Introducing 8 Extremely Powerful CPUs and GPUs to Datalore
Hello!
Today we are introducing 8 new powerful CPU and GPU machines to Datalore Professional and a flexible new way to spend your computation quota – Datalore credits!
Read on to learn how you can now easily analyze tens of gigabytes of data and train your Machine Learning and Deep Learning models on high-performance CPUs and GPUs!
Meet new CPUs and GPUs
Enjoy up to 96GB of RAM on CPU XXL and the power of up to 8 Nvidia T4 GPUs with a Multi-GPU L in Datalore!
The new CPU and GPU machines are available for Datalore Professional users only. Community users will get 120 hours of CPU S monthly computations, as before. To get access to the new machines, go to the Account Settings → Billing plans and upgrade your account to Datalore Professional!
Upgrade to Datalore Professional
Below you can find a full list of the available machines and their characteristics:
| Machine name | Type | RAM | nCPUs | nGPUs | Price per hour in Datalore Credits |
| CPU S | CPU | 4 | 2 | – | 0.0226 |
| CPU L | CPU | 16 | 2 | – | 0.0537 |
| CPU XL | CPU | 32 | 16 | – | 0.3425 |
| CPU XXL | CPU | 96 | 48 | – | 1.0287 |
| GPU S | GPU | 16 | 4 | 1 | 0.2125 |
| GPU M | GPU | 32 | 8 | 1 | 0.3028 |
| GPU L | GPU | 64 | 16 | 1 | 0.4843 |
| GPU XL | GPU | 128 | 32 | 1 | 0.8745 |
| GPU XXL | GPU | 256 | 64 | 1 | 1.7482 |
| Multi-GPU S | Multi-GPU | 192 | 48 | 4 | 1.5735 |
| Multi-GPU L | Multi-GPU | 384 | 96 | 8 | 3.7015 |
A glimpse behind the scenes of Datalore’s compute
Under the hood, for each new notebook computation Datalore launches a new AWS EC2 machine. By default a spot machine is launched, which helps keep the computation price as low as possible.
However, during peak hours on the AWS side there can be a wait time of 5-10 minutes to launch CPU XL, CPU XXL, and GPU M – Multi-GPU L. In some cases, large spot instances might not become available within a session, and Datalore will launch an on-demand machine instead. We will cover the extra costs (normally 3x) for launching on-demand machines in such cases.
As part of the feature roadmap, we will further introduce a choice between spot and on-demand types of machines to make the cost of each machine more transparent for users.
Datalore Credits
Many of our users have reported that they need only large CPU machine time, while some others need only GPU time. We want to make computation options flexible in Datalore, so we’re introducing Datalore credits for Professional users to spend on whichever machines they prefer.
As part of the Professional plan, users will get 12 Free Datalore credits monthly, which will be enough to cover 120 hours of computation on CPU L and 20 hours of Computations on GPU S. As before, the Professional plan will also include 750 hours of computations on CPU S to ensure access to your work, even if you run out of Datalore credits.
12 free Datalore credits will renew on a monthly basis and expire if you downgrade to a Community plan.
⚠️ Today, Professional users will receive 12 Datalore credits on their balance. The previous monthly quota for CPU M and GPU S will be reset.
New Machine and Storage Usage Reports
To make it clear how you are spending your new computation quota, we’ve redesigned the Machine and Storage users report and packed them with helpful information.
From the Account Settings → Billing Plans → Billing tab you can get a brief overview of the recent quota spendings and download the detailed Machine Usage and Storage Usage reports in CSV and XLSX formats.
Feature Roadmap
The roadmap for the future includes:
- Choosing between Spot and On-demand machines
- Purchasing additional Paid Datalore credits
All of these features are a prerequisite for launching a cloud Team plan, the beta version of which should be available in Q1 2023.
We believe Datalore credits and more powerful CPUs and GPUs will help you get more out of your Datalore Professional subscription and go beyond your available computation quota.
If you have any feedback or concerns, simply reply to this email. We’d be happy to hear from you!
Kind regards,
The Datalore team
Subscribe to Datalore News and Updates
