AMD EPYC 9535 vs Intel Xeon 6781P Comparison

Head-to-Head Benchmarks

The data shows a clear split between the AMD EPYC 9535 and the Intel Xeon 6781P: Intel wins the majority of the benchmark comparisons, but AMD takes the lead in the workloads that matter most for certain server roles. Out of 11 head-to-head tests, the Xeon 6781P claims 7 wins, while the EPYC 9535 secures 4. The most dramatic margin is in the physics test, where the Intel part scores 17,753 against AMD's 3,834 — a 78.4% advantage. This is not a close contest; it is a rout in that specific workload.

The Xeon 6781P also dominates in prime number finding, scoring 1,687 versus the EPYC's 874, a 48.2% lead. That gap suggests Intel's architecture is significantly stronger in certain integer-heavy, latency-sensitive loops. The extended instructions test also favors Intel, with a score of 199,048 versus 175,784, an 11.7% edge. Floating-point math goes Intel's way too: 507,406 against 457,047, a 9.9% win. Data compression is another Intel victory, 2,441,690 versus 2,308,822, a 5.4% margin. Random string sorting follows the same pattern, with Intel ahead 268,573 to 247,506, a 7.8% difference. The multithread score is close, but Intel still edges it out: 117,946 versus 114,528, a 2.9% lead.

AMD's wins are concentrated in areas where its Zen 5 architecture shows clear strengths. Single-thread performance goes decisively to the EPYC 9535: 3,720 versus 3,152, an 18% advantage. This is a substantial lead and directly relevant to lightly threaded or latency-sensitive tasks. Integer math is another big AMD win: 730,281 versus 584,834, a 24.9% margin. That is the largest percentage win for either side in the entire comparison. Data encryption also favors AMD, 127,372 versus 119,623, a 6.5% edge. The overall picture is that Intel wins on raw throughput in many parallel workloads, but AMD wins where per-core efficiency and specific instruction paths matter.

When looking at the broader benchmark context, the EPYC 9535 sits at the 100th percentile among all CPUs, while the Xeon 6781P is at the 99th. The average benchmark score tells a different story: the EPYC 9535 averages 379,408, while the Xeon 6781P averages 315,524. That is a 20% difference in favor of AMD, despite Intel winning more individual tests. The discrepancy is explained by the nature of the workloads — AMD's wins in single-thread and integer math may carry more weight in the aggregated score, or the specific benchmark suite weights those tests more heavily. The nearest rivals for each chip further contextualize the results: the EPYC 9535 is 1.6% ahead of the AMD EPYC 9655 and 3.9% ahead of the Intel Xeon 6960P, while the Xeon 6781P is 1.2% ahead of the AMD EPYC 9575F and 1.6% ahead of the AMD EPYC 9734.

FAQ

Q: Which CPU has better single-thread performance?

A: The AMD EPYC 9535 wins decisively. It scores 3,720 in the single-thread benchmark, while the Intel Xeon 6781P scores 3,152. That is an 18% advantage for AMD, which is a significant margin for a single-core metric.

Q: Does the Intel Xeon 6781P win in any major workload categories?

A: Yes, and by large margins. The Xeon leads in physics simulation by 78.4% (17,753 vs. 3,834), prime number finding by 48.2% (1,687 vs. 874), and floating-point math by 9.9% (507,406 vs. 457,047). It also wins in data compression, extended instructions, random string sorting, and the overall multithread score.

Q: How do the two compare in encryption performance?

A: The AMD EPYC 9535 is the winner here. It scores 127,372 in the data encryption test, versus 119,623 for the Intel Xeon 6781P, a 6.5% edge. This is relevant for workloads that rely heavily on cryptographic operations.

Q: What is the difference in overall average benchmark scores?

A: The AMD EPYC 9535 has a significantly higher average benchmark score of 379,408, compared to 315,524 for the Intel Xeon 6781P. This is roughly a 20% difference, favoring AMD. However, the Xeon wins more individual tests, so the aggregate score should be interpreted with caution.

Q: Are these CPUs in the same performance tier?

A: Both are top-tier server parts. The EPYC 9535 is at the 100th percentile among all CPUs, while the Xeon 6781P is at the 99th. Their nearest rivals are also high-end server chips, but the data shows they occupy different niches within that tier.

Q: Which CPU has higher multithread performance?

A: The Intel Xeon 6781P scores 117,946 in the multithread test, slightly ahead of the AMD EPYC 9535's 114,528. The margin is only 2.9%, which is narrow compared to the gaps seen in other tests.

The Verdict

The data points to a clear division of labor. If your workloads are dominated by physics simulation, prime number calculations, floating-point math, or data compression, the Intel Xeon 6781P is the stronger choice. Its 78.4% lead in physics and 48.2% lead in prime finding are not marginal improvements; they are transformative for those specific tasks. The Xeon also holds an edge in the overall multithread score, making it the safer bet for heavily parallel, throughput-oriented server workloads.

On the other hand, the AMD EPYC 9535 is the pick for single-threaded performance and integer-heavy tasks. An 18% single-thread advantage is substantial, and the 24.9% lead in integer math is the largest margin in any test. If your server runs a mix of light and heavy threads, or if you need strong encryption performance, the EPYC 9535's 6.5% edge in that area matters. Its average benchmark score of 379,408 also suggests better overall consistency across a broad range of tasks, even if it loses on specific individual tests.

The Xeon 6781P has more cores (80 vs. 64) and more threads (160 vs. 128), which helps explain its wins in parallel workloads. But the EPYC 9535's higher boost clock (4.30 GHz vs. 3.80 GHz) and smaller process node (4 nm vs. 5 nm) contribute to its per-core superiority. For a server that handles a diverse workload mix, the EPYC 9535's higher average score makes it the more versatile option. For a server dedicated to a narrow set of parallel, math-heavy tasks, the Xeon 6781P is the specialist that delivers where it counts.

Specification Differences

The two processors diverge on several core specifications. The Intel Xeon 6781P has 80 cores and 160 threads, while the AMD EPYC 9535 has 64 cores and 128 threads. Base clocks differ: the EPYC runs at 2.40 GHz, the Xeon at 2.00 GHz. Boost clocks also favor AMD: 4.30 GHz versus 3.80 GHz. Thermal design power (TDP) is higher on the Intel part at 350 watts, compared to 300 watts for the AMD. The socket is different: AMD uses Socket SP5, Intel uses Socket 4710.

Cache configurations are notably different. The EPYC 9535 has 80 KB of L1 per core and 1 MB of L2 per core, with 256 MB of shared L3. The Xeon 6781P has 112 KB of L1 per core and 2 MB of L2 per core, with 336 MB of shared L3. Memory support is DDR5 for both, but the EPYC uses a twelve-channel memory bus with 576.0 GB/s bandwidth, while the Xeon uses an eight-channel bus with 409.6 GB/s. PCIe support is also different: the EPYC offers Gen 5 with 128 lanes, the Xeon offers Gen 5 with 136 lanes. Both support ECC memory and have no integrated graphics.

Architecture Differences

The AMD EPYC 9535 is built on the Zen 5 architecture, codenamed Turin, and belongs to the EPYC 9005 series. It uses a 4 nm process node from TSMC, with a die size of 8x 70.6 mm² and 66,520 million transistors. The Intel Xeon 6781P uses the Granite Rapids architecture, part of the Xeon 6 (Granite Rapids-SP) generation. It is built on a 5 nm process node from Intel, with a die size of 2x 598 mm². The transistor count is not listed for the Intel part.

These architectural differences explain the benchmark results. AMD's smaller 4 nm process allows for higher clock speeds — 4.30 GHz boost versus 3.80 GHz — which directly contributes to its 18% single-thread win. The larger L1 and L2 caches per core on the Intel side (112 KB and 2 MB, respectively, versus 80 KB and 1 MB) likely help with its lead in cache-sensitive workloads like prime finding and physics simulation. The Xeon's larger shared L3 cache (336 MB vs. 256 MB) also plays a role in its multithread performance. The memory bandwidth gap is stark: AMD's twelve-channel bus delivers 576.0 GB/s, 40% more than Intel's 409.6 GB/s, yet the Xeon still wins the multithread test, indicating that Intel's core count and cache design compensate for the bandwidth disadvantage.

Where Each One Wins

The Intel Xeon 6781P is the clear winner in physics simulation, with a 78.4% lead. This makes it the go-to choice for scientific computing, finite element analysis, or any workload that relies heavily on rigid body dynamics or particle simulations. Its 48.2% edge in prime number finding suggests strength in number theory, cryptography-related math, or similar integer-heavy loops. The 9.9% lead in floating-point math makes it suitable for scientific computing and financial modeling that depends on FPU throughput. Data compression (5.4% lead) and random string sorting (7.8% lead) round out Intel's wins, making it strong for database workloads and data processing pipelines.

The AMD EPYC 9535 takes the win in single-thread performance by 18%, which is critical for database transaction processing, web serving, or any workload with a high degree of per-thread latency sensitivity. Its 24.9% lead in integer math makes it ideal for general-purpose computing, encryption, and integer-heavy business applications. The 6.5% edge in data encryption is a direct advantage for secure communications, VPN gateways, or any server that handles a lot of encrypted traffic. The EPYC's higher average benchmark score (379,408 vs. 315,524) also suggests it is the more balanced performer across a broad range of tasks, making it the safer default for mixed workloads. The data shows a simple rule: pick the Xeon for specialized parallel math, pick the EPYC for everything else.