AMD EPYC 7F72 vs Intel Xeon 638 Comparison

Head-to-Head Benchmarks

The benchmark data presents a clear split between the AMD EPYC 7F72 and the Intel Xeon 638. The Intel Xeon 638 secures more wins overall, taking 12 of the 17 recorded head-to-head comparisons, while the AMD EPYC 7F72 wins 5. However, the margin of victory tells a more nuanced story than the raw win count.

The Intel Xeon 638 dominates in Cinebench workloads across the board. In Cinebench R15 multi-core, it scores 4757 against the AMD's 4518, a 5% advantage. The single-core R15 result shows a similar gap, 671 versus 637, a 5.1% lead. This pattern repeats in Cinebench R20, where the Intel part scores 19824 versus 18828 in multi-core, and 2798 versus 2657 in single-core, both 5% margins. Cinebench R23 follows the same trajectory: 47202 versus 44829 in multi-core and 6663 versus 6328 in single-core, again a 5% delta in both cases.

The PassMark single-thread test produces the largest Intel advantage of the entire comparison. Here, the Xeon 638 scores 3670 while the EPYC 7F72 manages 2384, a 35% gap. This is a substantial difference in lightly threaded performance and indicates a significant architectural advantage for Intel in single-core execution.

The Intel part also wins in PassMark floating-point math, scoring 144757 against 108437, a 25.1% advantage. Extended instructions go Intel's way too, with 56498 versus 46936, a 16.9% margin. PassMark integer math is closer, with Intel scoring 184884 against 181103, a 2% edge. The PassMark multithread test also favors Intel, 55651 versus 52740, a 5.2% lead.

The AMD EPYC 7F72, however, shows decisive strengths in specific workloads. The largest AMD win is in data encryption, where it scores 56261 against the Intel's 36030, a massive 56.2% advantage. Random string sorting also heavily favors AMD, 102436 versus 74318, a 37.8% gap. Physics testing shows AMD ahead by 37.3%, scoring 6459 versus 4704. Data compression goes to AMD by 11.4%, with scores of 808795 versus 725818. Prime number finding is another AMD win, 498 versus 381, a 30.7% margin.

These results indicate that the Intel Xeon 638 is consistently faster in general-purpose compute and floating-point heavy tasks, while the AMD EPYC 7F72 excels in memory-intensive and security-related workloads. The overall average benchmark scores reflect this balance: the AMD EPYC 7F72 has an average score of 85072 across all benchmarks, while the Intel Xeon 638 trails slightly at 80723. The AMD part also holds a 96th percentile ranking among all CPUs, compared to the Intel's 95th percentile.

Architecture Differences

The two processors come from fundamentally different design philosophies. The AMD EPYC 7F72 is built on the Zen 2 architecture, codenamed Rome, using a 7 nm process from TSMC. It contains 3,800 million transistors on a die size of 74 mm². The Intel Xeon 638 uses the Granite Rapids architecture, also its codename, fabricated on Intel's 5 nm process with a die size of 598 mm². Transistor counts for the Intel part are not recorded in the database.

Core configurations differ significantly. The AMD EPYC 7F72 has 24 cores and 48 threads, while the Intel Xeon 638 has 16 cores and 32 threads. This gives AMD a 50% advantage in core and thread count. Both processors have a base clock of 3.20 GHz, but the Intel part boosts substantially higher at 4.80 GHz versus the AMD's 3.70 GHz boost.

Cache hierarchies reflect different approaches. The AMD EPYC 7F72 has 96 KB of L1 cache per core, 512 KB of L2 per core, and a massive 192 MB of shared L3 cache. The Intel Xeon 638 has 112 KB of L1 per core, 2 MB of L2 per core, but only 72 MB of shared L3. AMD's larger L3 cache likely contributes to its strong showing in data compression and encryption benchmarks.

Memory support also differs. The AMD part supports DDR4 memory across an eight-channel bus, while the Intel part supports DDR5 across a quad-channel bus. Interestingly, both achieve a recorded memory bandwidth of 204.8 GB/s. Both support ECC memory. PCIe connectivity differs as well: AMD offers Gen 4, while Intel provides Gen 5 with 80 lanes from the CPU only.

The socket platforms are entirely different. AMD uses Socket SP3, while Intel uses Socket 4710. The Intel Xeon 638 has an unlocked multiplier, whereas the AMD EPYC 7F72 does not. The Intel part has no integrated graphics, and the AMD part also lists no integrated graphics.

Release dates show a significant generational gap. The AMD EPYC 7F72 was released in April 2020, while the Intel Xeon 638 is dated February 2026. The Intel launch MSRP is $899, while the AMD part has no recorded launch MSRP.

Where Each One Wins

The Intel Xeon 638 is the clear choice for single-threaded and lightly threaded workloads. Its 35% lead in PassMark single-thread performance, combined with consistent 5% wins across all Cinebench versions in both single-core and multi-core, makes it suitable for applications that prioritize per-core performance. The 25.1% advantage in floating-point math suggests strength in scientific computing, financial modeling, and any workload heavy on double-precision arithmetic. The 16.9% edge in extended instructions points to better handling of SIMD-heavy code. The 2% win in integer math, while modest, adds to its overall general-purpose compute advantage.

The AMD EPYC 7F72 is the stronger part for memory-bandwidth-sensitive and security-centric workloads. The 56.2% lead in data encryption is the standout result, indicating hardware acceleration for cryptographic operations. The 37.8% advantage in random string sorting suggests superior memory access patterns and cache behavior. Physics simulation shows a 37.3% AMD lead, which may reflect its thread count advantage. Data compression, a workload that benefits from large caches, shows an 11.4% AMD edge. Prime number finding, another cache and thread dependent task, favors AMD by 30.7%.

For multithreaded general-purpose workloads, the results are mixed. The Intel Xeon 638 wins PassMark multithread by 5.2% and all Cinebench multi-core tests by 5%, despite having fewer cores. This indicates that Intel's higher boost clock and architectural efficiency overcome AMD's core count advantage in these specific tests. However, the AMD part's higher average benchmark score and 96th percentile ranking suggest it performs better across a broader range of tasks.

FAQ

Q: Which processor has more cores?

A: The AMD EPYC 7F72 has 24 cores and 48 threads, while the Intel Xeon 638 has 16 cores and 32 threads.

Q: How does single-thread performance compare?

A: The Intel Xeon 638 leads significantly in PassMark single-thread testing, scoring 3670 against 2384 for the AMD EPYC 7F72, a 35% advantage. Cinebench R23 single-core also goes Intel's way, 6663 versus 6328, a 5% margin.

Q: Which processor is better for data encryption workloads?

A: The AMD EPYC 7F72 is substantially ahead, scoring 56261 in PassMark data encryption versus 36030 for the Intel Xeon 638, a 56.2% advantage.

Q: What is the memory bandwidth of each processor?

A: Both the AMD EPYC 7F72 and the Intel Xeon 638 have a recorded memory bandwidth of 204.8 GB/s, though AMD uses eight-channel DDR4 while Intel uses quad-channel DDR5.

Q: How does the average benchmark score compare between the two?

A: The AMD EPYC 7F72 has an average benchmark score of 85072, while the Intel Xeon 638 scores 80723. The AMD part ranks in the 96th percentile of all CPUs, with the Intel in the 95th percentile.

Q: When was each processor released?

A: The AMD EPYC 7F72 was released in April 2020. The Intel Xeon 638 has a release date of February 2026.

Specification Differences

The AMD EPYC 7F72 has 24 cores and 48 threads, while the Intel Xeon 638 has 16 cores and 32 threads. The base clocks are identical at 3.20 GHz, but the Intel boosts to 4.80 GHz versus the AMD's 3.70 GHz. Thermal design power differs, with AMD at 240 and Intel at 180. The AMD part uses AMD Socket SP3, the Intel part uses Intel Socket 4710. The AMD architecture is Zen 2 with the Rome codename, while Intel uses Granite Rapids. The process nodes are 7 nm from TSMC for AMD and 5 nm from Intel for the Intel part. The AMD die size is 74 mm², the Intel die is 598 mm². L1 cache is 96 KB per core for AMD and 112 KB per core for Intel. L2 cache is 512 KB per core for AMD and 2 MB per core for Intel. L3 cache is 192 MB shared for AMD and 72 MB shared for Intel. Memory support is DDR4 for AMD and DDR5 for Intel. Memory buses are eight-channel for AMD and quad-channel for Intel. PCIe is Gen 4 for AMD and Gen 5 with 80 lanes for Intel. The Intel Xeon 638 has an unlocked multiplier, the AMD EPYC 7F72 does not. The Intel part number is SA2DN, while the AMD part number is 100-000000141100-000000141WOF. The launch MSRP for the Intel Xeon 638 is $899, while the AMD EPYC 7F72 has none recorded.