A new cloud service from Akamai is optimized for fast media processing. Processing footage runs up to 25 times faster than on CPUs.
Akamai’s cloud infrastructure achieves these performance gains thanks to Nvidia RTX 4000 Ada Generation GPUs, built on the latest Ada Lovelace architecture. Thanks to NVENC and NVDENC (Nvidia’s proprietary encoding and decoding solutions, respectively), video processing is many times faster than on a CPU.
NVENC and NVDENC support all prominent codecs such as H. 264, H. 265, VP9 and AV1. Akamai notes that processing virtual reality and augmented reality content can also be handled on the Nvidia hardware.
Other purposes?
Akamai has traditionally focused on content delivery networks (CDNs), with video processing being one of the most challenging use cases. It makes sense, therefore, that the company focuses primarily on this when announcing the new service. Still, there is more to run on Nvidia hardware, if Akamai is to be believed.
For example, Akamai talks about GenAI and machine learning purposes for which the RTX 4000 Ada Generation cards are also suitable. The company also mentions data analysis, scientific computing, gaming, graphics rendering and HPC. However, some of these are rather odd use cases for this kind of hardware in particular. After all, the RTX 4000 in question is essentially a workstation card and should not be confused with other Nvidia products. For example, it in no way comes close to the AI performance of an H100 and its physical GPU chip (AD104) is a lot less powerful than the AD102 found in the RTX 4090.
In other words: Akamai as of now offers a service that, while suitable for video processing, does not suddenly provide the best AI hardware in an accessible way. After all, the RTX 4000, for example, can be rented from CoreWeave on a cloud basis for $24 cents per hour, while the H100 costs roughly seventeen times more ($4.25 per hour).
Also read: Noname Security wants to sell itself to Akamai: an inevitable move?