AMD EPYC 9115 vs Intel Xeon 6517P Comparison

The Intel Xeon 6517P and AMD EPYC 9115 are both 16-core, 32-thread server processors aimed at similar workloads, yet they represent fundamentally different design philosophies. The data shows a close contest in many synthetic benchmarks, with the Intel part securing 12 head-to-head wins against 5 for the AMD part. However, the margin of victory varies significantly by workload type, revealing distinct architectural strengths. Both processors sit in the 94th percentile of all CPUs, indicating top-tier performance, but their average benchmark scores differ, with the Intel Xeon 6517P posting an average of 72350 compared to the EPYC 9115's 69288.

FAQ

Q: How do the two processors compare in overall average benchmark score?

A: The Intel Xeon 6517P has a higher average benchmark score of 72350, while the AMD EPYC 9115 averages 69288. This places the Intel part about 4.4% ahead in aggregate performance across the recorded test suite.

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

A: The Intel Xeon 6517P consistently wins all three Cinebench multi-core tests. It scores 4268 in R15, 17787 in R20, and 42352 in R23, compared to the AMD EPYC 9115's scores of 4233, 17641, and 42003 respectively, each with a 0.8% delta.

Q: Are there any workloads where the AMD EPYC 9115 is significantly faster?

A: Yes, the AMD EPYC 9115 wins decisively in PassMark integer math, scoring 181807 versus 162671 for the Intel part, a 10.5% advantage. It also leads in data encryption, random string sorting, and single-thread tests, though by smaller margins.

Q: What is the difference in memory bandwidth between the two?

A: The AMD EPYC 9115 offers significantly higher memory bandwidth at 576.0 GB/s, while the Intel Xeon 6517P provides 409.6 GB/s. The AMD part also uses a twelve-channel memory bus compared to the Intel's eight-channel configuration.

Q: Which processor has a larger L3 cache?

A: The Intel Xeon 6517P has a larger L3 cache at 72 MB shared, while the AMD EPYC 9115 has 64 MB shared. However, the AMD EPYC 9115 has a smaller L1 cache per core at 80 KB, compared to the Intel's 112 KB per core.

Q: When were these processors released?

A: The AMD EPYC 9115 was released on 2024-10-09, while the Intel Xeon 6517P was released later on 2025-02-23.

Architecture Differences

The architectural split between these two server CPUs is stark. The Intel Xeon 6517P is built on the Granite Rapids architecture, a 5 nm process from Intel's own foundry, while the AMD EPYC 9115 uses the Zen 5 architecture, codenamed Turin, fabricated by TSMC on a 4 nm node. The AMD part also has a notably different physical construction, with a die size of 2x 70.6 mm² and a transistor count of 16,630 million, details not provided for the Intel chip.

Cache hierarchies reveal different design priorities. The Intel Xeon 6517P allocates a larger per-core L1 cache at 112 KB and L2 cache at 2 MB per core, alongside a 72 MB shared L3. The AMD EPYC 9115 uses 80 KB L1 and 1 MB L2 per core, with a 64 MB shared L3. This suggests Intel's approach favors more aggressive per-core resource allocation, potentially benefiting latency-sensitive workloads, while AMD's smaller per-core caches might rely more on the efficiency of the Zen 5 core design itself.

Platform support diverges significantly. The Intel Xeon 6517P uses Intel Socket 4710 and provides 88 PCIe Gen 5 lanes, while the AMD EPYC 9115 uses AMD Socket SP5 and offers 128 PCIe Gen 5 lanes. The AMD part also features a twelve-channel memory bus with 576.0 GB/s bandwidth, compared to Intel's eight-channel bus with 409.6 GB/s. Both support DDR5 ECC memory, but the AMD platform clearly offers more headroom for memory-intensive and I/O-heavy deployments. Clock speeds differ as well, with the Intel part starting at 3.20 GHz base and boosting to 4.20 GHz, while the AMD part runs at 2.60 GHz base and 4.10 GHz boost. The Intel part has a higher TDP of 190 W, whereas the AMD EPYC 9115 is rated at 125 W.

Head-to-Head Benchmarks

The Cinebench suite shows a consistent, if narrow, advantage for the Intel Xeon 6517P. Across all three versions (R15, R20, R23) and both single-core and multi-core tests, the Intel part wins every round with a uniform 0.8% delta. For example, in Cinebench R23 multi-core, the Intel scores 42352 against 42003, and in single-core it scores 5979 against 5929. This pattern suggests a slight IPC and clock-speed edge for the Intel architecture in rendering workloads, likely driven by its higher boost clock of 4.20 GHz.

The PassMark suite reveals more divergent behavior. The Intel Xeon 6517P dominates in several compute-heavy tests. Its biggest win is in extended instructions, scoring 51891 versus 45477, a 14.1% advantage that points to stronger SIMD or specialized instruction set performance. It also leads in find prime numbers by 15.9% (335 vs 289) and floating point math by 12% (127497 vs 113853). Data compression shows an 8.9% lead (653338 vs 600099), and physics tests show a 6.3% edge (4452 vs 4188). In the broader multithread test, the Intel part wins by 1.7% (49786 vs 48936).

The AMD EPYC 9115's victories are concentrated in specific areas. Its most significant win is in integer math, where it scores 181807 against Intel's 162671, a 10.5% margin. This is a substantial reversal from the floating point result, suggesting different execution unit efficiencies. The AMD part also wins in data encryption (33489 vs 32385, a 3.3% edge) and random string sorting (70151 vs 67480, a 3.8% lead). In single-thread tests, the AMD EPYC 9115 edges ahead, scoring 3360 versus 3311, a 1.5% advantage, which is interesting given the Intel part's higher boost clock.

Specification Differences

The primary specification differences between these processors are clear and impactful. The socket and platform are entirely different: Intel Socket 4710 for the Xeon 6517P versus AMD Socket SP5 for the EPYC 9115, meaning they are not interchangeable in existing server infrastructure. The process nodes differ, with Intel using a 5 nm process and AMD using a 4 nm process. The AMD part has a specified transistor count of 16,630 million and a die size of 2x 70.6 mm², while these figures are not recorded for the Intel part.

Memory support is a major differentiator. Both use DDR5, but the AMD EPYC 9115 has a twelve-channel memory bus delivering 576.0 GB/s, while the Intel Xeon 6517P has an eight-channel bus delivering 409.6 GB/s. PCIe connectivity also favors AMD, with 128 Gen 5 lanes versus 88 on the Intel side. The base and boost clocks are higher on the Intel part (3.20 and 4.20 GHz) compared to the AMD (2.60 and 4.10 GHz), but the AMD part has a significantly lower TDP of 125 W versus 190 W. The cache configurations differ as detailed above, with Intel having larger per-core L1 and L2 caches and a larger L3 cache. The release dates differ, with the AMD EPYC 9115 launching earlier on 2024-10-09 and the Intel Xeon 6517P on 2025-02-23.

The Verdict

The recorded data presents two capable server processors with distinct profiles. The Intel Xeon 6517P is the better choice for workloads that rely on floating point math, extended instruction sets, and data compression. Its consistent wins in all Cinebench tests and the PassMark multithread test suggest it is the stronger general-purpose rendering and compute processor. The 14.1% lead in extended instructions and 12% lead in floating point math are substantial, making it the pick for scientific computing and simulation tasks that use these instruction paths.

The AMD EPYC 9115, however, is the superior option for integer-heavy workloads and memory-bandwidth-bound applications. Its 10.5% win in integer math is a clear signal for database operations, encryption, and certain types of data processing that are integer-centric. The higher memory bandwidth of 576.0 GB/s and twelve-channel bus, combined with 128 PCIe Gen 5 lanes, make it the more scalable platform for large-scale virtualization, high-throughput I/O, and memory-intensive data analytics. The lower TDP of 125 W also makes it a more power-efficient choice in dense server deployments.

For a buyer, the decision hinges on workload composition. If the application mix is dominated by floating point, compression, and general multi-threaded rendering, the Intel Xeon 6517P offers better raw performance in those areas. If the priorities are integer throughput, memory bandwidth, and platform I/O expandability, the AMD EPYC 9115 provides a stronger foundation. The Intel part's higher average benchmark score of 72350 versus 69288 aligns with its 12 wins in the head-to-head suite, but the AMD part's targeted victories in key server workloads prevent a blanket recommendation. The data indicates that the Intel Xeon 6517P is the faster chip overall in most recorded tests, but the AMD EPYC 9115 is the better fit for specific, high-value enterprise workloads.