├── .gitignore └── README.md /.gitignore: -------------------------------------------------------------------------------- 1 | tmp_* 2 | -------------------------------------------------------------------------------- /README.md: -------------------------------------------------------------------------------- 1 | # Cloud GPU Vendors 2 | 3 | This guide compares on-demand GPU vendors to help Machine Learning practitioners 4 | pick their preferred platform. Whether you're an enthusiast, newcomer or expert 5 | this guide will help you track updates and keep your costs low. 6 | 7 | ## Cloud GPU Vendors 8 | 9 | - "-" means not available 10 | - "???" means maybe available 11 | - GPU prices are per hour 12 | - Storage prices are per GB/month 13 | 14 | | GPU Vendor | K80 Price | V100 Price | Storage Price | Last Updated | 15 | | ----------------------------------------------------------------------------------------------------- | --------- | ---------- | -------------- | ------------ | 16 | | [Google Colaboratory](https://colab.research.google.com/) (Free TPU***) | Free | -@ | 15GB max | 8-Sep-18 | 17 | | [Kaggle Kernels](https://www.kaggle.com/dansbecker/running-kaggle-kernels-with-a-gpu) (Free TPU***) | Free | - | 8GB max | 8-Sep-18 | 18 | | [Salamander](https://salamander.ai/) | $0.36 | $1.17 | $0.10 | 8-Sep-18 | 19 | | [Paperspace](https://www.paperspace.com/) | $0.59 | $2.30 | $0.10 | 8-Sep-18 | 20 | | [Vast.AI](https://vast.ai/) | $0.64 | $2.30 | - | 19-Sep-18 | 21 | | [Oracle Cloud](https://cloud.oracle.com/compute/gpu/features) | - | $2.25 | $0.04 | 8-Sep-18 | 22 | | [Spell](https://spell.run/) | $0.9 | $3.06 | - | 24-Oct-19 | 23 | | [Crestle](https://www.crestle.com/) | $0.59 | - | $0.42 | 8-Sep-18 | 24 | | [Microsoft Azure](https://docs.microsoft.com/en-us/azure/virtual-machines/windows/sizes-gpu) | $0.87 | $2.95 | $0.10 | 8-Sep-18 | 25 | | [LeaderGPU](https://www.leadergpu.com/) | - | $3.00 | $0.13 | 8-Sep-18 | 26 | | [Amazon Web Services (AWS)](https://aws.amazon.com/ec2/instance-types/p2/) | $0.90 | $3.06 | $0.10 | 8-Sep-18 | 27 | | [DeepCognition](https://deepcognition.ai/) | $0.90 | $3.06 | ??? | 8-Sep-18 | 28 | | [Google Cloud Platform (GCP)](https://cloud.google.com/gpu/) | $1.21 | $3.24 | $0.03 | 8-Sep-18 | 29 | | [FloydHub](https://www.floydhub.com/) | $1.20 | $4.20 | 100GB included | 8-Sep-18 | 30 | | [IBM Cloud](https://www.ibm.com/cloud/gpu) | $2.65 | - | ??? | 8-Sep-18 | 31 | | [Alibaba Cloud](https://www.alibabacloud.com/product/gpu?spm=a3c0i.11728334.1144220.1.23794a1clKeiCs) | - | ??? | ??? | 8-Sep-18 | 32 | | [Lambda Labs](https://lambdalabs.com/) | - | ??? | ??? | 8-Sep-18 | 33 | | [Nvidia Cloud](https://www.nvidia.com/en-us/data-center/gpu-cloud-computing/) | ??? | ??? | ??? | 8-Sep-18 | 34 | | [Riseml](https://riseml.com/) | - | ??? | ??? | 8-Sep-18 | 35 | | [Valohai](https://valohai.com/) | ??? | ??? | ??? | 8-Sep-18 | 36 | | [VScaler](https://www.vscaler.com/) | ??? | ??? | ??? | 8-Sep-18 | 37 | | [Hostkey](https://www.hostkey.com/) | ??? | - | ??? | 6-Oct-20 | 38 | 39 | 40 | - *** Colaboratory provides free access to Google's TPUs v2, and Kaggle has TPUs v3. 41 | - @ - It's now possible to [connect Google Colaboratory to a V100 by connecting it to a GCP instance.](https://colab.research.google.com/drive/1Xh0slMoD1dfi8iTQJbO18uuCilJMLoKc) 42 | ### Other GPUs 43 | 44 | | GPU Vendor | P100 Price | GTX 1080 Price | Storage Price | Last Updated | 45 | | ----------------------------------------------------------------------------------------------------- | ---------- | -------------- | ------------- | ------------ | 46 | | [Snark AI](https://snark.ai/) | $0.10 | - | ??? | 8-Sep-18 | 47 | | [Vast.AI](https://vast.ai/) | $1.65 | $0.10 | - | 8-Sep-18 | 48 | | [Oracle Cloud](https://cloud.oracle.com/compute/gpu/features) | $1.28 | - | $0.04 | 8-Sep-18 | 49 | | [Vector Dash](https://vectordash.com/) | - | $0.51 | ??? | 8-Sep-18 | 50 | | [LeaderGPU](https://www.leadergpu.com/) | $1.50 | $0.75 | $0.13 | 8-Sep-18 | 51 | | [Paperspace](https://www.paperspace.com/) | $1.72 | - | $0.10 | 8-Sep-18 | 52 | | [Microsoft Azure](https://docs.microsoft.com/en-us/azure/virtual-machines/windows/sizes-gpu) | $2.00 | - | $0.13 | 8-Sep-18 | 53 | | [Ovh](https://www.ovh.com/world/public-cloud/instances/prices/) | - | $1.15 | $0.09 | 8-Sep-18 | 54 | | [Neptune ML](https://neptune.ml/) | $2.15 | - | $0.28 | 8-Sep-18 | 55 | | [Google Cloud Platform (GCP)](https://cloud.google.com/gpu/) | $2.22 | - | $0.03 | 8-Sep-18 | 56 | | [Alibaba Cloud](https://www.alibabacloud.com/product/gpu?spm=a3c0i.11728334.1144220.1.23794a1clKeiCs) | ??? | - | ??? | 8-Sep-18 | 57 | | [Exoscale](https://exoscale.com/) | ??? | - | ??? | 8-Sep-18 | 58 | | [IBM Cloud](https://www.ibm.com/cloud/gpu) | ??? | - | ??? | 8-Sep-18 | 59 | | [Hostkey](https://www.hostkey.com/) | ??? | - | ??? | 6-Oct-20 | 60 | 61 | ## Notes 62 | 63 | - Google Colaboratory, Kaggle Kernels & Crestle are only for Jupyter Notebooks 64 | and limited to 6 / 12 / 24 hour sessions 65 | 66 | - The prices for Google Cloud Platform assume you've attached the GPUs to a 67 | "n1-standard-16" instance. 68 | 69 | - 250GB Paperspace volumes are $0.04 per GB; larger or smaller volumes cost more 70 | 71 | ### Free Compute Credits 72 | 73 | - Google Cloud Platform gives [$300 credits](https://cloud.google.com/free/) to anyone 74 | 75 | - AWS gives [$110 credits](https://education.github.com/pack/) to students, in partnership with GitHub 76 | 77 | - Azure gives [$100 credits](https://azure.microsoft.com/en-in/free/students/) to students 78 | 79 | - Paperspace gives [$15 credits](https://github.com/asiedubrempong/fastai_deeplearn_part1/blob/master/tools/paperspace.md#summary-of-charges) to fast.ai students 80 | 81 | - [Vast.AI](https://vast.ai/), [DeepCognition](https://deepcognition.ai/), [Snark AI](https://snark.ai/), [Salamander](https://salamander.ai/) & [Crestle](https://www.crestle.com/) give approximately $1 credits to anyone 82 | 83 | - [Hostkey](https://www.hostkey.com/) give free trials to science projects and competitions 84 | 85 | ## GPU Datasheets 86 | 87 | Nvidia manufactures all the GPUs listed here 88 | 89 | | GPU | TFlops (double / single precision) | 90 | | -------------------------------------------------------------------------------------------------------------------------- | ---------------------------------- | 91 | | [K80](https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/tesla-product-literature/TeslaK80-datasheet.pdf) | 1.9 / 5.6 | 92 | | [V100](https://images.nvidia.com/content/technologies/volta/pdf/tesla-volta-v100-datasheet-letter-fnl-web.pdf) | 7.8 / 15.7 | 93 | | [P100](https://images.nvidia.com/content/tesla/pdf/nvidia-tesla-p100-datasheet.pdf) | 5.3 / 10.6 | 94 | | [GTX 1080](https://international.download.nvidia.com/geforce-com/international/pdfs/GeForce_GTX_1080_Whitepaper_FINAL.pdf) | 8.8 / 8.8 | 95 | --------------------------------------------------------------------------------