AMD EPYC 4585PX vs Intel Xeon 654 Comparison

Where Each One Wins

The head-to-head benchmark record is decisively lopsided. Out of 17 recorded comparisons, the AMD EPYC 4585PX takes 16 wins, while the Intel Xeon 654 claims a single victory. This is not a close contest in aggregate, but the one Intel win reveals a meaningful specialization worth examining.

The AMD EPYC 4585PX dominates every Cinebench iteration, both single-core and multi-core, with a consistent 15.9% margin across all six tests. In Cinebench R23 multi-core, the AMD scores 60,451 against Intel's 52,150. The single-core gap is equally consistent: 8,534 versus 7,362. This uniformity across rendering workloads suggests a fundamental per-thread advantage that scales cleanly into multi-threaded results.

PassMark results reinforce the AMD lead in most categories. The largest single margin is in find prime numbers, where AMD leads by 40.3% (547 versus 390). Integer math shows a 19.6% advantage (248,563 versus 207,745). Single-thread performance is 20.1% ahead (4,538 versus 3,778). Data encryption shows a 16.1% lead (47,224 versus 40,675). Physics simulation favors AMD by 18.2% (6,612 versus 5,596). Multithreaded PassMark scores put AMD ahead by 12.3% (68,908 versus 61,353).

The Intel Xeon 654 wins exactly one test: floating point math. Its score of 163,093 beats AMD's 153,219 by 6.1%. This is the only benchmark where Intel's architecture demonstrates a clear advantage, and it is a meaningful one for workloads that are heavily dependent on floating-point throughput rather than integer or encryption operations.

The use-case split is therefore straightforward. For rendering, encryption, compression, sorting, physics, and general integer workloads, the AMD EPYC 4585PX is the stronger part. For floating-point-heavy compute, the Intel Xeon 654 has a measurable edge, though it is the exception rather than the rule.

Architecture Differences

The two processors come from fundamentally different design philosophies. The AMD EPYC 4585PX is built on the Zen 5 architecture, codenamed Grado, and belongs to the EPYC 4005 series. It is fabricated on a 4 nm process at TSMC, with a die size of 2x 70.6 mm² and 16,630 million transistors. The Intel Xeon 654 uses the Granite Rapids architecture, part of the Xeon 600 series (Granite Rapids-WS), fabricated on a 5 nm process at Intel, with a die size of 2x 598 mm².

Core counts differ, with Intel offering 18 cores and 36 threads against AMD's 16 cores and 32 threads. Despite having two more cores, Intel cannot translate that into a multi-core win in most tests. The AMD part compensates with higher clock speeds: a base clock of 4.30 GHz and a boost clock of 5.70 GHz, versus Intel's 3.10 GHz base and 4.80 GHz boost. The AMD processor also carries a lower TDP of 170 watts compared to Intel's 200 watts.

Cache hierarchies are notably different. AMD provides 80 KB of L1 per core, 1 MB of L2 per core, and a 128 MB L3 cache. Intel provides 112 KB of L1 per core, 2 MB of L2 per core, but only 72 MB of shared L3. AMD's L3 is nearly double Intel's, which likely contributes to its strong performance in data compression (884,774 versus 818,902) and random string sorting (95,210 versus 82,828), workloads that benefit from large on-die caches.

Memory architecture is another major divergence. Both support DDR5 and ECC memory. But Intel uses an eight-channel memory bus with 409.6 GB/s of bandwidth, while AMD uses a dual-channel bus with 89.6 GB/s. Intel's memory bandwidth is over 4.5 times higher. This does not appear to rescue Intel in most tests, but the floating point win may be partially attributable to the memory subsystem feeding data to compute units more rapidly.

PCIe connectivity differs substantially. Intel provides Gen 5 with 128 lanes (CPU only), while AMD provides Gen 5 with 24 lanes (CPU only). This is a significant platform-level difference for workstation and server configurations that need many expansion devices, storage controllers, or GPUs. The Intel part also has an unlocked multiplier, whereas the AMD part is locked.

The AMD processor includes integrated Radeon Graphics, while the Intel Xeon 654 has no integrated graphics. The AMD part is built for Socket AM5, while Intel uses Socket 4710. Release dates differ as well, with AMD launching on 2025-05-12 and Intel on 2026-02-01. The AMD launch MSRP is $699, while the Intel launch MSRP is $1,199.

The Verdict

The benchmark data points to a clear overall winner. The AMD EPYC 4585PX wins 16 of 17 head-to-head tests, holds the 97th percentile among all CPUs in the database, and posts an average benchmark score of 99,324. The Intel Xeon 654 sits at the 96th percentile with an average score of 90,717. The AMD part is roughly 9.5% higher in average score, and its nearest rivals include the AMD Ryzen Threadripper PRO 9955WX at -1.7% and the AMD EPYC 7513 at -2.9%. The Intel part's nearest rivals include the AMD Ryzen AI Max+ 392 at 0.2% and the Intel Xeon w7-2575X at 2.9%.

For buyers prioritizing rendering, encryption, compression, integer math, physics, and single-thread responsiveness, the AMD EPYC 4585PX is the data-backed choice. Its consistent 15.9% lead across all Cinebench workloads and its 20.1% lead in PassMark single-thread performance make it particularly strong for mixed workstation duties. The lower TDP of 170 watts versus 200 watts also reduces thermal demands.

For buyers whose workloads are dominated by floating-point math, the Intel Xeon 654 offers a 6.1% advantage in that specific PassMark test. Additionally, the Intel platform provides 128 PCIe Gen 5 lanes versus 24, and eight-channel memory with 409.6 GB/s of bandwidth versus dual-channel at 89.6 GB/s. These platform capabilities matter for systems that need massive I/O expansion or memory-intensive parallel processing, even if the raw CPU benchmark scores favor AMD in most categories.

The Intel part also has more cores (18 versus 16) and an unlocked multiplier, which may appeal to users who plan to overclock. However, the recorded benchmark data shows no overclocked results, so any gains from that capability are not reflected in these measurements.

FAQ

Q: Which processor has the higher average benchmark score?

A: The AMD EPYC 4585PX has an average benchmark score of 99,324, compared to 90,717 for the Intel Xeon 654.

Q: How large is the AMD lead in Cinebench R23 multi-core?

A: AMD scores 60,451 against Intel's 52,150, a margin of 15.9%.

Q: Is there any benchmark where the Intel Xeon 654 wins?

A: Yes, PassMark floating point math. Intel scores 163,093 against AMD's 153,219, a 6.1% advantage.

Q: How do the core counts compare?

A: The Intel Xeon 654 has 18 cores and 36 threads. The AMD EPYC 4585PX has 16 cores and 32 threads.

Q: Which processor has more memory bandwidth?

A: The Intel Xeon 654 has an eight-channel memory bus with 409.6 GB/s of bandwidth. The AMD EPYC 4585PX has a dual-channel bus with 89.6 GB/s.

Q: What is the difference in PCIe lane counts?

A: The Intel Xeon 654 provides Gen 5 with 128 lanes (CPU only). The AMD EPYC 4585PX provides Gen 5 with 24 lanes (CPU only).

Head-to-Head Benchmarks

The most striking pattern in the head-to-head data is the uniformity of the AMD advantage in Cinebench. Every single Cinebench test, from R15 to R23, and both single-core and multi-core variants, shows exactly 15.9% delta. This consistency indicates that the AMD architecture scales its advantage identically across thread counts and workload sizes. In Cinebench R23 multi-core, AMD's 60,451 versus Intel's 52,150 translates to over 8,000 points of difference. In Cinebench R15 single-core, AMD's 860 versus Intel's 742 is a 118 point gap.

PassMark data shows more variation. The largest margin is in find prime numbers, where AMD leads by 40.3% (547 versus 390). This is a substantial gap for a workload that often rewards raw integer throughput and efficient branch handling. Integer math also favors AMD heavily at 19.6% (248,563 versus 207,745). Single-thread performance, measured twice in the database with identical scores, shows AMD ahead by 20.1% (4,538 versus 3,778).

Data encryption favors AMD by 16.1% (47,224 versus 40,675), and physics simulation favors AMD by 18.2% (6,612 versus 5,596). Random string sorting shows AMD ahead by 14.9% (95,210 versus 82,828), while data compression shows a narrower 8% lead (884,774 versus 818,902). Extended instructions favor AMD by 9.3% (69,457 versus 63,539).

The Intel Xeon 654's sole win in floating point math (163,093 versus 153,219) is worth contextualizing. The 6.1% margin is modest compared to AMD's larger wins elsewhere, but it is the only area where Intel's 18-core, eight-channel memory design outperforms AMD's 16-core, dual-channel design. For workloads that are purely floating-point bound, the Intel part is the stronger choice by the recorded data.

Overall, the head-to-head record favors AMD in 16 of 17 tests, with an average delta across all tests that heavily skews toward the EPYC 4585PX. The AMD part's percentile ranking of 97 versus Intel's 96 reflects this, as does the average benchmark score gap of roughly 8,600 points.