LeaderGPU is granting customers needing infrastructure for AI workloads access to Nvidia GeForce RTX 6000 Ada graphics cards, one of the best GPUs out there, on a rental basis from the end of November 2023.
The server configuration on offer includes eight RTX 6000 Ada GPUs with eight units of 48GB GDDR6 RAM, and each one includes 18,176 CUDA cores. This also boasts up to 960GB/s of memory bandwidth.
CUDA cores, built by Nvidia, are similar to CPU cores, and are designed to process multiple calculations at the same time – and you’d need hundreds of thousands to effectively undertake the most intensive generative AI workloads, with the more CUDA cores you have access to, the better.
Unleashing AI via the cloud
These configurations, which can be rented for €0.25 (or 27¢) per minute, also include two Intel Xeon Gold 6248R CPUs, fitted alongside 384GB RAM, as well as 2TB NVMe storage capacity. They can run on Ubuntu as well as Windows Server.
Beginning November 30, you can rent this configuration for €5,682.22 ($6,060) per month, €1,420.55 ($1,515) per week, or €284.11 ($300) per day.
For those seeking to get an hourly rate, this is equivalent to just €15 ($16), meaning that renting up to a million CUDA cores would set you back just €105 ($110) per hour.
The hardware you’d need to run AI workloads is incredibly expensive to purchase, maintain, and run, not to mention being incredibly energy-intensive. As a result, the prospect of leasing the IT infrastructure required – particularly the GPUs – is becoming tempting for more organizations.
This is where entities like LeaderGPU come in, with the cloud becoming a much more viable route for businesses to run their AI workloads.
Japanese company Third Wave also last month announced a GPU Cloud, which gives customers access to fleets of the best GPUs at relatively affordable prices.
More from TechRadar Pro
from TechRadar - All the latest technology news https://ift.tt/oZMVud4
No comments:
Post a Comment