WORK IN PROGRESS | Services are in active development and are subject to change.


Contents


Overview

Charity Engine provides compute resources in the form of standard instance-types.  Specs for these are outlined below.

Related resources:

  • Pricing 
  • Technical documentation (*includes details on the API, CLI, GnuParallel)
  • Our simple WebUI (*just try it…)

CPU Instance Types

During the beta test of the marketplace, two instance types are available, per the table below. (Other resource-types are available via custom arrangements; contact us for details.)

General Purpose Computing - Linux

vCPU

RAM

Benchmark

C.2x2

2

2

>500

C.2x4

2

4

>1000

C.4x4 (*coming soon)

4

4

>2000

Definitions

  • vCPU -  A full physical core.  Due to the heterogeneous nature of our network, specific brand and model are not specified, but provisioned CPUs will meet or exceed the benchmark specified for the Instance Type. (*See bullet below for further details on Benchmarks; and see "Special Features" below for functions to target specific hardware)
  • RAM -  The amount of RAM (GiB) allocated to each instance
  • Benchmark - Average Geekbench (v5) results (higher=better). Note: Benchmark results are for multi-threaded applications ("Multi-Core Score")
  • Other Instance Details
    • Data Transfer - See "4. Data Transfer", below. 

GPU Instance Types

GPU Computing -
Nvidia - Linux

GPU Model

GPU Count

vCPU

RAM

Single-GPU Benchmark

G.A5000.1.16.128

RTX A5000

1

16

128



202351
per card

G.A5000.2.32.256

RTX A5000

2

32

256

G.A5000.3.48.384

RTX A5000

3

48

384

G.A5000.4.64.512

RTX A5000

4

64

512

G.3090.1.16.128

RTX 3090

1

16

128



242754
per card

G.3090.2.32.256

RTX 3090

2

32

256

G.3090.3.48.384

RTX 3090

3

48

384

G.3090.4.64.512

RTX 3090

4

64

512

Definitions

  • GPU Model - The specific GPU hardware provided with the instance
  • GPU Count - The number of cards of the specified type
  • CPU - Due to the heterogeneous nature of our network, specific brand and model are not specified, but provisioned CPUs will meet or exceed capacity equal to a process with Geekbench score of 500 - or roughly one vCPU on an AWS c5.large instance. (*See "Special Features" below for functions to target specific hardware.)
  • RAM - The amount of RAM (GiB) allocated to each instance
  • Benchmark - Average Geekbench (v5) results (higher=better). For Nvidia: "CUDA Score"; for AMD: "OpenCL Score”
  • Other Instance Details:
    • Data Transfer - See "4. Data Transfer", below. 

Data Transfer

Each instance-hour includes a fixed allotment (MiB) of data-transfer (including import of the application container).  Additional transfer is billed per MiB. See the [Pricing page] for details.

Special Feature Flags

Workloads may require specific hardware and software features beyond those in the basic instance specification above.  This can be requested by use of flags in the instance request interface.  Note: use of such flags may {1} reduce available inventory, and {2} be subject to additional charges.

Supported flags are:

  • AVX - CPU with AVX support
  • Driver:nnn.nn - GPU driver version, e.g. 440.33.01
  • Tensorflow - CPU meeting common Tensorflow specs [*in development; contact us if needed.]

Flags are case insensitive.

Custom Arrangements

Custom arrangements may be possible; contact us if currently available instance types or feature-sets are not suitable for your workloads.

  • No labels