AMD EPYC 9575F vs Intel Xeon 696X Comparison

The AMD EPYC 9575F and Intel Xeon 696X are both 64-core, 128-thread server processors, yet their benchmark results diverge dramatically. The EPYC 9575F wins all 14 head-to-head comparisons, often by wide margins, while the Xeon 696X trails in every measured category. The data shows a clear performance hierarchy, but the Xeon 696X’s lower launch MSRP and distinct architectural choices suggest it still occupies a specific niche.

The Verdict

The data is unambiguous: the AMD EPYC 9575F is the superior performer in every benchmark recorded. It dominates multi-threaded workloads, with a 43.2% lead in Cinebench R23 multi-core (127739 vs 89227) and a 41% advantage in Passmark multithread (147998 vs 104974). For single-threaded tasks, the EPYC also leads, scoring 4173 versus 3742 in Passmark single-thread, an 11.5% edge. The EPYC’s overall average benchmark score is 311774, placing it in the 99th percentile of all CPUs, while the Xeon 696X averages 286102, also in the 99th percentile but with a lower absolute score.

Who should pick which? If the priority is raw compute throughput—whether for rendering, scientific simulations, or data compression—the EPYC 9575F is the clear choice, as it wins every head-to-head test. The Xeon 696X, however, is not without merit: its launch MSRP is $5599 compared to the EPYC’s $11791, and it offers an unlocked multiplier, which the EPYC lacks. This suggests the Xeon is aimed at scenarios where overclocking flexibility or a lower initial cost outweighs peak performance. For workloads that are less dependent on integer math or physics simulation—areas where the EPYC’s leads are 55.9% and 510.6% respectively—the Xeon’s 336 MB of L3 cache and 112 KB L1 per core might offer competitive performance in memory-bandwidth-sensitive tasks, though the benchmark data does not directly confirm this.

Architecture Differences

The two chips are built on fundamentally different designs. The EPYC 9575F uses AMD’s Zen 5 architecture (codename Turin), fabricated on a 4 nm TSMC process, with a die size of 8x 70.6 mm². The Xeon 696X, in contrast, is based on Intel’s Granite Rapids architecture, built on a 5 nm Intel process, with a die size of 2x 598 mm². This process node difference (4 nm vs 5 nm) likely contributes to the EPYC’s higher clock speeds: it has a base clock of 3.30 GHz and a boost of 5.00 GHz, versus the Xeon’s 2.40 GHz base and 4.80 GHz boost. The EPYC also has a higher TDP at 400W, compared to the Xeon’s 350W.

Cache layouts diverge significantly. The EPYC provides 80 KB of L1 and 1 MB of L2 per core, with 256 MB of shared L3. The Xeon offers more per-core cache—112 KB L1 and 2 MB L2—but its L3 is larger overall at 336 MB shared. Memory support also differs: the EPYC uses a twelve-channel DDR5 memory bus with 576.0 GB/s bandwidth, while the Xeon uses an eight-channel DDR5 bus with 409.6 GB/s. Both support ECC memory and PCIe Gen 5 with 128 lanes from the CPU. The EPYC uses AMD Socket SP5, while the Xeon uses Intel Socket 4710. Neither has integrated graphics, and both are marked for the server/workstation segment.

Head-to-Head Benchmarks

The most striking result is in Passmark physics, where the EPYC 9575F scores 20652 versus the Xeon’s 3382—a 510.6% difference. This is by far the largest margin in the dataset, suggesting either a significant architectural advantage in physics simulation or a uniquely optimized instruction path for the EPYC. In integer math, the EPYC leads 891817 to 572072, a 55.9% advantage, which is likely tied to its higher boost clock and Zen 5 core efficiency. Random string sorting shows a 93.2% gap (348500 vs 180392), indicating a strong performance in data manipulation tasks.

The Cinebench multi-core results are perfectly consistent: the EPYC wins R15, R20, and R23 by exactly 43.2% each, scoring 12876, 53650, and 127739 respectively, versus the Xeon’s 8994, 37475, and 89227. This uniformity suggests the performance delta is driven by sustained multi-threaded throughput rather than a specific benchmark quirk. Data encryption shows a 44.7% lead for the EPYC (162818 vs 112529), while floating-point math is closer, with the EPYC ahead 16.8% (526003 vs 450164). Extended instructions are the tightest multi-threaded test, with the EPYC winning 14.2% (197484 vs 172975). Even in single-threaded tests, where the Xeon’s higher per-core cache might help, the EPYC still leads by 11.5% in both Passmark single-thread and single-thread variants (4173 vs 3742).

FAQ

Q: Why does the AMD EPYC 9575F win the Passmark physics test by over 500%?

A: The EPYC scores 20652 in Passmark physics versus the Xeon’s 3382. This 510.6% delta is the largest in the head-to-head data, likely stemming from the EPYC’s higher base clock (3.30 GHz vs 2.40 GHz) and Zen 5 architecture, though the exact reason is not specified in the data.

Q: Is the Intel Xeon 696X ever ahead in any benchmark?

A: No. The head-to-head data lists 14 benchmarks, all won by the AMD EPYC 9575F, with winsA equal to 14 and winsB equal to 0.

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

A: The EPYC 9575F supports a twelve-channel DDR5 memory bus with 576.0 GB/s bandwidth, while the Xeon 696X uses an eight-channel DDR5 bus with 409.6 GB/s. This gives the EPYC a 40.6% higher theoretical memory bandwidth.

Q: How do the L3 cache sizes compare?

A: The EPYC 9575F has 256 MB of shared L3 cache, while the Xeon 696X has 336 MB of shared L3. The Xeon also has larger per-core L1 (112 KB vs 80 KB) and L2 (2 MB vs 1 MB) caches.

Q: Can the Xeon 696X be overclocked?

A: Yes, the Xeon 696X has an unlocked multiplier (multiplierUnlocked: true), whereas the EPYC 9575F does not (multiplierUnlocked: false). This means the Xeon allows user-controlled frequency adjustments.

Q: What is the launch MSRP of each processor?

A: The AMD EPYC 9575F has a launch MSRP of $11791, while the Intel Xeon 696X has a launch MSRP of $5599.

Where Each One Wins

Based strictly on benchmark data, the AMD EPYC 9575F wins every category: Cinebench multi-core (all three versions), Passmark data compression, encryption, extended instructions, prime number finding, floating-point math, integer math, multithread, physics, random string sorting, and single-thread tests. This makes it the definitive choice for compute-heavy workloads such as integer math processing, physics simulations, and data encryption. Its 55.9% lead in integer math (891817 vs 572072) is particularly notable for applications like financial modeling or cryptography. The 510.6% physics advantage suggests it could be ideal for engineering simulations or real-time physics calculations.

The Intel Xeon 696X, despite losing all head-to-head benchmarks, still has its own strengths indicated by the data. Its larger L3 cache (336 MB vs 256 MB) and higher per-core L1/L2 caches (112 KB / 2 MB vs 80 KB / 1 MB) could benefit workloads that repeatedly access the same data sets, though no benchmark in the pack directly measures this. The unlocked multiplier is a unique feature for server processors, potentially allowing users to manually tune performance—something the EPYC cannot offer. Additionally, its lower TDP (350W vs 400W) and lower launch MSRP ($5599 vs $11791) make it a more power-efficient and cost-effective option for environments where the EPYC’s performance headroom is not required.

Specification Differences

The two processors differ in several key specifications. The EPYC 9575F has a base clock of 3.30 GHz and boost clock of 5.00 GHz, while the Xeon 696X operates at 2.40 GHz base and 4.80 GHz boost. TDP is 400W for the EPYC and 350W for the Xeon. Process nodes differ: the EPYC uses TSMC’s 4 nm process with 66,520 million transistors on an 8x 70.6 mm² die, while the Xeon uses Intel’s 5 nm process with a 2x 598 mm² die (transistor count not provided). The EPYC’s cache is 80 KB L1 and 1 MB L2 per core with 256 MB L3, versus the Xeon’s 112 KB L1 and 2 MB L2 per core with 336 MB L3. Memory channels are twelve for the EPYC versus eight for the Xeon, with bandwidths of 576.0 GB/s and 409.6 GB/s respectively. Sockets are AMD Socket SP5 for the EPYC and Intel Socket 4710 for the Xeon. The EPYC is from the EPYC 9005 series (Zen 5 Turin), while the Xeon is from the Xeon 600 series (Granite Rapids-WS). The EPYC has a locked multiplier, while the Xeon is unlocked. Launch dates differ: October 9, 2024 for the EPYC, and February 1, 2026 for the Xeon.