Intel Xeon 6520P vs Intel Xeon 654 Comparison

Both the Intel Xeon 6520P and the Intel Xeon 654 are current-generation Granite Rapids server processors, but the benchmark data reveals they are tuned for distinctly different workloads. The 6520P, with its higher core count and larger shared cache, dominates the majority of compute-intensive tasks, winning 14 out of 17 head-to-head comparisons. The 654, however, is no slouch, leveraging its higher clock speeds to win the single-threaded performance crown and edge out a victory in floating-point math, making it a compelling option for latency-sensitive and lightly-threaded applications. The data suggests a clear split: the 6520P is the throughput champion, while the 654 is the responsiveness specialist.

Where Each One Wins

The performance split between these two processors is stark and consistent. The Intel Xeon 6520P is the clear winner for multi-threaded and parallel workloads. Its victories span across Cinebench R15, R20, and R23 multi-core tests, with a consistent 2.6% lead in each. This advantage extends to PassMark's integer math (3.1% ahead), data compression (2.8% ahead), and multithreaded tests (2.6% ahead). The most dramatic wins for the 6520P come in integer-heavy and physics-based tasks, where it outperforms the 654 by 34.9% in finding prime numbers and by 28.8% in physics calculations. Its 11.1% lead in data encryption and 15.6% lead in random string sorting further cement its position as the go-to chip for database, scientific, and security workloads that demand raw parallel processing power.

The Intel Xeon 654's wins are fewer but highly significant in their own right. It takes the single-threaded PassMark test with a score of 3778, a substantial 11.2% advantage over the 6520P's 3356. This is a massive swing, considering the 6520P wins all three Cinebench single-core tests by 2.6%. The 654 also narrowly wins the floating-point math test, scoring 163093 against 162862, a marginal 0.1% difference. This suggests that while the 6520P is better for general single-core tasks as measured by Cinebench, the 654's architecture and higher boost clock give it a specific edge in certain single-threaded and floating-point operations. For workloads that are not easily parallelized, the 654's responsiveness could be a deciding factor.

Architecture Differences

While both chips share the Granite Rapids architecture and are built on Intel's 5 nm process node, they diverge significantly in their physical design and feature sets. The most apparent difference is the core count: the Xeon 6520P packs 24 cores and 48 threads, while the Xeon 654 offers 18 cores and 36 threads. This 6-core advantage is the primary driver of the 6520P's multi-threaded dominance. The processors also have different die configurations. The 6520P is a single die with a size of 598 mm², whereas the 654 is a dual-die design, featuring 2x 598 mm². This dual-die approach may contribute to its higher clock speeds but also explains its smaller shared L3 cache.

Cache allocation is another critical differentiator. The 6520P features a massive 144 MB of shared L3 cache, double the 72 MB found on the 654. This larger cache pool is likely a key factor in its superior performance in data-heavy tasks like compression and encryption. Both processors have identical L1 and L2 cache per core (112 KB and 2 MB respectively), suggesting the per-core architecture is similar, but the larger shared pool on the 6520P allows for more data to be held closer to the cores. The 654 counters with a significantly higher boost clock of 4.80 GHz compared to the 6520P's 4.00 GHz, and a higher base clock of 3.10 GHz versus 2.40 GHz, which explains its strength in single-threaded tasks.

FAQ

Q: Which processor has more cores, and how does that affect performance?

A: The Intel Xeon 6520P has significantly more cores, offering 24 cores and 48 threads compared to the Intel Xeon 654's 18 cores and 36 threads. This 6-core advantage is directly reflected in the benchmark results, where the 6520P consistently leads in multi-core tests like Cinebench R23 (53495 vs 52150) and PassMark's multithreaded test (62936 vs 61353).

Q: The 654 has a higher clock speed. Why doesn't it win more benchmarks?

A: The Intel Xeon 654 does have a higher boost clock (4.80 GHz vs 4.00 GHz) and base clock (3.10 GHz vs 2.40 GHz). This translates to a significant 11.2% lead in the PassMark single-thread test. However, for most multi-threaded applications, the 6520P's six additional cores and double the L3 cache (144 MB vs 72 MB) outweigh the clock speed advantage, allowing it to win the majority of benchmarks.

Q: Are both processors based on the same architecture?

A: Yes, both the Intel Xeon 6520P and the Intel Xeon 654 are built on the Granite Rapids architecture and manufactured on Intel's 5 nm process node. They share the same socket (Intel Socket 4710) and support DDR5 memory over an eight-channel bus with identical 409.6 GB/s bandwidth.

Q: How do they compare in terms of memory and PCIe support?

A: Both processors are identical in memory support, offering DDR5 across an eight-channel bus with a bandwidth of 409.6 GB/s and ECC support. However, they differ in PCIe lanes, with the 6520P providing Gen 5 with 88 lanes, while the 654 offers significantly more with Gen 5 and 128 lanes.

Q: Which processor is better for cryptographic workloads?

A: The Intel Xeon 6520P is demonstrably better for cryptographic tasks. In the PassMark data encryption test, it scores 45188 compared to the 654's 40675, representing a substantial 11.1% performance advantage. This is likely due to its higher core count and larger cache.

Q: Is there a difference in their production status or market segment?

A: No, both processors are currently Active in production and are both targeted at the Server/Workstation market segment. They also share the same release date of 2025-02-23 for the 6520P and 2026-02-01 for the 654.

Specification Differences

The core architectural differences between the two processors are detailed in their specifications. The most significant divergence is in core and thread counts, where the 6520P holds a 24-core/48-thread configuration against the 654's 18-core/36-thread setup. This is complemented by a substantial difference in shared L3 cache, with the 6520P featuring 144 MB and the 654 featuring only 72 MB. Clock speeds are also a major point of differentiation, with the 654 having a higher base clock (3.10 GHz vs 2.40 GHz) and boost clock (4.80 GHz vs 4.00 GHz).

The physical design also differs, with the 6520P using a single 598 mm² die, while the 654 uses a dual-die design with 2x 598 mm². This may explain the 654's higher clock speeds but also its smaller cache. There are also differences in their feature sets: the 6520P has a launch MSRP of $1295 and a part number of SRVNQ, while the 654 has a launch MSRP of $1199 and part number SA2DP. Furthermore, the 654 has an unlocked multiplier, while the 6520P does not. Finally, the 654 offers more PCIe lanes (128) compared to the 6520P's 88 lanes.

Head-to-Head Benchmarks

The head-to-head data paints a clear picture of two distinct performance profiles. The 6520P's most impressive victories are in integer-heavy and physics-based workloads. In the PassMark find prime numbers test, it scores 526 against the 654's 390, a 34.9% advantage. Similarly, in PassMark physics, the 6520P scores 7209 versus 5596, a 28.8% lead. These are not marginal wins; they indicate a fundamental advantage in processing parallel integer operations. The 6520P also shows a strong 15.6% lead in random string sorting (95736 vs 82828) and an 11.1% lead in data encryption (45188 vs 40675). In the more general Cinebench tests, the 6520P maintains a consistent 2.6% lead across all R15, R20, and R23 multi-core and single-core benchmarks.

The 654's victories, while fewer, are equally telling. Its most significant win is in the PassMark single-thread test, where it scores 3778 against the 6520P's 3356, an 11.2% delta. This is the largest margin of victory in any benchmark, highlighting the 654's strength in lightly-threaded tasks. It also manages a narrow win in PassMark floating-point math, scoring 163093 against 162862, a 0.1% difference. This suggests that while the 6520P is the superior all-around processor, the 654's higher clock speeds give it a decisive edge in scenarios where single-thread performance is the bottleneck.

The Verdict

The data dictates a clear choice based on workload requirements. For users running heavily parallel, multi-threaded applications, the Intel Xeon 6520P is the superior processor. Its 24 cores, 48 threads, and 144 MB of L3 cache provide a commanding lead in Cinebench multi-core tests, integer math, data compression, encryption, and physics simulations. Its 34.9% and 28.8% wins in prime number finding and physics, respectively, make it the obvious choice for scientific computing, database processing, and any workload that can scale across many cores. The 6520P's overall average benchmark score of 93786, compared to the 654's 90717, further solidifies its position as the higher-performing chip for throughput-intensive tasks.

Conversely, the Intel Xeon 654 is the better option for workloads that prioritize single-threaded performance and lower latency. Its 11.2% lead in the PassMark single-thread test and its win in floating-point math make it ideal for applications like real-time signal processing, certain types of simulation, or legacy software that doesn't scale well across cores. Its higher base and boost clocks (3.10 GHz and 4.80 GHz) are the key enablers of this responsiveness. While it trails in multi-core benchmarks, its 2.6% deficit in Cinebench R23 is manageable, and its 128 PCIe lanes provide more I/O flexibility. Ultimately, the choice is between the 6520P's raw parallel power and the 654's exceptional single-core agility.