AMD EPYC 9384X vs Intel Xeon 6737P Comparison

The Intel Xeon 6737P and AMD EPYC 9384X are both 32-core, 64-thread server processors built for dense compute, but the recorded head-to-head data shows a clear overall winner. The Intel takes 15 of 16 benchmark comparisons, with the AMD's only victory coming in data encryption. The Intel's average benchmark score is 140694, placing it at the 98th percentile of all CPUs in the database, while the AMD averages 120427 and sits at the 97th percentile. The Intel's nearest rivals in the database are all within 2.5% of its average score, while the AMD's nearest rivals span a wider range.

Head-to-Head Benchmarks

The most decisive result is in PassMark floating point math, where the Intel scores 258811 against 174630 for the AMD, a 48.2% advantage. That is the largest single margin in the entire comparison and indicates a substantial difference in vector and floating-point throughput. The extended instructions test reinforces this: the Intel scores 105453 versus 74363, a 41.8% lead. Prime number finding also goes to the Intel by 16.9% (697 versus 596), and integer math favors the Intel by 11.1% (330756 versus 297833).

The Cinebench suite is uniformly in the Intel's favor. In Cinebench R15 multicore, the Intel scores 6822 against 5968, a 14.3% lead. The R15 single-core test shows 963 versus 842, a 14.4% margin. Cinebench R20 multicore gives 28428 versus 24870, also 14.3%, and R20 single-core gives 4013 versus 3510, again 14.3%. Cinebench R23 multicore follows the same pattern: 67688 versus 59215, a 14.3% lead. The PassMark multithread test lands at 14.3% as well (79634 versus 69665). The consistency of this 14.3% margin across all Cinebench versions and the multithread test suggests a uniform scaling advantage for the Intel across all cores.

Random string sorting goes to the Intel by 8.4% (129510 versus 119440). Data compression is closer, with the Intel at 1157255 versus 1119983, a 3.3% margin. The physics test is nearly a tie: 9362 versus 9332, a 0.3% edge for the Intel. Single-thread performance is also close: 3048 versus 3015, a 1.1% margin.

The AMD's only win is in data encryption, where it scores 72631 against 65615 for the Intel, a 9.7% advantage. This is a meaningful result for cryptographic workloads, and it is the one area where the AMD's architecture clearly outperforms the Intel.

Where Each One Wins

The Intel Xeon 6737P is the better choice for floating-point-heavy and SIMD-heavy workloads. The 48.2% lead in floating point math and the 41.8% lead in extended instructions point to strong vector and math library performance. The consistent 14.3% margins across the Cinebench suite and PassMark multithread indicate that the Intel scales well across all cores in general multithreaded tasks. Prime number finding, integer math, and random string sorting all favor the Intel, making it the better fit for general compute, data processing, and sorting workloads. The Intel also leads in single-thread performance, though by a small margin (1.1% in PassMark single-thread and 14.4% in Cinebench R15 single-core).

The AMD EPYC 9384X wins in data encryption by 9.7%, which makes it the better option for workloads that depend on encryption performance. The AMD also has a twelve-channel memory bus with 460.8 GB/s of bandwidth versus the Intel's eight-channel 409.6 GB/s, which can help in memory-bandwidth-bound scenarios even though the benchmark suite does not directly measure that. The AMD's 768 MB of shared L3 cache is substantially larger than the Intel's 144 MB, which can benefit workloads with large working sets that fit in cache.

The near-tie in physics (0.3%) and the small single-thread margin (1.1%) mean that the two are effectively interchangeable in these specific tests. The AMD's higher base clock (3.10 GHz versus 2.90 GHz) is offset by the Intel's higher boost clock (4.00 GHz versus 3.90 GHz) in the recorded results.

Architecture Differences

The Intel Xeon 6737P is built on the Granite Rapids architecture, using a 5 nm process at Intel's own foundry. The AMD EPYC 9384X uses the Zen 4 architecture in the Genoa-X variant, also on a 5 nm process but fabricated by TSMC. Both are 5 nm parts, but the design philosophies differ substantially.

The Intel has a monolithic die of 598 mm², while the AMD uses a chiplet design with eight dies of 72 mm² each. The AMD packs 90,160 million transistors across its chiplets; the Intel's transistor count is not recorded in the database.

Cache is the biggest architectural differentiator. The Intel provides 112 KB of L1 per core and 2 MB of L2 per core, with 144 MB of shared L3. The AMD provides 64 KB of L1 per core and 1 MB of L2 per core, but its shared L3 is 768 MB, substantially larger than the Intel's L3. That large L3 is a defining feature of the Genoa-X variant.

Memory architecture also differs. The Intel uses an eight-channel DDR5 memory bus with 409.6 GB/s of bandwidth. The AMD uses a twelve-channel DDR5 bus with 460.8 GB/s. Both support ECC memory. PCIe connectivity favors the AMD: 128 Gen 5 lanes versus 88 for the Intel.

The sockets are incompatible: the Intel uses Socket 4710, while the AMD uses Socket SP5.

Specification Differences

The two processors share the same core and thread counts (32 cores, 64 threads) and both are active production parts, but nearly every other specification differs.

The Intel has a base clock of 2.90 GHz and a boost clock of 4.00 GHz, while the AMD has a base of 3.10 GHz and a boost of 3.90 GHz. The AMD has a higher base clock but a lower boost clock. Thermal design power differs: the Intel is rated at 270 W, the AMD at 320 W.

The Intel's L1 cache is 112 KB per core versus 64 KB per core for the AMD. The Intel's L2 is 2 MB per core versus 1 MB per core. The L3 is 144 MB shared for the Intel versus 768 MB shared for the AMD.

Memory channels: eight for the Intel, twelve for the AMD. Memory bandwidth: 409.6 GB/s versus 460.8 GB/s. PCIe lanes: 88 for the Intel, 128 for the AMD.

The Intel was released on 2025-02-23, while the AMD was released on 2023-06-12. The Intel's launch MSRP is $4995; the AMD's launch MSRP is $5529. The Intel has a part number of SRVNZ; the AMD's part number is not recorded.

The Intel's nearest rivals in the database include the Intel Xeon 674X (average score 143103, 1.7% above the 6737P), the Intel Xeon 6732P (143444, 1.9% above), the AMD Ryzen 9 PRO 9965X3D (143735, 2.1% above), and the Intel Xeon w9-3575X (144323, 2.5% above). The AMD EPYC 9384X's nearest rivals include the Intel Xeon w7-3565X (118307, 1.8% below the EPYC), the AMD EPYC 7642 (124006, 2.9% above), the AMD Ryzen Threadripper PRO 5975WX (124171, 3% above), and the AMD EPYC 9255 (116388, 3.5% below).

FAQ

Q: Which processor wins more benchmarks in the head-to-head comparison?

A: The Intel Xeon 6737P wins 15 of the 16 recorded head-to-head tests. The AMD EPYC 9384X wins only the PassMark data encryption test.

Q: What is the largest performance gap between the two?

A: The largest gap is in PassMark floating point math, where the Intel scores 258811 versus 174630 for the AMD, a 48.2% advantage. The extended instructions test shows a 41.8% lead for the Intel.

Q: Does the AMD EPYC 9384X have any advantage in the recorded benchmarks?

A: Yes, the AMD wins the data encryption test with 72631 versus 65615, a 9.7% margin. It also has a larger L3 cache (768 MB versus 144 MB) and higher memory bandwidth (460.8 GB/s versus 409.6 GB/s), though those are not directly reflected in the head-to-head benchmark wins.

Q: How do the two compare in single-threaded performance?

A: The Intel leads by a small margin. In PassMark single-thread, the Intel scores 3048 versus 3015, a 1.1% difference. In Cinebench R15 single-core, the Intel scores 963 versus 842, a 14.4% margin.

Q: What are the core and thread counts for each processor?

A: Both have 32 cores and 64 threads.

Q: Which processor has a higher boost clock?

A: The Intel Xeon 6737P has a boost clock of 4.00 GHz, while the AMD EPYC 9384X has a boost clock of 3.90 GHz. The AMD has a higher base clock at 3.10 GHz versus 2.90 GHz for the Intel.