AMD EPYC 9354 vs Intel Xeon 658X Comparison

The AMD EPYC 9354 and Intel Xeon 658X are both 5 nm server processors aimed at the same socket-class market, yet they deliver almost opposite performance profiles. The EPYC 9354 is a 32-core Zen 4 part built for massive parallel throughput, while the Xeon 658X is a 24-core Granite Rapids chip with a dramatic boost clock advantage. Benchmark results show a near-total split: Intel wins every Cinebench test and single-threaded Passmark, while AMD dominates in integer, encryption, compression, and physics workloads. The data reveals a clear choice depending on whether your workload scales across many cores or depends on high per-thread speed.

Head-to-Head Benchmarks

The most striking pattern in the head-to-head results is the total sweep of Cinebench tests by the Intel Xeon 658X, though by very narrow margins. In Cinebench R23 multi-core, the Xeon scores 62,466 against the EPYC’s 61,722, a 1.2% advantage. The same 1.2% delta appears in R20 multi-core (26,235 vs 25,923) and R15 multi-core (6,296 vs 6,221). Single-core Cinebench results follow the same script: the Xeon leads by 1.1% in R15 (888 vs 878) and 1.2% in both R20 (3,703 vs 3,659) and R23 (8,818 vs 8,713). These are consistent wins, but they are so small that they are unlikely to translate into meaningful real-world differences for most applications.

The Passmark suite tells a completely different story, with AMD winning seven of the ten tests and often by huge margins. The EPYC 9354’s biggest wins come in data encryption (71,400 vs 52,357, a 36.4% lead), random string sorting (140,690 vs 103,028, a 36.6% lead), and find prime numbers (934 vs 649, a 43.9% lead). Integer math also favors AMD heavily at 304,828 vs 263,995, a 15.5% advantage. Data compression shows a 10% lead for AMD (1,168,626 vs 1,062,062), and physics is another AMD blowout at 9,281 vs 6,470, a 43.4% gap. Extended instructions are close, with AMD ahead just 1.8% (86,176 vs 84,626).

Intel’s Passmark wins are concentrated in fewer but significant areas. Floating point math goes to the Xeon at 210,480 vs 188,894, a 10.3% advantage. The multithread score is nearly identical, with Intel ahead 73,490 vs 72,615, only 1.2%. The single-thread score is where Intel dominates: 3,728 vs 2,601, a 30.2% lead. That single-thread number is the largest delta in the entire benchmark list, and it reflects the Xeon’s much higher boost clock of 4.90 GHz against the EPYC’s 3.80 GHz.

Architecture Differences

The two processors represent different design philosophies despite sharing a 5 nm process node. The AMD EPYC 9354 uses the Zen 4 architecture, codenamed Genoa, and is manufactured by TSMC. It packs 32 cores and 64 threads, with a base clock of 3.25 GHz and a boost clock of 3.80 GHz. The chip is built from 8 dies of 72 mm² each, totaling 52,560 million transistors. Cache is organized as 64 KB L1 per core, 1 MB L2 per core, and a massive 256 MB shared L3.

The Intel Xeon 658X uses the Granite Rapids architecture, also on a 5 nm process but fabricated by Intel itself. It has fewer cores at 24, with 48 threads, a lower base clock of 3.00 GHz, but a much higher boost clock of 4.90 GHz. The die size is listed as 2x 598 mm², and no transistor count is provided. Cache hierarchy differs significantly: L1 is larger at 112 KB per core, L2 doubles to 2 MB per core, but L3 is smaller at 144 MB shared.

Memory support is another major divergence. Both support DDR5 with ECC, but the EPYC 9354 uses a twelve-channel memory bus delivering 460.8 GB/s of bandwidth, while the Xeon 658X uses an eight-channel bus at 409.6 GB/s. PCIe is identical on paper: Gen 5 with 128 lanes (CPU only) for both. The EPYC 9354 has no integrated graphics, while the Intel part lists integrated graphics as N/A. The EPYC 9354’s multiplier is locked; the Xeon 658X’s multiplier is unlocked. The EPYC 9354 uses AMD Socket SP5, while the Xeon 658X uses Intel Socket 4710.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9354 has 32 cores and 64 threads, while the Intel Xeon 658X has 24 cores and 48 threads. That gives AMD a 33% advantage in core count and thread count.

Q: Why does the Intel Xeon 658X win single-threaded benchmarks by such a large margin?

A: The data shows the Xeon 658X has a boost clock of 4.90 GHz versus the EPYC 9354’s 3.80 GHz. In Passmark single-thread testing, the Xeon scores 3,728 against AMD’s 2,601, a 30.2% advantage that aligns with the clock speed gap.

Q: How do the two compare in multi-threaded workloads?

A: They are nearly identical despite the core count difference. In Cinebench R23 multi-core, the Xeon 658X scores 62,466 versus the EPYC 9354’s 61,722, a 1.2% Intel lead. Passmark multithread is also close, with Intel ahead 73,490 vs 72,615, again only 1.2%.

Q: Which processor has more L3 cache?

A: The AMD EPYC 9354 has 256 MB of shared L3 cache, while the Intel Xeon 658X has 144 MB. AMD’s L3 is 77% larger.

Q: What is the memory bandwidth difference?

A: The EPYC 9354 uses a twelve-channel memory bus with 460.8 GB/s bandwidth, while the Xeon 658X uses an eight-channel bus with 409.6 GB/s. AMD has a 12.5% bandwidth advantage.

Q: Are both processors on the same manufacturing process?

A: Yes, both are fabricated on a 5 nm process, but the EPYC 9354 is made by TSMC while the Xeon 658X is made by Intel.

Specification Differences

The following specifications differ between the AMD EPYC 9354 and Intel Xeon 658X:

• Cores: 32 (AMD) vs 24 (Intel)
• Threads: 64 (AMD) vs 48 (Intel)
• Base Clock: 3.25 GHz (AMD) vs 3.00 GHz (Intel)
• Boost Clock: 3.80 GHz (AMD) vs 4.90 GHz (Intel)
• TDP: 280 W (AMD) vs 250 W (Intel)
• Socket: AMD Socket SP5 vs Intel Socket 4710
• Architecture: Zen 4 (AMD) vs Granite Rapids (Intel)
• Codename: Genoa (AMD) vs Granite Rapids (Intel)
• Foundry: TSMC (AMD) vs Intel (Intel)
• Transistors: 52,560 million (AMD) vs not listed (Intel)
• Die Size: 8x 72 mm² (AMD) vs 2x 598 mm² (Intel)
• L1 Cache: 64 KB per core (AMD) vs 112 KB per core (Intel)
• L2 Cache: 1 MB per core (AMD) vs 2 MB per core (Intel)
• L3 Cache: 256 MB shared (AMD) vs 144 MB shared (Intel)
• Memory Bus: Twelve-channel (AMD) vs Eight-channel (Intel)
• Memory Bandwidth: 460.8 GB/s (AMD) vs 409.6 GB/s (Intel)
• Integrated Graphics: None (AMD) vs N/A (Intel)
• Multiplier Unlocked: No (AMD) vs Yes (Intel)
• Release Date: 2022-11-09 (AMD) vs 2026-02-01 (Intel)
• Launch MSRP: $3420 (AMD) vs $1699 (Intel)

Where Each One Wins

The AMD EPYC 9354 is the clear winner for workloads that hammer integer arithmetic, encryption, compression, and sorting. Its 15.5% lead in integer math, 36.4% lead in data encryption, and 36.6% lead in random string sorting make it the stronger choice for database queries, file compression pipelines, and security-related processing. The 43.4% advantage in physics simulation and the 43.9% lead in prime number finding reinforce that AMD’s extra cores and larger L3 cache pay off in parallel numeric workloads. The 460.8 GB/s memory bandwidth also gives it an edge in memory-bound tasks that the Xeon’s 409.6 GB/s cannot match.

The Intel Xeon 658X wins every Cinebench test, both single and multi-core, though by only about 1.2%. The more decisive wins are in floating point math (10.3% ahead) and single-thread performance (30.2% ahead). The 4.90 GHz boost clock makes it the better choice for lightly threaded applications, legacy software that doesn’t scale across many cores, or any workload where per-thread latency matters more than raw throughput. The 112 KB L1 and 2 MB L2 per core also suggest stronger per-core performance in latency-sensitive loops. For workloads that are predominantly single-threaded or floating-point heavy, the Xeon 658X is the data-backed pick.

The Verdict

The benchmark data paints a clear picture: the Intel Xeon 658X and AMD EPYC 9354 are not competing for the same users. If your applications are single-threaded or floating-point bound, the Xeon 658X is the obvious choice. It wins all six Cinebench results, including both single-core and multi-core, and its 30.2% single-thread advantage over the EPYC is decisive. The fact that it also wins multi-core Cinebench despite having 8 fewer cores suggests that its per-core efficiency and higher clock speed more than compensate for the core deficit in those specific render workloads.

Conversely, the EPYC 9354 is the right pick for broad parallel workloads that use integer math, encryption, compression, or physics. Its 43.9% lead in prime number finding and 43.4% lead in physics are massive, and its 36.4% encryption advantage is hard to ignore for security-focused servers. The 256 MB L3 cache and 12-channel memory bus provide headroom for large datasets that the Xeon’s 144 MB L3 and 8-channel bus cannot offer. The EPYC 9354 also carries a launch MSRP of $3420, which is higher than the Xeon 658X’s $1699 launch MSRP, but the benchmark page does not include pricing analysis.

Both processors sit at the 97th percentile among all CPUs, indicating they are both top-tier parts. The average benchmark score favors the EPYC 9354 at 126,810 versus the Xeon 658X’s 116,060, largely due to AMD’s commanding wins in Passmark’s integer-heavy tests. However, the Xeon 658X’s nearest rivals list includes the AMD EPYC 9255 with a score of 116,388, showing it is competitive with newer AMD parts. The EPYC 9354’s nearest rival is the Intel Xeon 6730P at 124,756, just 1.6% behind. Choose the EPYC 9354 for throughput-oriented, multi-threaded server tasks; choose the Xeon 658X for single-thread speed and floating-point performance.