AMD EPYC 9355P vs Intel Xeon w9-3575X Comparison

The AMD EPYC 9355P and Intel Xeon w9-3575X are both high-end server and workstation processors, yet the benchmark data reveals they are optimized for fundamentally different workloads. The EPYC 9355P, built on the Zen 5 architecture, dominates the majority of compute-heavy tasks, while the Xeon w9-3575X shows specific strengths in floating-point and extended instruction workloads. This analysis breaks down the recorded measurements to show where each processor excels, the architectural reasons for those differences, and which user profile aligns with each chip.

Where Each One Wins

The recorded head-to-head results show a clear split: the AMD EPYC 9355P wins 14 of the 16 benchmark comparisons, while the Intel Xeon w9-3575X takes only 2. The EPYC 9355P’s victories are concentrated in multithreaded, integer, and encryption-heavy tasks. Its most dominant win is in PassMark physics, where it scores 13,515 against the Xeon’s 6,836, a delta of 97.7%. It also leads massively in prime number finding (1,044 vs 687, a 52% advantage) and integer math (412,067 vs 298,224, a 38.2% lead). These results point to a processor that excels in scientific simulation, cryptography, and general integer processing.

The Xeon w9-3575X secures its two wins in floating-point math (273,398 vs 256,635, a 6.1% advantage) and extended instructions (109,843 vs 107,622, a 2% edge). These are specialized areas, but they matter for workloads such as financial modeling, certain engineering simulations, and vectorized code that leverages advanced instruction sets. The Xeon’s wins are narrower than the EPYC’s, but they are consistent in their categories.

For single-threaded tasks, the EPYC 9355P leads by a slim margin. In PassMark single-thread, it scores 3,747 versus 3,672, a 2% difference. That small edge carries through all Cinebench single-core tests, where the EPYC wins by 16.7% each time. In multi-core Cinebench tests, the EPYC’s margin is also 16.7% across R15, R20, and R23, showing a consistent architectural advantage rather than a workload-specific fluke.

Architecture Differences

The two processors come from different design philosophies. The AMD EPYC 9355P is part of the EPYC 9005 series, codenamed Turin, and uses the Zen 5 architecture. It is fabricated on a 4 nm process at TSMC, with a die size listed as 8x 70.6 mm² and a transistor count of 66,520 million. The Intel Xeon w9-3575X is codenamed Sapphire Rapids, part of the Xeon W generation, and uses a 10 nm process from Intel with a die size of 4x 477 mm². The process node difference is significant: the smaller 4 nm node allows AMD to pack more transistors into a smaller area, which likely contributes to its efficiency and performance per core.

Core counts differ substantially. The EPYC 9355P has 32 cores and 64 threads, while the Xeon w9-3575X has 44 cores and 88 threads. Despite having 12 fewer cores, the EPYC wins most multi-threaded benchmarks, which indicates higher per-core throughput. The Xeon’s base clock is lower at 2.20 GHz versus the EPYC’s 3.55 GHz, though the Xeon has a higher boost clock at 4.80 GHz versus 4.40 GHz. That boost advantage does not translate into single-thread wins in the recorded data.

Cache configurations also differ. The EPYC 9355P has 80 KB of L1 per core, 1 MB of L2 per core, and a massive 256 MB of shared L3 cache. The Xeon w9-3575X also has 80 KB of L1 per core but doubles the L2 to 2 MB per core, while its L3 is 97.5 MB. The EPYC’s L3 cache is more than 2.5 times larger, which helps in workloads that repeatedly access large datasets.

Memory architecture is another differentiator. The EPYC 9355P supports twelve-channel DDR5 memory with a bandwidth of 576.0 GB/s. The Xeon w9-3575X supports eight-channel DDR5 with a bandwidth of 307.2 GB/s. The EPYC offers nearly double the memory bandwidth, which is critical for data-intensive server workloads. Both support ECC memory, and both lack integrated graphics.

PCIe connectivity is close but not equal. The EPYC 9355P provides Gen 5 with 128 lanes (CPU only), while the Xeon w9-3575X provides Gen 5 with 112 lanes (CPU only). The EPYC has 16 more lanes, which matters for systems with many GPUs or NVMe drives.

The Xeon w9-3575X has an unlocked multiplier, making it overclockable, while the EPYC 9355P is locked. The Xeon also has a higher TDP at 340 watts versus the EPYC’s 280 watts. The EPYC was released on 2024-10-09, while the Xeon launched on 2024-08-23.

FAQ

Q: Which processor is faster in single-threaded workloads?

A: The AMD EPYC 9355P wins all single-thread benchmarks. It scores 1,176 in Cinebench R15 single-core, 4,901 in R20, 11,670 in R23, and 3,747 in PassMark single-thread. The Intel Xeon w9-3575X scores 1,008, 4,200, and 3,672 in the corresponding R15, R20, and PassMark tests. The Cinebench single-core deltas are 16.7% in favor of the EPYC.

Q: Does the Intel Xeon w9-3575X have any benchmark wins?

A: Yes, it wins two tests. It scores 273,398 in PassMark floating-point math versus the EPYC’s 256,635, a 6.1% advantage. It also scores 109,843 in PassMark extended instructions versus 107,622, a 2% lead.

Q: How does the core count compare?

A: The Intel Xeon w9-3575X has 44 cores and 88 threads, while the AMD EPYC 9355P has 32 cores and 64 threads. Despite fewer cores, the EPYC wins 14 of 16 head-to-head benchmarks.

Q: What is the memory bandwidth difference?

A: The AMD EPYC 9355P supports twelve-channel DDR5 with a bandwidth of 576.0 GB/s. The Intel Xeon w9-3575X supports eight-channel DDR5 with a bandwidth of 307.2 GB/s.

Q: Which processor has more L3 cache?

A: The AMD EPYC 9355P has 256 MB of shared L3 cache. The Intel Xeon w9-3575X has 97.5 MB of L3 cache.

Q: Are both processors currently in production?

A: Yes, both the AMD EPYC 9355P and the Intel Xeon w9-3575X have a production status of Active. Both are categorized as Server/Workstation processors.

Specification Differences

The core count is the most obvious difference: 32 cores for the EPYC 9355P versus 44 for the Xeon w9-3575X, with thread counts of 64 and 88 respectively. Base clocks differ at 3.55 GHz for the EPYC and 2.20 GHz for the Xeon. Boost clocks are 4.40 GHz for the EPYC and 4.80 GHz for the Xeon. TDP is 280 watts for the EPYC and 340 watts for the Xeon.

The process node differs: 4 nm for the EPYC (TSMC foundry) versus 10 nm for the Xeon (Intel foundry). The EPYC reports 66,520 million transistors and a die size of 8x 70.6 mm², while the Xeon has no transistor count listed and a die size of 4x 477 mm².

Cache structure differs in L2 and L3. Both have 80 KB of L1 per core. The EPYC has 1 MB of L2 per core, while the Xeon has 2 MB per core. L3 is 256 MB shared for the EPYC versus 97.5 MB for the Xeon.

Memory channels and bandwidth differ: twelve-channel with 576.0 GB/s for the EPYC, eight-channel with 307.2 GB/s for the Xeon. PCIe lanes also differ: Gen 5 with 128 lanes for the EPYC versus Gen 5 with 112 lanes for the Xeon.

The sockets are different: AMD Socket SP5 for the EPYC, Intel Socket 4677 for the Xeon. The Xeon has an unlocked multiplier, while the EPYC is locked. The launch MSRP is $2998 for the EPYC and $3789 for the Xeon. Release dates are 2024-10-09 for the EPYC and 2024-08-23 for the Xeon. Part numbers are 100-000001521 for the EPYC and SRN72 for the Xeon.

Head-to-Head Benchmarks

The largest margin in any test is PassMark physics, where the EPYC 9355P scores 13,515 against the Xeon’s 6,836, a 97.7% delta. This is a near-doubling of performance and indicates a massive advantage in physics simulation and related floating-point-heavy but integer-driven workloads.

The next biggest margin is PassMark find prime numbers, where the EPYC scores 1,044 versus 687, a 52% lead. This test is heavily dependent on integer arithmetic and cache efficiency. The EPYC’s 256 MB L3 cache and higher base clock likely contribute to this result.

PassMark integer math shows a 38.2% advantage for the EPYC: 412,067 versus 298,224. This is a core workload for many server applications, including database operations and general data processing. PassMark random string sorting also favors the EPYC by 31.2%, with scores of 176,697 versus 134,723.

PassMark data encryption is another strong EPYC win: 80,961 versus 62,258, a 30% delta. This matters for secure communications and storage encryption. PassMark data compression shows a 17.3% lead for the EPYC: 1,429,976 versus 1,219,584.

The Cinebench suite is uniformly 16.7% in favor of the EPYC across R15, R20, and R23, for both single-core and multi-core tests. R15 multi-core scores are 8,332 versus 7,140; R20 multi-core is 34,719 versus 29,751; R23 multi-core is 82,666 versus 70,837. Single-core scores are 1,176 versus 1,008 (R15), 4,901 versus 4,200 (R20), and 11,670 for the EPYC in R23 (the Xeon has no R23 single-core score recorded).

PassMark multithread shows a 15.9% lead for the EPYC: 96,603 versus 83,338. PassMark single-thread is close, with the EPYC at 3,747 and the Xeon at 3,672, a 2% delta.

The Xeon’s two wins are narrower. In floating-point math, it scores 273,398 versus 256,635, a 6.1% advantage. In extended instructions, it scores 109,843 versus 107,622, a 2% edge. These wins show the Xeon is competitive in vectorized and floating-point-heavy code, but they are isolated victories.

The average benchmark scores place the EPYC at 160,358 and the Xeon at 144,323. The EPYC’s nearest rivals include the AMD EPYC 9375F at 162,497 (-1.3% delta) and the AMD EPYC 7663 at 161,973 (-1% delta), while the Xeon’s nearest rival is the AMD EPYC 7643P at 144,824 (-0.3% delta). Both processors sit in the 98th percentile of all CPUs.

The Verdict

The AMD EPYC 9355P is the stronger processor for the majority of workloads. It wins 14 of 16 head-to-head comparisons, including all Cinebench tests, all multithreaded PassMark tests, and all encryption and compression tests. Its advantages in physics simulation (97.7%), prime number finding (52%), and integer math (38.2%) make it the clear choice for scientific computing, cryptography, data processing, and server-side virtualization. The 576.0 GB/s memory bandwidth and 256 MB L3 cache are the structural reasons for these wins.

The Intel Xeon w9-3575X is the better pick only for workloads that specifically rely on floating-point math and extended instructions. Its 6.1% lead in floating-point math and 2% lead in extended instructions are real but narrow. Users running financial simulations, certain engineering analysis tools, or code that heavily uses modern vector instruction sets may see a benefit from the Xeon. Its unlocked multiplier also allows overclocking, which is a feature the EPYC lacks.

For every other task, the data points to the EPYC 9355P. It delivers higher performance with lower TDP (280 watts versus 340 watts), more PCIe lanes (128 versus 112), and nearly double the memory bandwidth. The Xeon’s higher core count (44 versus 32) does not translate into benchmark wins, which means the EPYC’s per-core efficiency is superior. The verdict is straightforward: the EPYC 9355P for general compute, multithreaded server work, and integer-heavy tasks; the Xeon w9-3575X for niche floating-point and extended-instruction workloads.