Habana Labs, which is a Tel Aviv based Artificial Intelligence startup has announced today that they will be releasing their Gaudi AI training processor. This will promise to simply beat all the systems which are GPU based on four factors. All the individual Gaudi chips are able to beat other GPUs because of their raw performance.
Gaudi will be available in the market as a standard PCIe card and it is known to support a total of eight ports of 100 GB Ethernet along with the mezzanine card. The company will soon be working on a system with eight mezzanine cards.
In the year 2018, Habana Labs launched Goya inference solutions. With the help of Gaudi, they are able to offer a complete solution for businesses of all kinds which are interested in using hardware over GPUs. The only reason it is able to perform in such a great way is because of the hardware.
“The CPU and GPU architecture started from solving a very different problem than deep learning,” Habana CBO Eitan Medina told me. “The GPU, almost by accident, happened to be just better because it has a higher degree of parallelism. However, if you start from a clean sheet of paper and analyze what a neural network looks like, you can, if you put really smart people in the same room […] come up with a better architecture.”
That’s what Habana did for its Goya processor and it is now taking what it learned from this to Gaudi.
It is easier for all the developers because Habana Labs is able to support all kinds of Artificial Intelligence framework along with ONNX format. Switching from one processor to another is very simple in this case.
“Training AI models require exponentially higher compute every year, so it’s essential to address the urgent needs of the data center and cloud for radically improved productivity and scalability. With Gaudi’s innovative architecture, Habana delivers the industry’s highest performance while integrating standards-based Ethernet connectivity, enabling unlimited scale,” said David Dahan, CEO of Habana Labs. “Gaudi will disrupt the status quo of the AI Training processor landscape.”
The company has spilled a secret that the power is not stored in just the processor, but how it is connected to the entire system and with other processors.
“With its new products, Habana has quickly extended from inference into training, covering the full range of neural-network functions,” said Linley Gwennap, principal analyst of The Linley Group. “Gaudi offers strong performance and industry-leading power efficiency among AI training accelerators. As the first AI processor to integrate 100G Ethernet links with RoCE support, it enables large clusters of accelerators built using industry-standard components.”