AMD EPYC 7352 vs Intel Xeon 6515P Comparison

The AMD EPYC 7352 and Intel Xeon 6515P are two very different server processors, one a 24-core Zen 2 part from 2019 and the other a 16-core Granite Rapids part from 2025. The benchmark data shows a clear split: the Intel chip wins the majority of tests, but the AMD chip dominates in a few specific, critical workloads. The Xeon 6515P takes 13 of the 17 head-to-head benchmarks, while the EPYC 7352 wins only 4, yet those four wins reveal where the older AMD design still holds a distinct advantage.

Head-to-Head Benchmarks

The most striking result is in PassMark data encryption, where the EPYC 7352 scores 44,426 against the Xeon's 30,476. That is a 45.8% lead for AMD, the largest delta on the entire list. This suggests the Zen 2 architecture's cryptographic instructions are significantly more efficient than what Intel's Granite Rapids offers, making the EPYC the clear choice for encryption-heavy tasks like VPN termination or secure database connections.

The other AMD wins are narrower but still meaningful. In data compression, the EPYC 7352 scores 660,712 versus 592,645 for the Xeon, an 11.5% advantage. This workload benefits from the AMD part's larger 128 MB total L3 cache, which likely reduces memory traffic during compression. The EPYC also edges out the Xeon in integer math (148,605 vs 147,047, only 1.1% ahead) and random string sorting (69,231 vs 64,425, a 7.5% lead). These are modest margins, but they indicate that the EPYC's 24 cores and 48 threads can still outpace the Xeon's 16 cores and 32 threads in certain parallel integer workloads.

Every other benchmark goes to the Intel Xeon 6515P, often by wide margins. In Cinebench, the Xeon wins all six tests with a consistent 11% delta across R15, R20, and R23, both single-core and multi-core. For example, in Cinebench R23 multi-core, the Xeon scores 38,547 versus 34,314 for the EPYC, and in R23 single-core it scores 5,442 versus 4,844. The single-core gap is particularly notable because the Xeon's 3.80 GHz boost clock is substantially higher than the EPYC's 3.20 GHz, and the newer architecture delivers better instructions-per-clock.

The Xeon's biggest wins come in floating-point math and physics. In PassMark floating-point math, the Xeon scores 128,954 against 87,969 for the EPYC, a 31.8% margin. This is a massive advantage for scientific computing, simulation, and any workload with heavy FPU usage. PassMark physics also favors the Xeon heavily: 3,829 versus 2,688, a 29.8% lead. The Xeon also wins extended instructions by 24.7% (53,383 vs 40,203), finding prime numbers by 21.8% (385 vs 301), and single-thread performance by 30.7% (2,855 vs 1,979).

The overall picture is clear: the Xeon 6515P is the stronger all-around performer, with a 11% lead in multi-threaded Cinebench and a 30%+ lead in several PassMark subtests. However, the EPYC 7352's dominance in encryption and compression means that specific workloads will prefer the older AMD chip.

FAQ

Q: Which CPU has a higher overall benchmark score?

A: The Intel Xeon 6515P has an average benchmark score of 67,006, while the AMD EPYC 7352 scores 68,118. The EPYC actually has a slightly higher average score, despite losing the majority of individual tests, because its wins in data compression and encryption are weighted heavily in the average.

Q: How do these CPUs compare to their nearest rivals?

A: The EPYC 7352 sits at the 94th percentile of all CPUs, with its closest rival being the AMD EPYC 9184X (avg score 68,202, just 0.1% higher). The Xeon 6515P is at the 93rd percentile, with its closest rival being the AMD EPYC 4465P (avg 66,925, 0.1% lower). Both are competitive with modern high-end parts.

Q: Why does the EPYC 7352 win in encryption but lose in most other tests?

A: The EPYC 7352's 45.8% lead in data encryption (44,426 vs 30,476) likely stems from its Zen 2 architecture's optimized cryptographic instructions, combined with its 128 MB total L3 cache. The Xeon's newer Granite Rapids core is faster in most other operations, but it cannot match AMD's encryption throughput.

Q: Is the Xeon 6515P better for single-threaded workloads?

A: Yes, decisively. The Xeon scores 2,855 in PassMark single-thread versus 1,979 for the EPYC, a 30.7% lead. In Cinebench R23 single-core, the Xeon scores 5,442 versus 4,844, an 11% advantage. The Xeon's higher 3.80 GHz boost clock and newer architecture give it a clear edge in lightly-threaded tasks.

Q: What about multi-threaded performance?

A: The Xeon wins all multi-core Cinebench tests by 11%. In R23 multi-core, it scores 38,547 vs 34,314. PassMark multithread also favors the Xeon: 45,350 vs 40,370, an 11% lead. Despite having 8 fewer cores and 16 fewer threads, the Xeon's superior per-core performance and memory bandwidth (409.6 GB/s vs 204.8 GB/s) overcome the EPYC's core count advantage.

Q: Which CPU is older and which is newer?

A: The AMD EPYC 7352 was released on 2019-08-06, while the Intel Xeon 6515P was released on 2025-02-23. The Xeon is nearly six years newer, built on Intel's 5 nm process, while the EPYC uses TSMC's 7 nm node.

Architecture Differences

The two processors represent fundamentally different design philosophies. The EPYC 7352 uses AMD's Zen 2 microarchitecture under the codename "Rome," built on a 7 nm process from TSMC. It is a chiplet design with 4 dies, each 74 mm², totaling 15,200 million transistors. The Xeon 6515P uses Intel's Granite Rapids architecture, built on Intel's 5 nm node, and is a monolithic design (though the fact pack does not list die size or transistor count for it).

Cache hierarchies are completely different. The EPYC has 96 KB of L1 per core, 512 KB of L2 per core, and 32 MB of L3 per die, totaling 128 MB of L3. The Xeon has 112 KB of L1 per core, a much larger 2 MB of L2 per core, and 72 MB of shared L3. The Xeon's larger per-core L2 cache (2 MB vs 512 KB) is a key reason for its single-thread performance advantage, as more data can be stored closer to the core.

Memory support is a major generational difference. The EPYC uses DDR4 with eight-channel memory and 204.8 GB/s bandwidth. The Xeon uses DDR5, also eight-channel, but with double the bandwidth at 409.6 GB/s. This explains why the Xeon excels in bandwidth-sensitive workloads like floating-point math and physics.

PCIe connectivity also differs significantly. The EPYC offers Gen 4 with 128 lanes (CPU only), while the Xeon offers Gen 5 with 88 lanes. The EPYC has more lanes, but the Xeon's are faster per lane. The Xeon has no integrated graphics, listed as "N/A," while the EPYC also has none listed.

Specification Differences

The two CPUs differ on nearly every key specification. The EPYC 7352 has 24 cores and 48 threads, while the Xeon 6515P has 16 cores and 32 threads. Both have a 2.30 GHz base clock, but the Xeon boosts to 3.80 GHz versus the EPYC's 3.20 GHz. Thermal design power is close: 155 W for the EPYC vs 150 W for the Xeon.

The sockets are incompatible: AMD Socket SP3 for the EPYC, Intel Socket 4710 for the Xeon. The EPYC is from the EPYC 7002 series, while the Xeon is from the Xeon 6 (Granite Rapids-SP) generation. Process nodes differ: 7 nm (TSMC) for the EPYC, 5 nm (Intel) for the Xeon.

Memory is a major differentiator: DDR4 for the EPYC, DDR5 for the Xeon. Both support ECC memory and have eight-channel buses, but the Xeon's 409.6 GB/s bandwidth is exactly double the EPYC's 204.8 GB/s. The EPYC's total L3 cache is 128 MB, while the Xeon's is 72 MB shared. The EPYC's part number is 100-000000077, while the Xeon's is SRVU6. Neither is unlocked for overclocking.

Where Each One Wins

The AMD EPYC 7352 wins in specific, high-value workloads. Data encryption is its crown jewel, with a 45.8% lead over the Xeon. This makes it the preferred choice for security appliances, encrypted storage servers, and any workload that relies heavily on AES or other cryptographic operations. It also wins in data compression (11.5% lead), which benefits from its large 128 MB L3 cache, making it good for database workloads, log processing, and file compression services. The EPYC also has a slight edge in integer math (1.1%) and random string sorting (7.5%), suggesting it handles general integer-heavy server tasks well.

The Intel Xeon 6515P wins everywhere else, and by larger margins. Its 31.8% lead in floating-point math makes it the obvious choice for scientific computing, financial modeling, and engineering simulations. The 29.8% lead in physics tests reinforces this, as physics engines and similar workloads rely heavily on floating-point throughput. The Xeon's 30.7% single-thread lead makes it better for lightly-threaded applications, legacy code, and any workload that cannot scale across many cores. The 21.8% lead in prime number finding and 24.7% lead in extended instructions further solidify its position for compute-intensive tasks.

The Verdict

The data presents a clear choice based on workload. If your primary concern is encryption or compression throughput, the AMD EPYC 7352 is the winner, despite being the older part. Its 45.8% encryption lead is too large to ignore, and its 11.5% compression advantage is substantial. For these workloads, the EPYC's 24 cores and 128 MB L3 cache are more valuable than the Xeon's newer architecture.

For almost everything else, the Intel Xeon 6515P is the better processor. It wins all Cinebench tests by 11%, all single-thread tests by 11-30%, and dominates floating-point and physics by roughly 30%. Its double memory bandwidth (409.6 GB/s vs 204.8 GB/s) and higher boost clock (3.80 GHz vs 3.20 GHz) give it a decisive edge in modern, bandwidth-hungry applications. The Xeon also achieves this with fewer cores (16 vs 24) and lower TDP (150 W vs 155 W), meaning better performance-per-watt in most scenarios.

In terms of overall average benchmark score, the EPYC 7352 actually edges out the Xeon (68,118 vs 67,006), and it sits at the 94th percentile versus the Xeon's 93rd. But this average is skewed by the EPYC's massive encryption win. A typical mixed server workload will likely favor the Xeon due to its broader wins. The Xeon 6515P is the recommended choice for general-purpose compute, scientific workloads, and any task that benefits from high single-thread performance and memory bandwidth. The EPYC 7352 is the specialist, best deployed where encryption and data compression are the primary jobs.