoder_util>0 %</decoder_util>
</utilization>
<encoder_stats>
<session_count>0</session_count>
<average_fps>0</average_fps>
<average_latency>0</average_latency>
</encoder_stats>
<fbc_stats>
<session_count>0</session_count>
<average_fps>0</average_fps>
<average_latency>0</average_latency>
</fbc_stats>
<ecc_mode>
<current_ecc>Enabled</current_ecc>
<pending_ecc>Enabled</pending_ecc>
</ecc_mode>
<ecc_errors>
<volatile>
<sram_correctable>0</sram_correctable>
<sram_uncorrectable>0</sram_uncorrectable>
<dram_correctable>0</dram_correctable>
<dram_uncorrectable>0</dram_uncorrectable>
</volatile>
<aggregate>
<sram_correctable>0</sram_correctable>
<sram_uncorrectable>0</sram_uncorrectable>
<dram_correctable>0</dram_correctable>
<dram_uncorrectable>0</dram_uncorrectable>
</aggregate>
</ecc_errors>
<retired_pages>
<multiple_single_bit_retirement>
<retired_count>0</retired_count>
<retired_pagelist>
</retired_pagelist>
</multiple_single_bit_retirement>
<double_bit_retirement>
<retired_count>0</retired_count>
<retired_pagelist>
</retired_pagelist>
</double_bit_retirement>
<pending_blacklist>No</pending_blacklist>
<pending_retirement>No</pending_retirement>
</retired_pages>
<remapped_rows>N/A</remapped_rows>
<temperature>
<gpu_temp>40 C</gpu_temp>
<gpu_temp_max_threshold>96 C</gpu_temp_max_threshold>
<gpu_temp_slow_threshold>93 C</gpu_temp_slow_threshold>
<gpu_temp_max_gpu_threshold>85 C</gpu_temp_max_gpu_threshold>
<gpu_target_temperature>N/A</gpu_target_temperature>
<memory_temp>N/A</memory_temp>
<gpu_temp_max_mem_threshold>N/A</gpu_temp_max_mem_threshold>
</temperature>
<supported_gpu_target_temp>
<gpu_target_temp_min>N/A</gpu_target_temp_min>
<gpu_target_temp_max>N/A</gpu_target_temp_max>
</supported_gpu_target_temp>
<power_readings>
<power_state>P0</power_state>
<power_management>Supported</power_management>
<power_draw>26.78 W</power_draw>
<power_limit>70.00 W</power_limit>
<default_power_limit>70.00 W</default_power_limit>
<enforced_power_limit>70.00 W</enforced_power_limit>
<min_power_limit>60.00 W</min_power_limit>
<max_power_limit>70.00 W</max_power_limit>
</power_readings>
<clocks>
<graphics_clock>585 MHz</graphics_clock>
<sm_clock>585 MHz</sm_clock>
<mem_clock>5000 MHz</mem_clock>
<video_clock>810 MHz</video_clock>
</clocks>
<applications_clocks>
<graphics_clock>585 MHz</graphics_clock>
<mem_clock>5001 MHz</mem_clock>
</applications_clocks>
<default_applications_clocks>
<graphics_clock>585 MHz</graphics_clock>
<mem_clock>5001 MHz</mem_clock>
</default_applications_clocks>
<max_clocks>
<graphics_clock>1590 MHz</graphics_clock>
<sm_clock>1590 MHz</sm_clock>
<mem_clock>5001 MHz</mem_clock>
<video_clock>1470 MHz</video_clock>
</max_clocks>
<max_customer_boost_clocks>
<graphics_clock>1590 MHz</graphics_clock>
</max_customer_boost_clocks>
<clock_policy>
<auto_boost>N/A</auto_boost>
<auto_boost_default>N/A</auto_boost_default>
</clock_policy>
<voltage>
<graphics_volt>N/A</graphics_volt>
</voltage>
<supported_clocks>
<supported_mem_clock>
<value>5001 MHz</value>
<supported_graphics_clock>1590 MHz</supported_graphics_clock>
<supported_graphics_clock>1575 MHz</supported_graphics_clock>
<supported_graphics_clock>1560 MHz</supported_graphics_clock>
<supported_graphics_clock>1545 MHz</supported_graphics_clock>
<supported_graphics_clock>1530 MHz</supported_graphics_clock>
<supported_graphics_clock>1515 MHz</supported_graphics_clock>
<supported_graphics_clock>1500 MHz</supported_graphics_clock>
<supported_graphics_clock>1485 MHz</supported_graphics_clock>
<supported_graphics_clock>1470 MHz</supported_graphics_clock>
<supported_graphics_clock>1455 MHz</supported_graphics_clock>
<supported_graphics_clock>1440 MHz</supported_graphics_clock>
<supported_graphics_clock>1425 MHz</supported_graphics_clock>
<supported_graphics_clock>1410 MHz</supported_graphics_clock>
<supported_graphics_clock>1395 MHz</supported_graphics_clock>
<supported_graphics_clock>1380 MHz</supported_graphics_clock>
<supported_graphics_clock>1365 MHz</supported_graphics_clock>
<supported_graphics_clock>1350 MHz</supported_graphics_clock>
<supported_graphics_clock>1335 MHz</supported_graphics_clock>
<supported_graphics_clock>1320 MHz</supported_graphics_clock>
<supported_graphics_clock>1305 MHz</supported_graphics_clock>
<supported_graphics_clock>1290 MHz</supported_graphics_clock>
<supported_graphics_clock>1275 MHz</supported_graphics_clock>
<supported_graphics_clock>1260 MHz</supported_graphics_clock>
<supported_graphics_clock>1245 MHz</supported_graphics_clock>
<supported_graphics_clock>1230 MHz</supported_graphics_clock>
<supported_graphics_clock>1215 MHz</supported_graphics_clock>
<supported_graphics_clock>1200 MHz</supported_graphics_clock>
<supported_graphics_clock>1185 MHz</supported_graphics_clock>
<supported_graphics_clock>1170 MHz</supported_graphics_clock>
<supported_graphics_clock>1155 MHz</supported_graphics_clock>
<supported_graphics_clock>1140 MHz</supported_graphics_clock>
<supported_graphics_clock>1125 MHz</supported_graphics_clock>
<supported_graphics_clock>1110 MHz</supported_graphics_clock>
<supported_graphics_clock>1095 MHz</supported_graphics_clock>
<supported_graphics_clock>1080 MHz</supported_graphics_clock>
<supported_graphics_clock>1065 MHz</supported_graphics_clock>
<supported_graphics_clock>1050 MHz</supported_graphics_clock>
<supported_graphics_clock>1035 MHz</supported_graphics_clock>
<supported_graphics_clock>1020 MHz</supported_graphics_clock>
<supported_graphics_clock>1005 MHz</supported_graphics_clock>
<supported_graphics_clock>990 MHz</supported_graphics_clock>
<supported_graphics_clock>975 MHz</supported_graphics_clock>
<supported_graphics_clock>960 MHz</supported_graphics_clock>
<supported_graphics_clock>945 MHz</supported_graphics_clock>
<supported_graphics_clock>930 MHz</supported_graphics_clock>
<supported_graphics_clock>915 MHz</supported_graphics_clock>
<supported_graphics_clock>900 MHz</supported_graphics_clock>
<supported_graphics_clock>885 MHz</supported_graphics_clock>
<supported_graphics_clock>870 MHz</supported_graphics_clock>
<supported_graphics_clock>855 MHz</supported_graphics_clock>
<supported_graphics_clock>840 MHz</supported_graphics_clock>
<supported_graphics_clock>825 MHz</supported_graphics_clock>
<supported_graphics_clock>810 MHz</supported_graphics_clock>
<supported_graphics_clock>795 MHz</supported_graphics_clock>
<supported_graphics_clock>780 MHz</supported_graphics_clock>
<supported_graphics_clock>765 MHz</supported_graphics_clock>
<supported_graphics_clock>750 MHz</supported_graphics_clock>
<supported_graphics_clock>735 MHz</supported_graphics_clock>
<supported_graphics_clock>720 MHz</supported_graphics_clock>
<supported_graphics_clock>705 MHz</supported_graphics_clock>
<supported_graphics_clock>690 MHz</supported_graphics_clock>
<supported_graphics_clock>675 MHz</supported_graphics_clock>
<supported_graphics_clock>660 MHz</supported_graphics_clock>
<supported_graphics_clock>645 MHz</supported_graphics_clock>
<supported_graphics_clock>630 MHz</supported_graphics_clock>
<supported_graphics_clock>615 MHz</supported_graphics_clock>
<supported_graphics_clock>600 MHz</supported_graphics_clock>
<supported_graphics_clock>585 MHz</supported_graphics_clock>
<supported_graphics_clock>570 MHz</supported_graphics_clock>
<supported_graphics_clock>555 MHz</supported_graphics_clock>
<supported_graphics_clock>540 MHz</supported_graphics_clock>
<supported_graphics_clock>525 MHz</supported_graphics_clock>
<supported_graphics_clock>510 MHz</supported_graphics_clock>
<supported_graphics_clock>495 MHz</supported_graphics_clock>
<supported_graphics_clock>480 MHz</supported_graphics_clock>
<supported_graphics_clock>465 MHz</supported_graphics_clock>
<supported_graphics_clock>450 MHz</supported_graphics_clock>
<supported_graphics_clock>435 MHz</supported_graphics_clock>
<supported_graphics_clock>420 MHz</supported_graphics_clock>
<supported_graphics_clock>405 MHz</supported_graphics_clock>
<supported_graphics_clock>390 MHz</supported_graphics_clock>
<supported_graphics_clock>375 MHz</supported_graphics_clock>
<supported_graphics_clock>360 MHz</supported_graphics_clock>
<supported_graphics_clock>345 MHz</supported_graphics_clock>
<supported_graphics_clock>330 MHz</supported_graphics_clock>
<supported_graphics_clock>315 MHz</supported_graphics_clock>
<supported_graphics_clock>300 MHz</supported_graphics_clock>
</supported_mem_clock>
<supported_mem_clock>
<value>405 MHz</value>
<supported_graphics_clock>645 MHz</supported_graphics_clock>
<supported_graphics_clock>630 MHz</supported_graphics_clock>
<supported_graphics_clock>615 MHz</supported_graphics_clock>
<supported_graphics_clock>600 MHz</supported_graphics_clock>
<supported_graphics_clock>585 MHz</supported_graphics_clock>
<supported_graphics_clock>570 MHz</supported_graphics_clock>
<supported_graphics_clock>555 MHz</supported_graphics_clock>
<supported_graphics_clock>540 MHz</supported_graphics_clock>
<supported_graphics_clock>525 MHz</supported_graphics_clock>
<supported_graphics_clock>510 MHz</supported_graphics_clock>
<supported_graphics_clock>495 MHz</supported_graphics_clock>
<supported_graphics_clock>480 MHz</supported_graphics_clock>
<supported_graphics_clock>465 MHz</supported_graphics_clock>
<supported_graphics_clock>450 MHz</supported_graphics_clock>
<supported_graphics_clock>435 MHz</supported_graphics_clock>
<supported_graphics_clock>420 MHz</supported_graphics_clock>
<supported_graphics_clock>405 MHz</supported_graphics_clock>
<supported_graphics_clock>390 MHz</supported_graphics_clock>
<supported_graphics_clock>375 MHz</supported_graphics_clock>
<supported_graphics_clock>360 MHz</supported_graphics_clock>
<supported_graphics_clock>345 MHz</supported_graphics_clock>
<supported_graphics_clock>330 MHz</supported_graphics_clock>
<supported_graphics_clock>315 MHz</supported_graphics_clock>
<supported_graphics_clock>300 MHz</supported_graphics_clock>
</supported_mem_clock>
</supported_clocks>
<processes>
<process_info>
<gpu_instance_id>N/A</gpu_instance_id>
<compute_instance_id>N/A</compute_instance_id>
<pid>675</pid>
<type>G</type>
<process_name>/usr/lib/xorg/Xorg</process_name>
<used_memory>22 MiB</used_memory>
</process_info>
<process_info>
<gpu_instance_id>N/A</gpu_instance_id>
<compute_instance_id>N/A</compute_instance_id>
<pid>5762</pid>
<type>C</type>
<process_name>python</process_name>
<used_memory>1005 MiB</used_memory>
</process_info>
</processes>
<accounted_processes>
</accounted_processes>
</gpu>
</nvidia_smi_log>