In September of final 12 months, Intel launched its Xeon 6900P Granite Ridge processors and obtained themselves again within the sport towards AMD. This week, Intel continued to develop its processor line with the launch of the Xeon 6500P and Xeon 6700P server processors.
The brand new processors can be found in as much as 86 cores per socket and may scale to 4 and eight-socket platforms. Whereas the brand new processors are a step up when it comes to sockets, they’re just about a step down when it comes to their options; fairly than the 12 reminiscence channels of DDR5 provided by the 6900P, the 6500P and 6700P present as much as 8 DDR5 channels with some barely slower reminiscence choices. When it comes to PCI lanes, the 6500P and 6700P supply as much as 88 lanes, in comparison with 96 lanes for the 6900P. Like their massive brother, the 6900P, each processor traces help MRDIMM reminiscence. They’ve as much as 64 lanes of CLX 2.0, Intel AVX 512, and AMX (Superior Matrix Extensions). Different variations may be seen within the determine under (additionally together with the brand new Xeon 6 SoC edge processor line).
(Supply: Intel)
As a part of the discharge, Intel additionally touted the AI capabilities of the brand new processors optimized to seize a big share of the rising market, delivering management efficiency in CPU machine studying, smaller GenAI fashions, and GPU-accelerated workloads as a number CPU. The power to run the just lately launched DeepSeek mannequin on x86 servers with good reminiscence bandwidth and no GPU does level a spot for conventional CPU in utilizing fashions for inference.
When it comes to HPC efficiency, Intel launched a set of benchmarks that indicated an uptick of 1.2 to 1.5 occasions quicker than the Xeon 8592+ 64 core processor (with equal and most cores. The enhancements are notable as a result of they function within the energy envelope of 350. The graphic under depicts the outcomes. Additional impartial assessments are anticipated within the close to future.
(Supply: Intel)
Michelle Johnston Holthaus, interim co-CEO of Intel and CEO of Intel Merchandise, commented on the processors, “The Xeon 6 household delivers the business’s greatest CPU for AI and groundbreaking options for networking whereas concurrently driving effectivity and bringing down the overall price of possession.”
Edge processors
Intel additionally introduced the Xeon 6 SoC, or Granite Rapids D. With built-in 200G Ethernet, the Xeon 6 SoC is designed as an edge processor. The flagship model, Xeon 6726P-B, has 42 cores, 4 channels of DDR5-6400 reminiscence, 2.3GHz base frequency (2.9GHz turbo), a 235-watt TDP, and 200G Ethernet. The complete line begins at 12-cores with the 6503P-B Granite Rapids D SoC.
The Xeon 6 SoC line ranges from 12-42 cores (worth vary $934-$3,795), with a few of the decrease core depend fashions dropping the Ethernet to 100G. All fashions help Intel AMX (Superior Matrix Extensions)
Talking of Ethernet
Intel additionally introduced two new Ethernet controller and community adapter product traces to handle the rising calls for of enterprise, telecommunications, cloud, high-performance computing (HPC), edge, and AI functions. Preliminary availability contains dual-port 25GbE PCIe and OCP 3.0-compliant adapters.
The E830 line of controllers and community adapters delivers as much as 200GbE bandwidth with help for PCIe 5.0, versatile port configurations (x200GbE, 2×100/50/25/10GbE and eight×25/10GbE), and superior precision time capabilities, together with Precision Time Measurement (PTM). These adapters are optimized for high-density virtualized workloads, providing strong safety features and efficiency.
The E610 sequence controllers and community adapters present 10GBASE-T, 5G, 2.5G, and 1000BASE-T connectivity optimized for management aircraft operations. The 610 sequence provides excellent energy effectivity, 50% much less energy than earlier generations, superior manageability, and strong safety features that simplify community administration.
Course of ahead
As just lately reported at a convention in San Jose, California, senior principal engineer Steve Carson stated Intel has produced 30,000 wafers, yielding 1000’s of computing chips, in a single quarter with its new ASML’s excessive numerical aperture (NA) lithography machines.
This text first appeared on HPCwire.
