Amazon announced the next generation Graviton3 processors and the preview of the EC2 C7g instances. The cloud provider claims that the new instance type running the latest Arm-based processors will provide better compute, higher floating-point and faster cryptographic performances.
The C7g instances are designed for compute-intensive workloads such as HPC, batch processing, electronic design automation (EDA), media encoding, scientific modeling, distributed analytics, and CPU-based machine learning inferencing. As for the current generation C6 instances, Amazon has a Graviton first approach, announcing the Arm based (C6g) instances earlier than the Intel based ones (C6i).
Jeff Barr, vice president and chief evangelist at AWS, explains the benefits of the C7g instances:
While we are still optimizing these instances, it is clear that the Graviton3 is going to deliver amazing performance. In comparison to the Graviton2, the Graviton3 will deliver up to 25% more compute performance and up to twice as much floating point & cryptographic performance. On the machine learning side, Graviton3 includes support for bfloat16 data and will be able to deliver up to 3x better performance.
As for all Graviton processors, the new ones will include dedicated cores and caches for each vCPU and they will rely on the AWS Nitro System. Matthew Wilson, VP and distinguished engineer at AWS, adds:
Increasing the width of memory access is also part of improving performance. Graviton3 is the first public cloud instance to utilize DDR5 memory, which increases memory bandwidth by 50%.
Colin Percival, computer Scientist and FreeBSD security officer emeritus, comments:
Does it really count as "first public cloud instance" when it's not GA yet? (...) As long as it's just a preview I think it's cheating to claim that you got here first.
Focusing on sustainability, Adam Selipsky, CEO at AWS, tweets:
AWS Graviton3 processors are more energy efficient, using up to 60% less energy for the same performance versus comparable EC2 instances. This is a big step for a sustainable cloud!
For a server with two sockets/server and an industry-standard 42U rack, AWS suggests that traditional configurations ran out of rack power before space. Running Graviton3 instances, Amazon runs out of space before power. Using a chiplet design with seven different dies, the Graviton3 servers are apparently running on Arm’s Neoverse V1 core.
While some developers on Reddit question the choice of one more processor, user gnocchicotti comments:
Five years ago when Amazon laid the groundwork for these chips and their ecosystem, Intel had them over a barrel. Now AWS has 3 viable CPU options while their competitors only have 2 (...) Even if they only break even on the entire project, the risk that they mitigated by multisourcing probably made it worthwhile.
Prices for the G7g instances have not been announced yet. To try the new instances, customers have to apply to join the preview.
