Intel Xeon 654 vs Intel Xeon 6731P Comparison

Both the Intel Xeon 654 and Intel Xeon 6731P are active, Granite Rapids-based server/workstation processors sharing the Intel Socket 4710, yet benchmark data reveals two distinctly different performance profiles. The Xeon 654, with 18 cores and 36 threads, dominates the majority of head-to-head tests, winning 13 of 17 comparisons, while the Xeon 6731P, with 32 cores and 64 threads, counters with decisive victories in specialized workloads. These results show that core count alone does not dictate overall performance, as the Xeon 654's higher clock speeds consistently translate into superior scores across rendering and general-purpose computing.

Where Each One Wins

The performance split between these two processors is stark and workload-dependent. The Intel Xeon 654 claims victory in the vast majority of tested scenarios, establishing itself as the stronger all-rounder for mainstream server and workstation tasks. Its winning bench areas include every Cinebench iteration—R15, R20, and R23—in both single-core and multi-core modes, alongside PassMark's multithread, integer math, floating-point math, data compression, data encryption, and single-thread tests. This sweep indicates the Xeon 654 excels at rendering, compilation, and typical enterprise application workloads where clock speed per core drives responsiveness.

The Intel Xeon 6731P, despite losing the overall benchmark count, wins in four specific areas: PassMark extended instructions, find prime numbers, physics, and random string sorting. These wins reveal a processor tuned for highly parallel, instruction-heavy tasks that scale with core count and raw thread throughput. The physics score of 7105 against the Xeon 654's 5596 is particularly notable, suggesting the 6731P handles simulation and physics calculations more efficiently. Similarly, the find prime numbers score of 541 versus 390 shows an advantage in integer-heavy algorithmic work, while random string sorting (88019 vs 82828) indicates strength in data manipulation tasks that benefit from larger cache hierarchies.

Architecturally, the 6731P's 32 cores and 144 MB of shared L3 cache provide a clear advantage in workloads that can saturate all threads, whereas the 654's higher base clock of 3.10 GHz and boost clock of 4.80 GHz give it the edge in latency-sensitive operations. The data suggests a user facing mixed workloads—web serving, database queries, and content creation—would favor the 654, while a user running dedicated simulation or data-sorting pipelines might prefer the 6731P.

Architecture Differences

Both processors share the Granite Rapids architecture and are fabricated on Intel's 5 nm process node, but their physical and logical configurations diverge significantly. The Xeon 654 belongs to the Xeon 600 (Granite Rapids-WS) generation, featuring a dual-die design with a combined die size of 2x 598 mm². Its cache hierarchy includes 112 KB of L1 per core, 2 MB of L2 per core, and a shared 72 MB L3 cache. The Xeon 6731P, from the Xeon 6 (Granite Rapids-SP) generation, uses a single 598 mm² die and doubles the shared L3 cache to 144 MB, while maintaining the same per-core L1 and L2 allocations.

Clock speeds represent the most critical architectural divergence. The Xeon 654 operates at a 3.10 GHz base clock and boosts to 4.80 GHz, while the Xeon 6731P runs at a 2.50 GHz base and 4.10 GHz boost. This 0.6 GHz base-clock gap and 0.7 GHz boost-clock gap directly explain the 654's single-core dominance, as evidenced by its 79.3% lead in PassMark single-thread scores. The 6731P compensates with 32 cores versus 18, adding 28 threads to its total, and its larger L3 cache provides more on-die data capacity for multi-threaded workloads.

Memory support is identical on both parts: DDR5 with an eight-channel memory bus delivering 409.6 GB/s of bandwidth, and ECC memory support included. PCIe connectivity differs slightly, with the Xeon 654 offering Gen 5 across 128 lanes (CPU only), while the Xeon 6731P provides Gen 5 across 136 lanes. Both lack integrated graphics and target the server/workstation market segment. The Xeon 654 has an unlocked multiplier, whereas the 6731P is locked, a feature relevant to overclocking scenarios. The 6731P carries a higher TDP of 245 watts compared to the 654's 200 watts, reflecting its additional cores and cache.

Head-to-Head Benchmarks

The Cinebench suite delivers a consistent story: the Xeon 654 outperforms the 6731P by exactly 16.2% in every multi-core and single-core test. In Cinebench R23 multi-core, the 654 scores 52150 against 44871, and in R23 single-core it posts 7362 versus 6334. The identical deltaPct across all six Cinebench tests—16.2% for R15, R20, and R23 in both modes—indicates a uniform clock-speed advantage that scales linearly across rendering workloads, independent of the 6731P's additional 14 cores.

PassMark results reveal a more nuanced picture. The largest single-core gap appears in PassMark single-thread, where the 654 scores 3778 against 2107, a 79.3% advantage. This dramatic difference stems from the 654's 4.80 GHz boost clock and the 6731P's lower 4.10 GHz ceiling, combined with the 6731P's larger core count potentially impacting per-core efficiency. In multi-threaded PassMark, the 654 again wins with 61353 versus 52790, a 16.2% margin matching the Cinebench pattern. Integer math (207745 vs 198761, +4.5%) and floating-point math (163093 vs 157330, +3.7%) both favor the 654, though by smaller margins.

The 6731P's wins are concentrated in specific PassMark subtests. Extended instructions sees the 6731P ahead at 65656 versus 63539, a 3.2% edge. Find prime numbers produces the largest 6731P victory at 541 versus 390, a 27.9% lead. Physics shows a 21.2% advantage (7105 vs 5596), and random string sorting yields a 5.9% win (88019 vs 82828). These results indicate the 6731P's 144 MB L3 cache and higher core count excel at operations with high data reuse or parallelizable algorithms, even if its lower clock speeds hurt in latency-bound tasks. Data compression (818902 vs 799474, +2.4%) and data encryption (40675 vs 40087, +1.5%) narrowly favor the 654.

FAQ

Q: Which processor has a higher clock speed?

A: The Intel Xeon 654 runs at 3.10 GHz base and 4.80 GHz boost, while the Intel Xeon 6731P operates at 2.50 GHz base and 4.10 GHz boost.

Q: How many cores does each processor have?

A: The Xeon 654 has 18 cores and 36 threads, whereas the Xeon 6731P has 32 cores and 64 threads.

Q: What is the L3 cache size difference?

A: The Xeon 654 features 72 MB of shared L3 cache, while the Xeon 6731P doubles that with 144 MB of shared L3 cache.

Q: Which processor wins in Cinebench R23 multi-core?

A: The Xeon 654 scores 52150, beating the Xeon 6731P's 44871 by 16.2%.

Q: Are there any workloads where the Xeon 6731P beats the Xeon 654?

A: Yes, the 6731P wins in PassMark extended instructions, find prime numbers, physics, and random string sorting tests, with its largest lead being 27.9% in find prime numbers.

Q: What is the memory bandwidth for both processors?

A: Both support DDR5 with an eight-channel memory bus, providing identical 409.6 GB/s memory bandwidth.

The Verdict

Benchmark results indicate the Intel Xeon 654 is the superior choice for the majority of computing tasks, winning 13 of 17 head-to-head comparisons. Its 16.2% advantage across all Cinebench multi-core and single-core tests, combined with a 79.3% lead in PassMark single-thread performance, makes it the clear pick for rendering, application development, and general server workloads where per-core speed matters. The 654's 4.80 GHz boost clock and 3.10 GHz base clock, paired with 36 threads, deliver responsiveness that the 6731P's 64 threads cannot match in latency-sensitive operations. With an average benchmark score of 90717 versus 87756, the 654 sits in the 96th percentile of all CPUs, matching the 6731P's percentile but with a 3.2% higher average score.

The Intel Xeon 6731P is the appropriate selection for narrowly defined, highly parallel workloads. Its wins in physics (7105 vs 5596), find prime numbers (541 vs 390), and random string sorting (88019 vs 82828) demonstrate that its 32 cores and 144 MB L3 cache provide tangible benefits in simulation, algorithmic computation, and data manipulation. The 6731P's average benchmark score of 87756 places it roughly on par with the Intel Xeon 6736P (87864, -0.1%) and Intel Xeon w7-2575X (88172, -0.5%), indicating it competes effectively within its own performance tier. Users running dedicated physics simulations or data-sorting pipelines should consider the 6731P, but anyone needing a versatile processor for mixed workloads will find the Xeon 654 demonstrably faster across the broader benchmark suite.