Intel touts the superiority of its processors over the competition in many AI-related applications, and also introduces a plethora of related product.
Today, Intel is launching its new 3rd Gen Intel Xeon Scalable processors, formerly code-named “Ice Lake,” the foundation of Intel’s data center platform. The first server products to use Intel’s 10nm manufacturing process, they deliver up to 40 cores per processor. While Ice Lake was the highlight of the news, Intel also made multiple related announcements, including a new 200 Gbe Ethernet adapter, the Intel Optane Persistent Memory 200 Series, which has 32% higher memory bandwidth compared to the 100 series, and the availability of the new Intel Agilex FPGA.
“With our 3rd Gen Xeon Scalable processors, we are bringing artificial intelligence everywhere, to be proficient at serving all workloads from the edge to the cloud through the network,” said Lisa Spelman, Corporate Vice President, Xeon and Memory Group, Data Platforms Group at Intel.
Intel stressed that over 50 million Xeon Scalable Processors have been shipped, and they have been the historic market leader, but they have come under pressure from the second and third generations of AMD’s EPYC processors, the latter of which just came out in the middle of March. This launch has been eagerly anticipated by the industry as a result.
Intel’s 3rd Gen processors do have one fundamental disadvantage over AMD’s – they have fewer cores. Intel now offers a 40-core max, compared to AMD’s 64. Intel countered that unpleasant reality by emphasizing the acceleration capabilities of their processors through the building in of Intel DL Boost’s AI, and of Intel Crypto Acceleration, which reduces the penalty of implementing pervasive data encryption, and provides higher throughput with fast and strong encryption for AVX-512 ISA.
“This is the only data centre processor with built in AI acceleration – AI made flexible,” Spelman said. “We really see a huge amount of opportunity and we think customers will see this flexible performance deliver for them.” Intel says that the hardware and software optimizations in the new processors deliver 74% faster AI performance compared with the prior generation. They also say their testing shows up to 1.5 times higher performance across a broad mix of 20 popular AI workloads versus AMD EPYC 7763 and up to 1.3 times higher performance on a broad mix of 20 popular AI workloads versus Nvidia A100 GPU.
The core metrics of the 3rd Gen platform include support for up to 6TB of system memory per socket, up to 8 channels of DD-3200 memory per socket, up to 64 lances of PCIe Gen 4 per socket, and up to 2.6x memory increase compared to 2nd Gen Intel Scalable.
Spelman also touted how security has been enhanced in the new chips.
“We now bring Intel SGX into 2 socket processors, to give the smallest attack surface of any data centre computing technology,” she said. It also boasts a 4000x increase in protected enclave size to 1TB.
The new processors also introduce two new security features. Intel Total Memory Encryption is for basic bulk encryption of the entire memory space to protect against physical attacks. Intel Platform Firmware Resilience is for defending and recovering the underlying firmware layer to protect against permanent denial of service.
Intel also introduced its next generation of Intel Optane persistent memory, the 200 series.
“Intel Optane continues to grow as a competitive differentiator for Intel, with over 1000 Proof-of-Concepts, and over 80% converted into revenue deployments, Spelman said. Intel says using Intel Optane persistent memory 200 series with the 3rd Gen Intel Xeon Scalable processors lower costs per VM by up to 25% while delivering the same performance, achieves up to 2.5 times more transactions, and delivers up to 63% higher throughput for live video content and 33% more memory capacity, compared with the previous generations.
Intel also announced the Intel Optane SSD P5800X, which it termed the world’s fastest SSD for the data center.
“These new SSDs have up to 66x better QoS and up to 13x lower average latency compared to NAND SSDs,” Spelman said. They also have 4x greater random 4K mixed read/write IOPs, and 67% higher endurance.”
Also announced was the Intel SSD D5-P5316, the industry’s first 144-L PCIe QLC NAND. It has Intel’s most advanced Quad-Level Cell NAND technology, and delivers industry-leading SSD storage density for the data center and the high bandwidth of PCIe 4.0 interface. It has up to 2 times higher sequential read performance, up to 38% higher random read performance, up to 48% better latency and up to 5 times higher endurance compared with the previous-generation Intel QLC SSD family.
Intel also announced a new 200 Gbe Ethernet adapter, the E810-2CQDA2, which is capable of delivering up to 100 Gbps port speeds with features only available on Intel platforms. It increases network data throughput up to 200 Gbps per adapter for bandwidth intensive workloads, and has up to 2 times increase in resources for virtualized and containerized networks compared with the Intel Ethernet 700 series.
Intel also announced that the Intel Agilex FPGA family, which are critical to the rapidly growing category of Intel SmartNICs, is now shipping.
“This was announced in 2019 and is shipping today,” Spelman said. “It is a long cycle for FPGAs. It has 2x better fabric performance per watt compared to competing 7 nm FPGAs.” It also has up to 50% faster fabric performance than prior generations for 5G fronthaul gateway applications.
Spelman noted that the new Intel Xeon Scalable processors are supported by more than 500 ready-to-deploy Intel Market Ready and Intel Select Solutions that help to accelerate customer deployments. More than 80% of these have been refreshed to take advantage of 3rd Gen Intel Xeon Scalable processors.
“Solutions Delivery is an area where we have a head start over competitors, with 500-plus partner solutions optimized for Intel platforms that span from the edge to the cloud,” she said.