Intel Xeon 6728P

The Intel Xeon 6728P is a 24-core, 48-thread server processor built on the Granite Rapids architecture, targeting the Server/Workstation market segment. With a base clock of 2.70 GHz and a boost clock of 4.10 GHz, this chip sits within the Xeon 6 family, fabricated on Intel’s 5 nm process node with a die size of 598 mm². The data shows a processor positioned for dense, multi-threaded workloads, though its benchmark percentile ranking of 50 indicates it lands exactly at the midpoint of all CPUs tracked in the database, suggesting a balanced rather than extreme performance profile.

Benchmark Performance

The benchmark data for the Intel Xeon 6728P is notable for what it lacks: the `benchmarks` array is empty, and the `avgBenchmarkScore` is 0. Consequently, the `nearestRivals` list is also empty, meaning there are no direct comparative scores or deltaPct values available for this entry. The only quantitative anchor is the `percentileVsAllCpus` of 50, which places this processor at the median of the entire CPU landscape in the database. This percentile is not a score itself but a relative ranking, implying that half of all tracked CPUs perform better and half perform worse in aggregated benchmark terms.

Without concrete benchmark scores, the performance picture must be inferred from architectural specifications. The combination of 24 cores and 48 threads, paired with a 144 MB shared L3 cache, suggests strong parallel throughput potential. The boost clock of 4.10 GHz is respectable for a server chip with this core count, indicating that single-threaded bursts can reach competitive levels. However, the absence of any benchmark results means that statements about exact performance deltas versus rivals cannot be made, the data simply does not support such claims. The percentile of 50 does indicate that this is not an outlier in either direction; it is a mainstream performer within the database’s population of CPUs.

Who Should Consider It

Given the empty benchmark array, recommendations must be derived from the specification sheet rather than measured results. The 24 cores and 48 threads make this processor a natural fit for heavily parallel server workloads, such as virtualization hosts running many concurrent VMs, database servers handling transaction-heavy loads, or scientific computing tasks that scale linearly with core count. The eight-channel DDR5 memory bus, with a theoretical bandwidth of 409.6 GB/s, further supports memory-bound applications like in-memory analytics or large-scale data processing.

For workstation users, the 2.70 GHz base clock and 4.10 GHz boost clock offer enough frequency headroom for interactive tasks like code compilation or 3D rendering, where both multi-threaded rendering and single-threaded UI responsiveness matter. Office productivity workloads, which typically rely on single-thread performance, would benefit from the 4.10 GHz boost, but the processor’s server orientation means it is likely overkill for such tasks. Gaming is not a primary use case for a 210 W TDP server chip, though the high boost clock could handle game servers or streaming workloads where multiple concurrent threads are required. The data suggests this processor is best suited for environments where core count and memory bandwidth take precedence over raw clock speed.

Single-Thread vs Multi-Thread Behavior

The split between single-thread and multi-thread capability is defined by the clock range: 2.70 GHz base versus 4.10 GHz boost. The 1.40 GHz gap between these two figures is substantial, indicating that the processor can aggressively ramp up frequency when fewer cores are active. This behavior is typical of server chips that must balance power draw across 24 cores but can still deliver snappy single-threaded responses when needed. For workloads that are primarily single-threaded, such as legacy database queries, certain scripting languages, or single-threaded physics simulations, the 4.10 GHz boost clock provides a meaningful performance ceiling.

In contrast, all-core workloads will likely settle near the 2.70 GHz base clock, given the 210 W TDP constraint. This means that multi-threaded scaling is not purely linear; doubling the core count from 12 to 24 does not necessarily double throughput if frequency drops proportionally. The 144 MB shared L3 cache mitigates some of this by reducing memory latency across cores, but the data implies that the Xeon 6728P is optimized for workloads that can tolerate lower all-core frequencies in exchange for higher core counts. The eight-channel memory bus helps feed those cores, but the actual frequency behavior under full load is not specified in the fact pack.

How It Compares

Because the `nearestRivals` array is empty, there are no direct rival comparisons available from the fact pack. The processor’s percentile of 50 against all CPUs provides a general reference point, but it does not identify specific competing parts. In the absence of rival names, scores, or deltaPct values, any comparison would require outside knowledge, which is prohibited by the rules. The data simply does not include comparative figures, so this section can only note that no rival data is present. The architecture, Granite Rapids, is Intel’s latest server design, but without benchmark scores, its standing relative to prior generations or competing architectures cannot be quantified. The benchmark database entry is incomplete on this front, leaving a gap in the analytical record.

Power and Thermals

The TDP for the Intel Xeon 6728P is 210 W, which classifies it as a high-power server processor. This thermal design power dictates the cooling tier required: a capable air cooler with a large heatsink and high-static-pressure fans would be the minimum, while a robust liquid cooling solution may be preferable in dense server racks where ambient temperatures are high. The 598 mm² die size, produced on Intel’s 5 nm process, indicates a large chip that spreads heat across a wide area, which can aid in thermal dissipation but also requires efficient contact with the cooling solution.

The 210 W TDP is not extreme for a 24-core server part, but it is significant enough that passive cooling is unlikely to suffice. Server chassis with high airflow, such as 1U or 2U systems with redundant fan walls, would be appropriate. The lack of an integrated GPU means all thermal headroom is dedicated to the CPU cores, simplifying cooling design. The boost clock of 4.10 GHz will generate peak heat, so the cooling solution must handle sustained all-core loads at the base clock and burst loads at boost without throttling. The fact pack does not specify a power limit beyond TDP, but the eight-channel memory controller and 88 PCIe Gen 5 lanes also draw power, meaning the total system power budget will exceed the 210 W CPU figure.

FAQ

Q: What is the socket type for the Intel Xeon 6728P?

A: The processor uses Intel Socket 4710, which is specific to the Granite Rapids-SP generation.

Q: How much L3 cache does the Xeon 6728P have?

A: It has 144 MB of shared L3 cache, alongside 112 KB of L1 cache per core and 2 MB of L2 cache per core.

Q: What memory type and bus width does it support?

A: The processor supports DDR5 memory with an eight-channel memory bus, providing a theoretical memory bandwidth of 409.6 GB/s. It also supports ECC memory.

Q: What is the launch MSRP of the Xeon 6728P?

A: The launch MSRP is $2478.

Q: Does the Xeon 6728P have an integrated GPU?

A: The fact pack lists `integratedGraphics` as null, indicating no integrated GPU is present; a discrete GPU would be required for display output.

Q: What is the production status of this processor?

A: The production status is listed as "Active," and the release date is 2025-02-23.

Platform and Compatibility

The Intel Xeon 6728P is built for the Intel Socket 4710, which is exclusive to the Granite Rapids-SP generation. This socket is not backward compatible with older Xeon platforms, so a new motherboard is required for any system upgrade. Memory support is limited to DDR5, with an eight-channel configuration that delivers a theoretical bandwidth of 409.6 GB/s. The memory controller supports ECC memory, which is essential for server reliability in data-critical workloads. The platform does not support DDR4, meaning existing DDR4 server RAM cannot be reused.

For expansion, the processor provides 88 PCIe Gen 5 lanes from the CPU itself, enabling high-bandwidth connectivity for GPUs, NVMe storage, and network adapters. This lane count is generous for a 24-core chip, supporting multiple accelerators or high-speed I/O devices simultaneously. The upgrade path within this socket is limited to other Granite Rapids-SP parts, but the fact pack does not list specific sibling models. The lack of an integrated GPU means a discrete graphics card is necessary for any display output, though this is standard for server processors. The platform is designed for dual-socket configurations, though the fact pack does not explicitly confirm this; the 88 PCIe lanes and eight-channel memory bus suggest a high-end server motherboard with extensive connectivity options. The 210 W TDP requires a motherboard with robust power delivery, typically found in workstation or enterprise server boards, and the Socket 4710 form factor dictates a larger chassis to accommodate the cooling solution.