Intel debuts the Xeon 6 processor, aiming to modernize records facilities to kind out challenge AI workloads

Intel has announced its Xeon 6 chip to set up determined that records facilities can contend with the workloads mandatory as extra firms originate to deploy AI apps and models. The processor comes with two microarchitectures, an efficient core (E-core) and a performance core (P-core) and will seemingly be available in the market in the 6700 and 6900 platform offerings. Nonetheless, very finest the 6700 E-core version will launch on June 4. The 6900 P-core version will debut in Q3 2024, with others available in the market in Q1 2025.

“It’s all about enabling prospects in the business to converse accurate business outcomes,” Matt Langman, Intel’s vp and general supervisor, acknowledged in a records convention. “We witness every company turning into an AI company, whether these firms deserve to be extra efficient in their operations, extra efficient in their product trend, extra efficient—or unbiased correct extra practical—with their buyer engagements.”

As successfully as to the Xeon 6 processor, Intel is revealing pricing for its Gaudi 2 and Gaudi 3 accelerator chips for the first time. A fashioned AI equipment that contains the Gaudi 2, supplied in 2022, will tag $65,000, while the Gaudi 3, which became unveiled in April, will checklist at $125,000.

Xeon 6: What it is crucial to know

Intel calls the Xeon 6 a “sturdy computing platform” that “excels at each and each performance and efficiency,” two areas that it claims are essential for “meeting the ever-rising demands of the records center.” The processor supports a large different of divulge instances, from compute-intensive AI and high-performance computing to archaic challenge applications and these that are strength-efficient and high-density.

The draw is for firms to modernize their aging records center methods with Xeon 6 chips, which guarantees to enlarge tag savings, reduction meet sustainability desires, optimize bodily floor and rack station, and generate original digital capabilities.

The Xeon 6 P-core and E-core, previously code-named Granite Rapids and Sierra Wooded discipline, share a hardware platform basis and instrument stack. They’re geared up to toughen an increased core depend, hold a increased reminiscence bandwidth with DDR5, feature multiplexed combined depraved DIMM, increased inter-socket bandwidth with UPI 2.0, Compute Insist Link 2.0, Frequent OS and firmware, and extra.

Citing different requirements records facilities could per chance per chance per chance need, Intel presents two different microarchitectures for its Xeon 6. It boils down to workloads and capabilities. A P-core is continuously extra appropriate for these handling high-performance computing and compute-intensive AI. Nonetheless, it’s furthermore precious for the “broadest array of challenge applications.” Langman explains, “It in fact is these forms of latency-sensitive workloads that the advantages of a P-core will reduction meet or exceed these usages for prime single-threaded and high per-core performance capabilities.” Nonetheless, workloads gripping cloud-native web, microservices and digital providers will in all likelihood hold the benefit of Xeon 6 E-core, “taking unbiased correct thing about the increased density and improved performance per watt.”

So, what are the differences between the Xeon 6 6700 Series and the 6900 Series?

Xeon 6 Processors (6700 Series)

• Up to 144 Efficient-cores / 86 Efficiency-cores
• Socket Improve: 1S/2S and 4S/8S (P-core very finest) toughen
• Max TDP: Up to 350W per CPU
• Mem Channels: 8 channel reminiscence, up to 6400 MT/s DDR5 reminiscence, 8000 MT/s MCR DIMM reminiscence (P-core)
• PCIe/CXL: Up to 88 lanes PCIe 5.0/CXL 2.0
• UPI Hyperlinks: 4 UPI 2.0 hyperlinks, up to 24 GT/s

Xeon 6 Processors (6900 Series)

• Up to 288 Efficient-cores / 128 Efficiency-cores
• Socket Improve: 1S/2S toughen
• Max TDP: Up to 500W per CPU
• Mem Channels: 12 channel reminiscence, up to 6400 MT/s DDR5 reminiscence, 8800 MT/s MCR DIMM reminiscence (P-core)
• PCIe/CXL: Up to 96 lanes PCIe 5.0/CXL 2.0
• UPI Hyperlinks: 6 UPI 2.0 hyperlinks, up to 24 GT/s

What’s with the staggered launch?

No longer every variation of the Xeon 6 will be available in the market this present day. Intel says it is miles deliberately staggering the launch of different processor SKUs because it’s what prospects curiously need. “The difficulty that made the most sense for the differ of prospects we’re preserving with Xeon 6 is the roughly staggered present you witness. And what’s moving is as we’re getting closer to launching the first one, it is miles in a large present, and we’re getting a vary of market excitement around it, and it’s fun staring at it happen,” Ryan Tabrah, Intel’s vp and general supervisor for its E-core product line, clarifies.

Xeon 6 E-core bragging rights

Because the Xeon 6 E-core is the first to debut—and the first Xeon version to hold an E-core—Intel has supplied some metrics, asserting it enables rack-diploma consolidation of three-to-1, offers rack-diploma performance features of up to 4.2 instances and presents performance per watt features of up to 2.6 when in comparison with its 2d-generation Intel Xeon processors on media transcode workloads.

“As we uncover at this present day’s records center, and let’s advise you employ approximately 200 racks, which could per chance per chance per chance be a typical mid-sized records center deployment, about 15-kilowatt racks and 22 u servers, and you uncover at how they’re delivering media streams per 2d on 2d-gen Intel Xeon. You uncover on the advancements we’re delivering on Intel Xeon 6 with the perf per watt and overall perf performance development, and you procure that 3-to-1 rack consolidation down to 66 racks—extraordinary, unbelievable savings in rack station and rack functionality,” Langman says.

“Nonetheless…it’s now not very finest the savings in racks, on the different hand it’s the vitality savings and the carbon footprint. And what we witness is, from a like a flash vitality usage, a savings of upwards of 84,000-megawatt hours over the period of four years—essential vitality savings—combined with lowering carbon emissions of 34,000 metric plenty…over that identical period of time. So we procure the advantages of rack consolidation and enabling the business sustainability desires.”

Gaudi 3: Pricing and original system providers

Moreover the Xeon 6 records, Intel is releasing pricing for two of its latest Gaudi products. Designed to compete against Nvidia’s H100, these accelerator chips reduction put collectively and infer large language models however at a decrease total running tag.

Firms the usage of a identical old AI equipment with eight Intel Gaudi 2 accelerators with a current baseboard pays $65,000, which Intel estimates to be one-third of the price of comparable seemingly choices. A equivalent equipment with eight Intel Gaudi 3 accelerators will be $125,000, or about two-thirds the price of Intel’s opponents.

Despite the fact that Intel disclosed pricing, it didn’t present any additional miniature print about Gaudi 3’s availability past declaring that the chip is on the steady song for launch in Q3 2024.

To aid with its Chase-To-Market approach, the company is partnering with now not decrease than ten system providers, including Dell, Hewlett-Packard Endeavor, Lenovo, Supermicro and original additions Asus, Foxconn, Gigabyte, Inventec, Quanta and Wistron.

AI divulge instances for Xeon and Gaudi

When requested how Xeon and Gaudi will work collectively, Tabrah describes the connection as complementary. “It’s unleashing prospects’ records facilities,” he states. “They’re working into these strength bottlenecks. The records center itself is form of becoming the bottleneck. And whenever you happen to can unleash that records center for yourself unbiased correct by rapidly and without effort transferring over to a truly efficient general compute architecture and now not deserve to the touch anything, then you positively can traipse and unbiased correct innovate with the relaxation of your infrastructure to attain extra AI; that’s unbiased correct superior.”

Anil Nanduri, one other of Intel’s vice presidents, aspects out that it’s all about helping the challenge free up its records. With the introduction of generative AI, organizations are searching for to search out the accurate strategies to leverage large language models to hold solutions correct for the workloads they need and what’s very finest for the computational costs they’re joyful investing in. “You’re going to watch where accelerators will quiet speed these [LLMs], however a RAG-love divulge case where I can employ my datasets extra original, and I can recover outcomes for the prospects who are doing records discovery or different divulge instances, employ their vector embeddings on the Xeon and then connect it to a model that’s working on an accelerator,” he shares. “This is a large divulge case where we can ask to watch a vary of lawful compatibility, performance and price Xeon and Gaudi can converse.”

The bulletins about Xeon, Gaudi, and even Intel’s Lunar Lake, attain because the chipmaker responds to opponents swiping at what they witness because the company’s waning market dominance. At Computex, Intel Chief Executive Pat Gelsinger took exception to 1 comment from Nvidia CEO Jensen Huang, who has claimed Moore’s Law is mindless. Gelsinger then showed off the company’s latest chip for the AI PC.

“Intel is one in all the accurate firms on this planet innovating all the device during the beefy spectrum of the AI market opportunity—from semiconductor manufacturing to PC, community, edge and records center methods,” Gelsinger, Intel’s chief executive, remarks in an announcement. “Our latest Xeon, Gaudi and Core Extremely platforms, combined with the capacity of our hardware and instrument ecosystem, is delivering the versatile, staunch, sustainable and price-effective solutions our prospects need to maximize the extraordinary opportunities forward.”