CPU Comparison

The AMD EPYC 9455P and Intel Xeon 6740E are both high-core-count server processors aimed at dense compute workloads, but the benchmark data positions them very differently. Across the head-to-head suite, the AMD EPYC 9455P wins 16 of 17 tests, while the Intel Xeon 6740E takes a single victory. The average benchmark score for the EPYC 9455P is 218063, compared to 187718 for the Xeon 6740E, a gap of 16.2% according to the nearestRivals data. This page breaks down where each processor excels, the architectural reasons behind those results, and what the raw numbers mean for workload selection.

Where Each One Wins

The AMD EPYC 9455P dominates nearly every category in the head-to-head results. Its largest margins come in compute-heavy integer and floating-point workloads: passmark_physics shows a 118.8% lead, passmark_find_prime_numbers a 107% lead, and passmark_single_thread an 87.8% lead. These are not marginal differences, they represent a fundamental per-core performance advantage that carries through all Cinebench tests, where the EPYC 9455P is consistently 53.2% to 53.3% ahead across R15, R20, and R23, in both single-core and multi-core variants.

The Intel Xeon 6740E wins only passmark_data_encryption, scoring 137106 against 115202 for the EPYC 9455P, a 16% advantage. This is a meaningful result for encryption-heavy workloads such as VPN termination, secure socket layer processing, or database encryption. The Xeon 6740E also comes closer in passmark_data_compression (1786845 vs 1932632, an 8.2% deficit) and passmark_random_string_sorting (220684 vs 241849, a 9.6% deficit), suggesting that while the AMD part still wins, the gap narrows for memory-bandwidth-bound tasks that involve large data movement.

For multi-threaded throughput, the EPYC 9455P’s passmark_multithread score of 116713 is 53.2% higher than the Xeon’s 76167. The pattern is consistent: single-thread performance, multi-thread performance, integer math, floating-point math, and extended instructions all favor the AMD chip. The Xeon 6740E’s single win in encryption, combined with its closer margins in compression and sorting, suggests it is best suited to specific data-transformation pipelines where its unique architecture, discussed below, provides an edge.

Architecture Differences

The two processors take fundamentally different design approaches. The AMD EPYC 9455P uses the Zen 5 architecture, codenamed Turin, fabricated on a 4 nm process at TSMC. It integrates 48 cores and 96 threads, with a boost clock of 4.40 GHz and a base clock of 3.15 GHz. The Intel Xeon 6740E uses the Sierra Forest architecture, codenamed Sierra Forest, built on Intel’s 5 nm process. It packs 96 cores but only 96 threads, no simultaneous multithreading, with a boost clock of 3.20 GHz and a base clock of 2.40 GHz.

Cache configurations differ sharply. The EPYC 9455P has 80 KB of L1 per core, 1 MB of L2 per core, and a shared 256 MB L3 cache. The Xeon 6740E has 96 KB of L1 per core, 4 MB of L2 per module, and a shared 96 MB L3 cache. The AMD part’s larger L3 cache (256 MB vs 96 MB) is likely a factor in its dominance in tests like passmark_find_prime_numbers, where working sets can exceed smaller caches. The Xeon’s larger L2 per module (4 MB) suggests a different strategy for data locality, but the benchmark results show it does not compensate for the overall L3 deficit.

Memory subsystems also differ. The EPYC 9455P uses a twelve-channel DDR5 memory bus with 576.0 GB/s bandwidth, while the Xeon 6740E uses an eight-channel DDR5 bus with 409.6 GB/s. The AMD part’s 40.6% higher memory bandwidth aligns with its wins in data compression and random string sorting, both of which are sensitive to memory throughput. PCIe lanes favor AMD as well: the EPYC 9455P provides 128 Gen 5 lanes (CPU only), while the Xeon 6740E provides 88 Gen 5 lanes.

The physical package differs too. The EPYC 9455P uses AMD Socket SP5, while the Xeon 6740E uses Intel Socket 4710. The AMD chip has a die size listed as 8x 70.6 mm², while the Intel chip has a single 578 mm² die. The AMD part’s TDP is 300 W, the Intel part’s TDP is 250 W. Both support DDR5 and ECC memory, and both are active production parts with no integrated graphics.

Head-to-Head Benchmarks

The Cinebench results are remarkably uniform. In cinebench_r15_multicore, the EPYC 9455P scores 9999 against 6525 for the Xeon 6740E, a 53.2% delta. The single-core variant shows the same 53.2% delta (1411 vs 921). This pattern repeats in R20 (multicore: 41666 vs 27191, 53.2%; singlecore: 5882 vs 3838, 53.3%) and R23 (multicore: 99206 vs 64741, 53.2%; singlecore: 14005 vs 9140, 53.2%). The consistency across all three Cinebench versions indicates a core-level efficiency advantage that scales uniformly with thread count, not a workload-specific artifact.

Outside Cinebench, the margins vary. Passmark_integer_math shows the EPYC 9455P at 606087 vs 457614, a 32.4% lead. Passmark_floating_point_math shows 358673 vs 309333, a 16% lead. Passmark_extended_instructions gives the AMD part its second-largest margin at 75.4% (138505 vs 78968), suggesting a significant advantage in vectorized or specialized instruction throughput. The largest single margin is passmark_physics at 118.8% (16647 vs 7608), followed by passmark_find_prime_numbers at 107% (1089 vs 526).

The Intel Xeon 6740E’s only win is passmark_data_encryption, where it scores 137106 against 115202, a 16% advantage. This is notable because encryption workloads often rely on dedicated instruction sets (such as AES-NI) rather than raw core count or clock speed. The Xeon’s architecture appears to handle this specific instruction mix more efficiently, even though it loses in extended instructions overall. In data compression, the EPYC 9455P wins by 8.2% (1932632 vs 1786845), and in random string sorting by 9.6% (241849 vs 220684), both closer than the Cinebench margins, indicating that the Xeon’s memory bandwidth, while lower, is not the sole bottleneck in these tasks.

FAQ

Q: Why does the AMD EPYC 9455P win so many more benchmarks despite having half the cores?

A: The EPYC 9455P has 48 cores and 96 threads, while the Xeon 6740E has 96 cores and 96 threads. However, the AMD part’s boost clock is 4.40 GHz versus 3.20 GHz, and its per-core L2 cache is 1 MB versus 4 MB per module (which covers multiple cores). The benchmark data shows a 53.2% to 53.3% lead in all Cinebench tests, indicating that per-core performance, not raw core count, drives the results.

Q: What is the single biggest benchmark gap between the two?

A: The largest delta is in passmark_physics, where the EPYC 9455P scores 16647 and the Xeon 6740E scores 7608, a 118.8% difference. The second-largest is passmark_find_prime_numbers at 107% (1089 vs 526).

Q: Is there any workload where the Intel Xeon 6740E is better?

A: Yes, in passmark_data_encryption, the Xeon 6740E scores 137106 versus 115202 for the EPYC 9455P, a 16% advantage. This is the only head-to-head test the Intel part wins, out of 17 total.

Q: How do the memory systems compare?

A: The EPYC 9455P uses a twelve-channel DDR5 bus with 576.0 GB/s bandwidth. The Xeon 6740E uses an eight-channel DDR5 bus with 409.6 GB/s bandwidth. Both support ECC memory.

Q: What are the core and thread counts?

A: The EPYC 9455P has 48 cores and 96 threads (2 threads per core). The Xeon 6740E has 96 cores and 96 threads (1 thread per core, no SMT).

Q: How does the average benchmark score compare?

A: The EPYC 9455P’s average benchmark score is 218063, while the Xeon 6740E’s is 187718. According to nearestRivals data, the EPYC 9455P is 16.2% ahead of the Xeon 6740E.

Specification Differences

• Cores: 48 (AMD) vs 96 (Intel)
• Threads: 96 (AMD) vs 96 (Intel)
• Base Clock: 3.15 GHz (AMD) vs 2.40 GHz (Intel)
• Boost Clock: 4.40 GHz (AMD) vs 3.20 GHz (Intel)
• TDP: 300 W (AMD) vs 250 W (Intel)
• Socket: AMD Socket SP5 (AMD) vs Intel Socket 4710 (Intel)
• Architecture: Zen 5 (AMD) vs Sierra Forest (Intel)
• Process Node: 4 nm (AMD, TSMC) vs 5 nm (Intel, Intel)
• Die Size: 8x 70.6 mm² (AMD) vs 578 mm² (Intel)
• Transistors: 66,520 million (AMD) vs not listed (Intel)
• L1 Cache: 80 KB per core (AMD) vs 96 KB per core (Intel)
• L2 Cache: 1 MB per core (AMD) vs 4 MB per module (Intel)
• L3 Cache: 256 MB shared (AMD) vs 96 MB shared (Intel)
• Memory Bus: Twelve-channel (AMD) vs Eight-channel (Intel)
• Memory Bandwidth: 576.0 GB/s (AMD) vs 409.6 GB/s (Intel)
• PCIe: Gen 5, 128 Lanes (AMD) vs Gen 5, 88 Lanes (Intel)
• Part Number: 100-000001563 (AMD) vs SRPFV (Intel)

The Verdict

The data is unambiguous for most workloads: the AMD EPYC 9455P is the stronger processor. It leads by 53.2% in every Cinebench test, by 32.4% in integer math, by 16% in floating-point math, and by 75.4% in extended instructions. For general-purpose compute, multi-threaded rendering, scientific simulation, or any task that stresses per-core throughput, the EPYC 9455P is the clear choice based on these benchmarks.

The Intel Xeon 6740E’s single win in data encryption (16% ahead) makes it a niche option for specialized encryption pipelines. Its closer margins in data compression (8.2% behind) and random string sorting (9.6% behind) suggest it can handle memory-intensive data transformation tasks respectably, but not better than the AMD part. The Xeon 6740E’s 96 cores provide no multi-threaded advantage in this dataset, it loses passmark_multithread by 53.2%, because the EPYC 9455P’s higher clocks and superior cache configuration compensate for its smaller core count.

For buyers prioritizing raw performance across a broad range of benchmarks, the EPYC 9455P wins 16 of 17 tests and holds a 16.2% average score advantage. For buyers with encryption-dominated workloads, the Xeon 6740E’s 16% win in that specific test may justify its selection, but the overall evidence points to the AMD part as the more capable processor in nearly every other scenario.