GPUServer
Params
name
A human readable description of the object.
server_type
description to be done
gpu_power
On premise gpu server gpu power from estimating the carbon footprint of bloom in watt / gpu.
gpu_idle_power
On premise gpu server gpu idle power from estimating the carbon footprint of bloom in watt / gpu.
ram_per_gpu
On premise gpu server ram per gpu from estimating the carbon footprint of bloom in gigabyte_ram / gpu.
carbon_footprint_fabrication_per_gpu
On premise gpu server carbon footprint one gpu from estimating the carbon footprint of bloom in kilogram / gpu.
average_carbon_intensity
Average carbon intensity of on premise gpu server electricity in gram / kilowatt_hour.
compute
Nb gpus of on premise gpu server from e-footprint hypothesis in gpu.
carbon_footprint_fabrication_without_gpu
On premise gpu server carbon footprint without gpu from estimating the carbon footprint of bloom in kilogram.
lifespan
Lifespan of on premise gpu server in year.
power_usage_effectiveness
Pue of on premise gpu server in dimensionless.
utilization_rate
On premise gpu server utilization rate in dimensionless.
base_compute_consumption
Base gpu consumption of on premise gpu server from e-footprint hypothesis in gpu.
base_ram_consumption
Base ram consumption of on premise gpu server from e-footprint hypothesis in gigabyte_ram.
storage
An instance of Storage.
fixed_nb_of_instances
User defined number of on premise gpu server instances. Can be an EmptyExplainableObject in which case the optimum number of instances will be computed, or an ExplainableQuantity with a dimensionless value, in which case e-footprint will raise an error if the object needs more instances than available.
Backwards links
Calculated attributes
carbon_footprint_fabrication
ExplainableQuantity in kilogram, representing the On premise gpu server carbon footprint fabrication.
Example value: 3100.0 kilogram
Depends directly on:
- on premise GPU server carbon footprint without GPU from Estimating the Carbon Footprint of BLOOM
- Nb gpus of on premise GPU server from e-footprint hypothesis
- on premise GPU server carbon footprint one GPU from Estimating the Carbon Footprint of BLOOM
through the following calculations:
You can also visit the link to on premise GPU server carbon footprint fabrication’s full calculation graph.
power
ExplainableQuantity in watt, representing the On premise gpu server power.
Example value: 1600.0 watt
Depends directly on:
- on premise GPU server GPU power from Estimating the Carbon Footprint of BLOOM
- Nb gpus of on premise GPU server from e-footprint hypothesis
through the following calculations:
You can also visit the link to on premise GPU server power’s full calculation graph.
idle_power
ExplainableQuantity in watt, representing the On premise gpu server idle power.
Example value: 200.0 watt
Depends directly on:
- on premise GPU server GPU idle power from Estimating the Carbon Footprint of BLOOM
- Nb gpus of on premise GPU server from e-footprint hypothesis
through the following calculations:
You can also visit the link to on premise GPU server idle power’s full calculation graph.
ram
ExplainableQuantity in gigabyte_ram, representing the On premise gpu server ram.
Example value: 320.0 gigabyte_ram
Depends directly on:
- on premise GPU server RAM per GPU from Estimating the Carbon Footprint of BLOOM
- Nb gpus of on premise GPU server from e-footprint hypothesis
through the following calculations:
You can also visit the link to on premise GPU server RAM’s full calculation graph.
hour_by_hour_ram_need
On premise gpu server hour by hour ram need in gigabyte_ram.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in GB_ram:
first 10 vals [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
last 10 vals [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
Depends directly on:
- Hourly Manually defined GPU job average occurrences across usage patterns
- RAM needed on server on premise GPU server to process Manually defined GPU job from e-footprint hypothesis
- Hourly Generative AI model job average occurrences across usage patterns
- No additional GPU RAM needed because model is already loaded in memory from e-footprint hypothesis
through the following calculations:
You can also visit the link to on premise GPU server hour by hour ram need’s full calculation graph.
hour_by_hour_compute_need
On premise gpu server hour by hour compute need in gpu.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in gpu:
first 10 vals [0.0, 0.01, 0.02, 0.0, 0.01, 0.01, 0.01, 0.01, 0.01, 0.01],
last 10 vals [0.0, 0.0, 0.01, 0.0, 0.02, 0.02, 0.01, 0.01, 0.0, 0.01]
Depends directly on:
- Hourly Manually defined GPU job average occurrences across usage patterns
- gpus needed on server on premise GPU server to process Manually defined GPU job from e-footprint hypothesis
- Hourly Generative AI model job average occurrences across usage patterns
- Generative AI model job nb of required GPUs during inference
through the following calculations:
You can also visit the link to on premise GPU server hour by hour compute need’s full calculation graph.
occupied_ram_per_instance
ExplainableQuantity in gigabyte_ram, representing the Occupied ram per on premise gpu server instance including services.
Example value: 17.52 gigabyte_ram
Depends directly on:
- Base RAM consumption of on premise GPU server from e-footprint hypothesis
- Generative AI model base RAM consumption
through the following calculations:
You can also visit the link to Occupied RAM per on premise GPU server instance including services’s full calculation graph.
occupied_compute_per_instance
ExplainableQuantity in gpu, representing the Occupied cpu per on premise gpu server instance including services.
Example value: 0.0 gpu
Depends directly on:
through the following calculations:
You can also visit the link to Occupied CPU per on premise GPU server instance including services’s full calculation graph.
available_ram_per_instance
ExplainableQuantity in gigabyte_ram, representing the Available ram per on premise gpu server instance.
Example value: 270.48 gigabyte_ram
Depends directly on:
- on premise GPU server RAM
- on premise GPU server utilization rate
- Occupied RAM per on premise GPU server instance including services
through the following calculations:
You can also visit the link to Available RAM per on premise GPU server instance’s full calculation graph.
available_compute_per_instance
ExplainableQuantity in gpu, representing the Available cpu per on premise gpu server instance.
Example value: 3.6 gpu
Depends directly on:
- Nb gpus of on premise GPU server from e-footprint hypothesis
- on premise GPU server utilization rate
- Occupied CPU per on premise GPU server instance including services
through the following calculations:
You can also visit the link to Available CPU per on premise GPU server instance’s full calculation graph.
raw_nb_of_instances
Hourly raw number of on premise gpu server instances in concurrent.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in concurrent:
first 10 vals [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
last 10 vals [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0]
Depends directly on:
- on premise GPU server hour by hour ram need
- Available RAM per on premise GPU server instance
- on premise GPU server hour by hour compute need
- Available CPU per on premise GPU server instance
through the following calculations:
You can also visit the link to Hourly raw number of on premise GPU server instances’s full calculation graph.
nb_of_instances
Hourly number of on premise gpu server instances in concurrent.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in concurrent:
first 10 vals [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
last 10 vals [1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0]
Depends directly on:
- Hourly raw number of on premise GPU server instances
- User defined number of on premise GPU server instances
- Server type of on premise GPU server from e-footprint hypothesis
through the following calculations:
You can also visit the link to Hourly number of on premise GPU server instances’s full calculation graph.
instances_fabrication_footprint
Hourly on premise gpu server instances fabrication footprint in kilogram.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in kg:
first 10 vals [0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06],
last 10 vals [0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06, 0.06]
Depends directly on:
- Hourly number of on premise GPU server instances
- on premise GPU server carbon footprint fabrication
- Lifespan of on premise GPU server
through the following calculations:
You can also visit the link to Hourly on premise GPU server instances fabrication footprint’s full calculation graph.
instances_energy
Hourly energy consumed by on premise gpu server instances in kilowatt_hour.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in kWh:
first 10 vals [0.24, 0.25, 0.25, 0.24, 0.25, 0.24, 0.24, 0.24, 0.25, 0.24],
last 10 vals [0.24, 0.24, 0.25, 0.24, 0.25, 0.25, 0.24, 0.24, 0.24, 0.24]
Depends directly on:
- Hourly number of on premise GPU server instances
- on premise GPU server idle power
- PUE of on premise GPU server
- Hourly raw number of on premise GPU server instances
- on premise GPU server power
through the following calculations:
You can also visit the link to Hourly energy consumed by on premise GPU server instances’s full calculation graph.
energy_footprint
Hourly on premise gpu server energy footprint in kilogram.
Example value: 26298 values from 2024-12-31 23:00:00+00:00 to 2028-01-01 17:00:00+00:00 in kg:
first 10 vals [0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02],
last 10 vals [0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02, 0.02]
Depends directly on:
- Hourly energy consumed by on premise GPU server instances
- Average carbon intensity of on premise GPU server electricity
through the following calculations:
You can also visit the link to Hourly on premise GPU server energy footprint’s full calculation graph.