Deep Learning: Intel launches its new processors for data centers

Intel took advantage of its Intel Vision event to announce on Tuesday the launch of its new deep learning processors for training and inference, the Habana Gaudi2 and Habana Greco, making AI more accessible and more valuable for data. center operators. At its Intel Vision event, the chipmaker also shared details about its IPU and GPU portfolios, all of which are aimed at business customers.

These new deployments are not important for Intel. “AI drives the data center,” said Eitan Medina, COO of Habana Labs, Intel’s data center team dedicated to in-depth study of AI processor technologies. “It’s even the largest and fastest growing application. But different customers use different blends for different applications,” he said.

Different use cases explain Intel’s investment in different data center chips. Habana processors are designed for customers who require in-depth computing analysis. The new Gaudi2 processor, for example, could improve vision modeling for applications used in autonomous vehicles, medical imaging and fault detection in manufacturing. As a reminder, Intel bought Habana Labs, an Israel-based programmable chipmaker, for approximately $ 2 billion in 2019.

Chips etched at 7 nm

The second -generation Gaudi2 and Greco chips are both engraved at 7 nanometers, an improvement over the 16 nanometers of the first generation. They are built on the high efficiency architecture of Havana. According to Intel, the Gaudi2 provides twice the training throughput of Nvidia’s A100-80GB GPU for the ResNet-50 computer vision model and the BERT natural language processing model.

“Compared to the A100 GPU, which is implemented on the same process node and is roughly the same size, the Gaudi2 offers higher learning performance, as shown by the comparison of apples-to-apples with main task ”, we explained on the side of Habana Labs. “The architecture of this deep learning acceleration is fundamentally better and supported by a strong roadmap.”

Compared to the first generation Gaudi, the Gaudi2 offers up to 40% better price performance on the AWS Cloud with Amazon EC2 DL1 instances and is in place with the Supermicro X12 Gaudi Learning Server. The chip introduces an integrated media processing engine for compressed media and host subsystem offloading. The Gaudi2 triples the built-in memory capacity from 32 GB to 96 GB of the HBM2E with a bandwidth of 2.45 TB/sec. This chip includes 24 x 100GbE RoCE RDMA NICs, on-chip, for scaling and scaling using standard Ethernet.

The new Arctic Sound GPU has been announced

Gaudi2 processors are now available for Habana customers. Habana has partnered with Supermicro to bring the Supermicro Gaudi2 training server to market this year. Meanwhile, the second -generation Greco inference chip will be available to select customers in the second half of this year. Greco’s second generation includes increased on-board memory, which increases bandwidth five times and increases on-chip memory from 50 to 120 MB. It also adds additional performance to media decoding and processing, while offering a smaller form factor for computational efficiency.

“Gaudi2 can help Intel customers train larger and more complex deep learning workloads with speed and efficiency, and we look forward to the inference efficiencies that Greco will bring.” Sandra Rivera, executive vice president of Intel, in a statement.

Intel also released on Tuesday an expanded roadmap for its infrastructure processing unit (IPU) portfolio. Intel originally built IPUs for cloud computing giants – hyperscalers like Google and Facebook – but it’s now expanding the access of those units to other customers. Intel will ship two IPUs next year: Mount Evans, Intel’s first ASIC IPU, and Oak Springs Canyon, Intel’s second -generation FPGA IPU shipped to Google and other service providers.

In 2023 and 2024, Intel plans to launch its third -generation 400GB IPUs, codenamed Mount Morgan and Hot Springs Canyon. In 2025 and 2026, Intel plans to ship 800 GB IPUs to customers and partners. Intel also gave news about its data center GPU, named Arctic Sound. Designed for media transcoding, visual graphics and inference in the cloud, the Arctic Sound-M (ATS-M) is the industry’s first discrete GPU with an AV1 hardware encoder. It offers performance of 150 trillion operations per second (TOPS). The ATS-M will be available in two form factors and more than 15 system designs from partners including Dell, Supermicro, Inspur and H3C. Its launch is scheduled for the third quarter of this year.


Leave a Comment