AMD EPYC 9175F vs Intel Xeon 654 Comparison

The AMD EPYC 9175F and Intel Xeon 654 represent two distinct philosophies for the server/workstation segment, with the data showing a decisive pattern of wins for the AMD part across nearly every benchmark category. While both processors are active production parts targeting the same market segment, their architectural approaches and resulting performance profiles diverge sharply. The EPYC 9175F, built on a 4 nm process with Zen 5 architecture, consistently outpaces the Xeon 654, which uses a 5 nm process and Granite Rapids architecture, in both single-threaded and most multi-threaded workloads. However, the Intel part counters with a significant victory in floating-point math, and it offers a substantially lower launch MSRP, presenting a nuanced choice for buyers who prioritize specific workload types over raw aggregate performance.

FAQ

Q: Which processor has more physical cores?

A: The Intel Xeon 654 has 18 cores, while the AMD EPYC 9175F has 16 cores. This gives the Intel part a two-core advantage in core count.

Q: How much larger is the L3 cache on the AMD EPYC 9175F?

A: The AMD EPYC 9175F features a massive 512 MB shared L3 cache, compared to 72 MB shared on the Intel Xeon 654. This represents a significant difference in cache capacity.

Q: What is the single-thread performance gap in Cinebench R23?

A: The AMD EPYC 9175F scores 7895 in Cinebench R23 single-core, which is 7.2% higher than the Intel Xeon 654's score of 7362. This consistent 7.2% delta appears across all Cinebench tests.

Q: Are both processors unlocked for overclocking?

A: No. The Intel Xeon 654 has an unlocked multiplier, while the AMD EPYC 9175F does not have an unlocked multiplier. This is a notable difference for users seeking manual tuning.

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

A: The AMD EPYC 9175F supports twelve-channel DDR5 memory with a bandwidth of 576.0 GB/s, while the Intel Xeon 654 supports eight-channel DDR5 with a bandwidth of 409.6 GB/s. The AMD part offers 166.4 GB/s more theoretical bandwidth.

Q: Which CPU has a higher boost clock?

A: The AMD EPYC 9175F has a boost clock of 5.00 GHz, which is higher than the Intel Xeon 654's 4.80 GHz boost clock. The base clocks also differ, with AMD at 4.20 GHz versus Intel's 3.10 GHz.

Architecture Differences

The architectural divide between these two processors is substantial. The AMD EPYC 9175F is built on TSMC's 4 nm process node from the EPYC 9005 series, codenamed Turin, using the Zen 5 architecture. In contrast, the Intel Xeon 654 uses Intel's own 5 nm process from the Xeon 600 generation, codenamed Granite Rapids. This process difference contributes to the AMD part's ability to reach higher clock speeds—4.20 GHz base and 5.00 GHz boost—compared to Intel's 3.10 GHz base and 4.80 GHz boost.

Core and cache configurations further separate the two. Despite having fewer cores (16 vs. 18), the AMD EPYC 9175F allocates 80 KB of L1 and 1 MB of L2 per core, while the Intel Xeon 654 allocates 112 KB of L1 and 2 MB of L2 per core. The L3 cache tells a different story: AMD provides a colossal 512 MB shared L3, dwarfing Intel's 72 MB shared L3. This 440 MB difference likely explains AMD's dominance in cache-sensitive workloads like data compression and random string sorting.

Memory architecture also diverges. The EPYC 9175F uses a twelve-channel memory bus with 576.0 GB/s bandwidth, whereas the Xeon 654 uses an eight-channel bus with 409.6 GB/s bandwidth. Both support DDR5 and ECC memory, and both offer PCIe Gen 5 with 128 lanes. The Intel part has an unlocked multiplier, a feature absent on the AMD chip, and it sports a die size of 2x 598 mm², while the AMD chip uses 16x 70.6 mm² chiplets totaling 133,040 million transistors. The AMD processor also carries a higher TDP of 320 watts, compared to Intel's 200 watts.

The Verdict

The data points decisively toward the AMD EPYC 9175F for most buyers. With 16 wins out of 17 head-to-head benchmarks, the AMD part establishes superiority across a broad spectrum of workloads. The Cinebench results show a uniform 7.2% advantage for AMD in both single-core and multi-core tests, indicating a consistent architectural efficiency lead. In PassMark tests, the AMD chip wins by margins ranging from 4% in data encryption to a staggering 90% in find prime numbers. It also leads in multithread performance by 10.2%, which is notable given that the Intel part has two additional cores.

However, the Intel Xeon 654 is not without merit. It wins in floating-point math by 10.5%, a significant margin that suggests a superior floating-point unit design. This could tip the scales for workloads heavily reliant on FP calculations, such as certain scientific simulations or financial modeling. Furthermore, the Intel part's launch MSRP of $1199 is substantially lower than AMD's $4256, and it has an unlocked multiplier for potential overclocking, though the higher TDP and lower memory bandwidth are trade-offs. For users whose primary workloads involve floating-point math or who have strict power budgets, the Xeon 654 deserves consideration. For virtually everything else, the EPYC 9175F is the stronger choice based on raw performance metrics.

Specification Differences

The two processors differ in several key specification fields. The AMD EPYC 9175F has 16 cores and 32 threads, while the Intel Xeon 654 has 18 cores and 36 threads. Clock speeds differ: AMD's base clock is 4.20 GHz versus Intel's 3.10 GHz, and AMD's boost clock is 5.00 GHz versus Intel's 4.80 GHz. TDP ratings are 320 watts for AMD and 200 watts for Intel. Sockets are incompatible: AMD uses Socket SP5, while Intel uses Socket 4710. Process nodes are 4 nm (TSMC) for AMD and 5 nm (Intel) for Intel.

Cache configurations diverge significantly: AMD has 80 KB L1 per core, 1 MB L2 per core, and 512 MB shared L3; Intel has 112 KB L1 per core, 2 MB L2 per core, and 72 MB shared L3. Memory support is DDR5 for both, but AMD uses a twelve-channel bus with 576.0 GB/s bandwidth, while Intel uses an eight-channel bus with 409.6 GB/s bandwidth. The launch MSRP is $4256 for AMD and $1199 for Intel. The multiplier is locked on AMD and unlocked on Intel. The AMD part is from the EPYC 9005 series (Zen 5, Turin), while the Intel part is from the Xeon 600 series (Granite Rapids-WS). The AMD chip has a transistor count of 133,040 million and a die size of 16x 70.6 mm², whereas the Intel die size is 2x 598 mm². Release dates also differ: AMD on 2024-10-09 and Intel on 2026-02-01.

Head-to-Head Benchmarks

The head-to-head data reveals a clear pattern of AMD dominance. In Cinebench R15, R20, and R23, the AMD EPYC 9175F wins every single-core and multi-core test by exactly 7.2%. For example, Cinebench R23 multi-core shows AMD at 55923 versus Intel's 52150, and single-core shows 7895 versus 7362. This uniformity suggests a fundamental per-clock efficiency advantage for Zen 5 in these rendering workloads.

The PassMark suite shows even wider gaps in some areas. The AMD chip wins data compression by 5.9% (867186 vs. 818902), data encryption by 4% (42297 vs. 40675), extended instructions by 11% (70529 vs. 63539), and integer math by 5.8% (219800 vs. 207745). The multithread test shows a 10.2% lead for AMD (67634 vs. 61353), and single-thread tests show a 12.7% advantage (4256 vs. 3778). The most dramatic wins come in find prime numbers, where AMD scores 741 versus Intel's 390—a 90% delta—and in physics, where AMD leads 9984 to 5596, a 78.4% gap. Random string sorting also favors AMD by 15.6% (95783 vs. 82828).

The lone Intel victory is in floating-point math, where the Xeon 654 scores 163093 against AMD's 145939, a 10.5% margin. This is the only test where Intel outperforms, and the margin is substantial, indicating a specialized strength rather than a fluke.

Where Each One Wins

The AMD EPYC 9175F wins across most general-purpose and compute-intensive workloads. It excels in integer math, data compression, encryption, extended instructions, and multithreaded tasks, making it well-suited for database operations, data analytics, general server virtualization, and code compilation. Its massive 512 MB L3 cache and higher memory bandwidth likely drive its wins in cache-sensitive tasks like random string sorting and prime number finding. The physics test result, with a 78.4% lead, suggests AMD's architecture handles certain physics simulation workloads far better. For rendering tasks as measured by Cinebench, the AMD part is uniformly 7.2% faster in both single-threaded and multi-threaded scenarios, making it the clear pick for content creation and 3D rendering workflows.

The Intel Xeon 654's single win in floating-point math is its defining strength. This result implies that for workloads dominated by floating-point operations—such as scientific computing, financial risk modeling, or certain engineering simulations—the Intel part can deliver better performance. The 10.5% margin is significant and could translate to meaningful time savings in batch processing of FP-heavy tasks. Additionally, the Intel part's lower TDP (200 watts vs. 320 watts) makes it more power-efficient on paper, which could be a deciding factor in dense server environments where thermal management is critical. The unlocked multiplier also offers overclocking potential, though the data does not quantify what gains might be achievable. For users with FP-centric workloads or strict power constraints, the Xeon 654 presents a compelling case; for all other tested workloads, the EPYC 9175F is the data-backed winner.