The Wait Is Finally Over! Intel Sapphire Rapids “4th Gen Xeon” CPU Family Is Here But Was It Worth The Wait?
Originally planned to launch more than a year ago, Intel’s Sapphire Rapids Xeon CPUs have slipped time after time, facing continued production woes and revisions that were highlighted throughout its production, however, the wait is finally over and Intel finally can call it a day as it launches their full-fledged Xeon chip lineup. With today’s launch, Intel is boasting some big numbers, utilizing the various built-in accelerators of any CPU on the planet. Intel states that its 4th Gen Xeon Scalable processors codenamed Sapphire Rapids offer (vs 3rd Gen Xeon CPUs):
+53% Average Performance Gain in General Purpose Computing Up To 10x High inference & training performance in Artificial Intelligence Up To 2x capacity for vRAN workloads at the same power envelope (Network 5G vRAN) Up To 2x higher data compression with 95% fewer cores (Networking & Storage) Up To 3x higher performance in Data Analytics Up To 3.7x performance on memory-bound workloads in HPC (Xeon Max CPU Series)
Following is a list of all the Built-In Accelerators and Security Engines you would get on the Intel 4th Gen Xeon Sapphire Rapids CPUs:
Intel Advanced Matrix Extensions (AMX) Intel QuickAssist Technology (QAT) Intel Data Streaming Accelerator (DSA) Intel Dynamic Load Balancer (DLB) Intel In-Memory Analytics Accelerators (DLB) Intel Advanced Vector Extensions 512 (AVX-512) Intel Advanced Vector Extensions 512 (AVX-512) for vRAN Intel Crypto Acceleration Intel Speed Select Technology (SST) Intel Data Direct I/O Technology (DDIO) Intel Software Guard Extensions (SGX) Intel TrustDomain Extension (TDX) Intel Control-Flow Enforcement Technology (CET)
AMD EPYC Genoa vs Intel Xeon Sapphire Rapids-SP Server CPU Platforms
Those are certainly some big numbers but we have to see how Intel is achieving these figures and to do that, we have to talk specs. For Sapphire Rapids-SP, Intel is using a quad multi-tile chiplet design which will come in HBM and non-HBM flavors. The HBM flavors will be branded under the new Xeon Max CPU series while the Non-HBM chips will retain the standard Xeon branding. While each tile is its own unit, the chip itself acts as one singular SOC and each thread has full access to all resources on all tiles, consistently providing low-latency & high cross-section bandwidth across the entire SOC. We have already taken an in-depth look at the P-Core over here but some of the key changes that will be offered to the data center platform will include AMX, AiA, FP16, and CLDEMOTE capabilities. The Accelerator Engines will increase the effectiveness of each core by offloading common-mode tasks to these dedicated accelerator engines which will increase performance & decrease the time taken to achieve the necessary task. In terms of I/O advancements, Sapphire Rapids-SP Xeon CPUs will introduce CXL 1.1 for accelerator and memory expansion in the data center segment. There’s also an improved multi-socket scaling via Intel UPI, delivering up to 4 x24 UPI links at 16 GT/s and a new 8S-4UPI performance-optimized topology. The new tile architecture design also boosts the cache beyond 100 MB along with Optane Persistent Memory 300 series support. The lineup will also come in HBM flavors which will be using a different packaging design:
Intel Sapphire Rapids-SP Xeon (Standard Package) - 4446mm2 Intel Sapphire Rapids-SP Xeon (HBM2E Package) - 5700mm2 AMD EPYC Genoa (12 CCD Package) - 5428mm2
Further detailing the performance and efficiency figures, Intel claims that 4th Gen Xeon customers can expect a 2.9x average performance per watt efficiency improvement with up to 70 Watts worth of power savings per CPU in optimized power mode with minimal performance loss and 55% lower TCO and power consumption.
Intel Sapphire Rapids 4th Gen Xeon CPU Platform
The Intel 4th Gen Xeon CPU family is made up of nearly 50 SKUs, offering a wide range of options for customers. The Sapphire Rapids lineup will make use of 8-channel DDR5 memory with speeds of up to 4800 Mbps & support PCIe Gen 5.0 on the Eagle Stream platform (C740 chipset). The expansiveness of built-in accelerators included in 4th Gen Xeon means Intel delivers platform-level power savings, lessening the need for additional discrete acceleration and helping our customers achieve their sustainability goals. Additionally, the new Optimized Power Mode can deliver up to 20% socket power savings with a less than 5% performance impact for selected workloads11. New innovations in air and liquid cooling reduce total data center energy consumption further; and the manufacturing of 4th Gen Xeon, it’s been built with 90% or more renewable electricity at Intel sites with state-of-the-art water reclamation facilities. Artificial Intelligence In AI, and compared to the previous generation, 4th Gen Xeon processors achieve up to 10x5,6 higher PyTorch real-time inference and training performance with built-in Intel® Advanced Matrix Extension (Intel® AMX) accelerators. Intel’s 4th Gen Xeon unlocks new levels of performance for inference and training across a wide breadth of AI workloads. The Xeon CPU Max Series expands on these capabilities for natural language processing, with customers seeing up to a 20x12 speed-up on large language models. With the delivery of Intel’s AI software suite, developers can use their AI tool of choice, while increasing productivity and speeding time to AI development. The suite is portable from the workstation, enabling it to scale out in the cloud and all the way out to the edge. And it has been validated with over 400 machine learning and deep learning AI models across the most common AI use cases in every business segment. Networking 4th Gen Xeon offers a family of processors specifically optimized for high-performance, low-latency networks, and edge workloads. These processors are a critical part of the foundation driving a more software-defined future for industries ranging from telecommunications and retail to manufacturing and smart cities. For 5G core workloads, built-in accelerators help increase throughput and decrease latency, while advances in power management enhance both the responsiveness and the efficiency of the platform. And, when compared to previous generations, the 4th Gen Xeon delivers up to twice the virtualized radio access network (vRAN) capacity without increasing power consumption. This enables communications service providers to double the performance per watt to meet their critical performance, scaling, and energy efficiency needs. The Eagle Stream platform will also introduce the LGA 4677 socket which will be replacing the LGA 4189 socket for Intel’s upcoming Cedar Island & Whitley platform which would house Cooper Lake-SP and Ice Lake-SP processors, respectively. The Intel Sapphire Rapids-SP Xeon CPUs will also come with CXL 1.1 interconnect that will mark a huge milestone for the blue team in the server segment. Coming to the configurations, the top part is started to feature 60 cores with a TDP of 350W. What is interesting about this configuration is that it is listed as a low-bin split variant which means that it will be using a tile or MCM design. The Sapphire Rapids-SP Xeon CPU will be composed of a 4-tile layout with each tile featuring 14 cores. The Intel Sapphire Rapids-SP Xeon CPUs will come in six tiers:
Xeon Max (9400 Series) Xeon Platinum (8000 Series) Xeon Gold (6000 Series) Xeon Gold (5000 Series) Xeon Silver (4000 Series) Xeon Bronze (3000 Series)
The TDPs listed here are at PL1 rating so the PL2 rating, as seen earlier, is going to be very high in the 400W+ range and the BIOS limit is expected to hover at around 700W+. Furthermore, there are ten segments within the lineup itself that indicate the workload they are aimed at. These are listed as follows:
2S Performance General Purpose 2S Mainline General Purpose Liquid Cooled General Purpose (-Q) Single-Socket General Purpose (-U) Long-Life Use IOT General Purpose (-T) IMDB/Analytics/Virtualization Optimized (-H) 5G / Networking Optimized (-N) Cloud Optimized IaaS (-P) SaaS Optimized (-V) Media Optimized (-M) Storage & Hyperconverged Infrastructure (HCI) Optimized (-S) HPC Optimized (Xeon CPU Max Series)
The flagship of the lineup is the Intel Xeon Platinum 8490H which offers 60 Golden Cove cores, 120 threads, 112.5 MB of L3 cache, a single-core boost of up to 3.5 GHz and a 2.9 GHz all-core boost, & a base TDP figure of 350W. Intel is charging a very premium $17,000 US which is over $5000 more expensive than the top Genoa chip that packs 96 cores, 192 threads, 384 MB of L3 cache, and more I/O capabilities. It is likely that Intel is charging more for the expanded scalability that this chip will offer with up to 8S support. This means that customers can go with up to 8 8480H CPUs on a single server, allowing for up to 480 cores, 960 threads, and a pool of 900 MB of cache. That will also have an insanely high price of $136,000 US for the 8 Xeon Platinum chips alone.
Intel 4th Gen Sapphire Rapids Xeon CPUs (2S Performance General Purpose)
Intel 4th Gen Sapphire Rapids Xeon CPUs (2S Mainline General Purpose)
Intel 4th Gen Sapphire Rapids Xeon CPUs (Liquid-Cooled/ Single-Socket/ IOT)
Intel 4th Gen Sapphire Rapids Xeon CPUs (IMDB/Analytics/Virtualization Optimized)
Intel 4th Gen Sapphire Rapids Xeon CPUs (Networking / 5G)
Intel 4th Gen Sapphire Rapids Xeon CPUs (Cloud / HCI Optimized)
Intel 4th Gen Sapphire Rapids Xeon Max Series CPUs
It looks like AMD will still hold the upper hand in the number of cores & threads offered per CPU with their Genoa chips pushing for up to 96 cores and Bergamo going up to 128 cores whereas Intel Xeon chips would max out at 60 cores if they don’t plan on making SKUs with a higher number of tiles. Intel will have a wider and more expandable platform that can support up to 8 CPUs at once so unless Genoa offers more than 2P (dual-socket) configurations, Intel will have the lead in the most number of cores per rack with an 8S rack packing up to 480 cores and 960 threads. But we have to wait for the reviews on this lineup as AMD may hold a major lead in the efficiency and performance department with its EPYC family.