Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Intel’s ISC 2022 announcement was a three-part attempt to compete across high-performance computing: add high-bandwidth memory (HBM) to Xeon CPUs, build a faster successor to its Ponte Vecchio GPU, and eventually combine CPU and GPU chiplets in an “XPU.” The strategy was real, but the outcomes diverged. Sapphire Rapids HBM reached customers as the Intel Xeon CPU Max Series; Rialto Bridge was later discontinued; and Falcon Shores remained a roadmap architecture rather than a proven shipping product in the available Intel material.
What Intel announced at ISC 2022
The May 31, 2022 announcement covered three different solutions to three different HPC bottlenecks. Some scientific and engineering programs are limited by memory bandwidth, some by parallel floating-point throughput, and others by the complexity of moving data between CPUs, accelerators and memory. Intel’s answer was a portfolio rather than one universal processor.
- Sapphire Rapids HBM: a Xeon CPU with on-package high-bandwidth memory for bandwidth-sensitive workloads.
- Rialto Bridge: a planned successor to the Ponte Vecchio data-center GPU.
- Falcon Shores: a future chiplet-based XPU combining Xeon CPU tiles and Xe GPU tiles.
The original coverage was a roadmap story, not an independent benchmark review. Intel supplied the performance comparisons and future projections, so those claims must be separated from later product results.
Free tools Windows power users keep installed
One-click scans. No signup required.
Read the original ISC 2022 announcement context.
Why HBM matters on a CPU
Conventional server memory offers substantial capacity, but many HPC applications repeatedly stream large arrays and are limited by how quickly data can reach the cores. HBM uses a very wide, short connection to memory stacked in or alongside the processor package, providing much higher bandwidth than DDR5 alone.
#1 Best Overall
- MODEL P86771-005: Ultra-compact HPE ProLiant MicroServer Gen11 featuring Intel Xeon 6325P 3.5GHz 4-core processor, ideal for SMB workloads and edge deployments
- FLEXIBLE MEMORY & STORAGE: Includes 32GB DDR5 UDIMM memory (expandable to 128GB) and 4 LFF-NHP drive bays. Features new MR408i-p controller support for enhanced storage performance
- READY TO RUN: Includes 1 x HPE 4TB SATA 6G Business Critical HDD, 180W external power adapter, and 1/1/1 year warranty for dependable plug-and-play server operation
- WHISPER-QUIET & SPACE-SAVING: Ultra-compact mini tower design fits easily in small office spaces; supports wall, flat, or vertical placement for deployment flexibility
- REMOTE MANAGEMENT READY: Includes HPE iLO6 with Silicon Root of Trust, TPM 2.0, and dedicated iLO-M.2 port kit for secure and efficient remote server administration
That does not make every CPU operation faster. A compute-bound, cache-friendly or poorly vectorized program may see little benefit. HBM is most valuable when profiling shows that memory bandwidth—not arithmetic throughput—is the limiting factor. Capacity also remains a separate issue: a package-level HBM pool does not replace the larger external memory available in a complete server.
Sapphire Rapids HBM became Xeon CPU Max
Intel’s fourth-generation Xeon Scalable architecture, Sapphire Rapids, launched with an HBM-equipped branch later branded Intel Xeon CPU Max Series. Intel lists five Max CPU models, all launched in Q1 2023, with 32 to 56 performance cores, 64 GB of in-package HBM and a 350 W TDP.
| Model | Cores | Base / max turbo | Cache | HBM |
|---|---|---|---|---|
| Xeon Max 9462 | 32 | 2.70 / 3.50 GHz | 75 MB | 64 GB |
| Xeon Max 9460 | 40 | 2.20 / 3.50 GHz | 97.5 MB | 64 GB |
| Xeon Max 9468 | 48 | 2.10 / 3.50 GHz | 105 MB | 64 GB |
| Xeon Max 9470 | 52 | 2.00 / 3.50 GHz | 105 MB | 64 GB |
| Xeon Max 9480 | 56 | 1.90 / 3.50 GHz | 112.5 MB | 64 GB |
The package uses four compute tiles linked with EMIB and supports PCIe 5.0 and CXL 1.1. Intel describes HBM operation as flexible for different application needs, including arrangements that use HBM as cache, as directly addressable memory, or alongside conventional memory; system documentation and software configuration determine the practical mode.
Recommended Free Tools
Intel’s 2022 material compared HBM Sapphire Rapids with third-generation Xeon Scalable processors at roughly 2× to 3× performance, depending on workload. Intel’s later Max Series page advertises up to 4.8× better performance than competing processors on selected HPC and AI workloads. Neither figure means a universal CPU speed increase. The result depends on the application, compiler, libraries, data set, thread count, memory mode and comparison system. Treat these as Intel-supplied results, not neutral laboratory benchmarks.
Rank #2
- Windows server license is not included
Intel Xeon CPU Max overview · Intel ARK model specifications
Ponte Vecchio, Aurora and the GPU push
Ponte Vecchio was Intel’s first major Xe-HPC data-center GPU and a foundation of the Aurora supercomputer at Argonne National Laboratory. Aurora was designed around Intel Xeon CPUs and Ponte Vecchio accelerators as Intel competed with AMD-powered Frontier and NVIDIA-based systems.
In 2022 Intel expected Aurora eventually to exceed Frontier. That was a forward-looking expectation, not a measured result available at the time of the announcement, and should not be presented as a historical benchmark fact without a dated source.
Rialto Bridge: ambitious successor, canceled product
Intel described Rialto Bridge as Ponte Vecchio’s successor with:
Rank #3
- Intel Xeon E5-2699 V4 Docosa-core (22 Core) 2.20 Ghz Processor - Socket Lga 2011-v3 - 5.50 Mb - 55 Mb Cache - 64-bit Processing - 14 Nm - 145 W
- up to 160 Xe cores, versus up to 128 for Ponte Vecchio;
- faster memory signaling (Intel used the term “more GT/s”);
- greater I/O bandwidth; and
- support for the second Open Accelerator Module specification.
OAM v2’s specification allowed power delivery of up to 800 W. That is a capability of the module specification, not proof that a final Rialto board would continuously consume 800 W. Intel’s original article also speculated about which interconnect would account for the I/O increase; that detail was not confirmed in the supplied sources.
The crucial update is that Rialto Bridge never became a current product. Intel later said it was discontinuing the design because it would have delivered an incremental step over the existing architecture. The company instead described a two-year data-center-GPU cadence and positioned Falcon Shores as the next major Max-family GPU architecture.
Intel’s later roadmap update and Rialto Bridge decision
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Falcon Shores and the XPU idea
Falcon Shores was presented as a configurable combination of Xeon CPU tiles and Xe GPU tiles. The proposed XPU could balance scalar CPU work and massively parallel GPU work while targeting high memory bandwidth and capacity. The chiplet approach was intended to let system designers choose a more CPU-heavy or GPU-heavy composition instead of treating those as entirely separate products.
Rank #4
- Dell T7810 Precision Tower Workstation
- 2x Intel Xeon E5-2690 v4 14-Core/28 Threads 3.1GHz (3.5GHz Turbo)
- 128GB Memory DDR4 – Nvidia Quadro K620 2GB
- Add your own Hard Drives/ SSDs
- Add your own Operating System
Intel’s 2022 presentation projected improvements of more than 5× in areas including performance per watt, compute density, memory bandwidth and memory capacity versus then-current parts. Those were roadmap projections, not measured Falcon Shores results. Intel’s retrieved roadmap material targeted introduction in 2025, but the available sources do not establish a final shipping specification or commercial availability as of August 16, 2026.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What shipped, and what did not?
| 2022 name or promise | Later status |
|---|---|
| Sapphire Rapids HBM | Shipped as Intel Xeon CPU Max Series in Q1 2023. |
| Ponte Vecchio | Commercialized as Intel Data Center GPU Max. |
| Rialto Bridge | Discontinued before becoming a shipping successor. |
| Falcon Shores | Roadmap CPU/GPU architecture; final product status is not established by the supplied material. |
Who should consider CPU Max?
CPU Max is most compelling when a real workload is bandwidth-bound, the code is already CPU-oriented, and avoiding a discrete accelerator simplifies deployment. Examples can include computational fluid dynamics, finite-element analysis, molecular dynamics, genomics and graph analytics—but those categories are not guarantees. Benchmark the actual application with its compiler, MPI implementation, libraries, data size and memory configuration.
A conventional Xeon may be better when capacity matters more than bandwidth, the workload is compute-bound, power or cooling is restricted, or the HBM premium does not improve total cost of ownership. A discrete GPU may be preferable for massively parallel code with mature CUDA or other accelerator libraries, provided the organization can support porting, scheduling and accelerator operations.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Architects should also evaluate NUMA and memory placement, software support through Intel oneAPI and MPI, interconnects, storage, cooling, OEM availability and lifecycle support. Raw core counts cannot be compared directly between CPU and GPU products, and 64 GB of HBM is not equivalent to 64 GB of total server memory.
Terminology guide
| 2022 codename | Later branding or status |
|---|---|
| Sapphire Rapids HBM | Intel Xeon CPU Max Series |
| Ponte Vecchio | Intel Data Center GPU Max Series |
| Rialto Bridge | Discontinued |
| Falcon Shores | Future/roadmap Max-family GPU/XPU architecture in the cited roadmap |
Bottom line for infrastructure decisions
Intel’s 2022 announcement contained a credible technical strategy: use HBM for bandwidth-starved CPU workloads, offer Xe GPUs for parallel acceleration and eventually combine both in a chiplet XPU. Execution was uneven. The HBM CPU concept reached the market as Xeon CPU Max, while the planned intermediate GPU successor was canceled, and Falcon Shores should be judged by documented products rather than its early projections. For a purchase decision, measure the full application stack and system economics—not a headline multiplier.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
