AMD EPYC 7663 vs Intel Xeon w9-3575X Comparison

Head-to-Head Benchmarks

The recorded head-to-head data shows a clear split between the AMD EPYC 7663 and the Intel Xeon w9-3575X, with the Intel part taking 11 of the 16 compared tests while the AMD part wins 5. The overall average benchmark score favors the EPYC 7663 at 161973 against 144323 for the Xeon, but that aggregate hides the fact that the two processors dominate in completely different workload types.

The most lopsided victory in either direction belongs to the EPYC 7663 in PassMark data encryption, where it scores 111725 against 62258 for the Xeon, a 79.5% advantage. This is the single largest delta in the entire comparison and indicates a fundamental architectural strength for the AMD part in cryptographic workloads. Similarly, the EPYC 7663 crushes the Xeon in integer math, posting 468096 versus 298224, a 57% lead. Random string sorting also goes heavily to AMD at 184367 versus 134723, a 36.8% edge. Data compression shows an 18.6% win for the EPYC 7663 with scores of 1447020 against 1219584, and the AMD part leads in physics tests at 8020 versus 6836, a 17.3% margin.

The Xeon w9-3575X counterattacks most decisively in extended instructions, where it posts 109843 versus 73719, a 32.9% advantage. Single-thread performance is also a major win for Intel: the Xeon scores 3672 in PassMark single-thread tests against 2606 for the EPYC 7663, a 29% lead. Floating-point math goes to Intel by 8% (273398 versus 251535). The remaining Intel wins are narrow, consistent victories across the Cinebench suite and multithreaded tests. In Cinebench R15, R20, and R23 multicore, the Xeon leads by 1.5% in every case, with scores of 7140 versus 7032, 29751 versus 29304, and 70837 versus 69773 respectively. Single-core Cinebench results mirror this pattern: R15 at 1008 versus 992 (1.6% lead), R20 at 4200 versus 4137 (1.5% lead). PassMark multithread shows the Xeon ahead at 83338 versus 82087, also a 1.5% margin. The find prime numbers test is nearly a tie, with the Xeon at 687 versus 676, a 1.6% edge.

These results tell a clear story: the EPYC 7663 is overwhelmingly stronger in integer-heavy, encryption, compression, and sorting workloads, while the Xeon w9-3575X wins in floating-point, extended instructions, and everything that depends on raw clock speed. The Cinebench multicore margins are small enough to be considered effectively equivalent, but the single-thread gap is substantial.

FAQ

Q: Which processor has the higher average benchmark score?

A: The AMD EPYC 7663 has an average benchmark score of 161973, while the Intel Xeon w9-3575X averages 144323. The EPYC 7663 sits at the 98th percentile of all CPUs, and the Xeon is also at the 98th percentile.

Q: How large is the single-thread performance gap?

A: The Intel Xeon w9-3575X leads by 29% in PassMark single-thread tests, scoring 3672 versus 2606 for the AMD EPYC 7663. This is one of the largest deltas in the comparison.

Q: Which chip wins in encryption workloads?

A: The AMD EPYC 7663 wins decisively in PassMark data encryption, scoring 111725 versus 62258 for the Intel Xeon w9-3575X, a 79.5% advantage. This is the biggest win for either chip in the entire head-to-head set.

Q: Are the Cinebench multicore results close?

A: Yes, the Intel Xeon w9-3575X leads by exactly 1.5% in Cinebench R15, R20, and R23 multicore tests. The scores are 7140 versus 7032, 29751 versus 29304, and 70837 versus 69773, respectively.

Q: What does the nearest rivals data show for the EPYC 7663?

A: The EPYC 7663's nearest rival is the AMD EPYC 9375F, which has an average score of 162497, just 0.3% higher. The EPYC 9355P trails by 1%, the Intel Xeon 676X trails by 2.2%, and the AMD EPYC 7C13 leads by 3.5% with a score of 167788.

Q: How does the Xeon w9-3575X compare to its nearest rivals?

A: The AMD EPYC 7643P is 0.3% higher in average score at 144824, the AMD Ryzen 9 PRO 9965X3D trails by 0.4% at 143735, the Intel Xeon 6732P trails by 0.6% at 143444, and the Intel Xeon 674X trails by 0.9% at 143103.

The Verdict

The data directs different buyers toward different chips. The AMD EPYC 7663 is the correct choice for workloads dominated by integer math, data compression, encryption, random string sorting, and physics simulation. Its 79.5% encryption advantage and 57% integer math lead are not marginal differences; they are decisive gaps that would translate into substantially faster job completion in those specific categories. The 36.8% edge in random string sorting and 18.6% lead in data compression reinforce the pattern of AMD dominance in data manipulation tasks.

The Intel Xeon w9-3575X is the pick for single-threaded applications and floating-point heavy compute. The 29% single-thread lead is massive, and the 32.9% advantage in extended instructions suggests a strong vector-processing capability. The 8% floating-point math win adds to this profile. However, the Cinebench multicore results are nearly identical, with the Xeon leading by only 1.5% across all three versions of the benchmark. This means that for general multi-core rendering, the two processors are effectively tied, and the decision should hinge on the specific workload mix.

For a database of server and workstation parts, the EPYC 7663 offers the higher overall average score (161973 versus 144323) and sits in the same 98th percentile as the Xeon. The Xeon w9-3575X counters with a higher boost clock of 4.80 GHz versus 3.50 GHz for the EPYC, which explains its single-thread dominance. The EPYC's 56 cores and 112 threads give it a raw parallelism advantage over the Xeon's 44 cores and 88 threads, yet the Xeon still manages to win multithreaded PassMark by 1.5%, showing that clock speed can offset core count in some scenarios.

Specification Differences

The AMD EPYC 7663 has 56 cores and 112 threads, while the Intel Xeon w9-3575X has 44 cores and 88 threads. The base clocks differ: 2000.00 MHz for the AMD part versus 2200.00 MHz for the Intel part. The boost clocks are further apart, with the Xeon reaching 4800.00 MHz against 3500.00 MHz for the EPYC. Thermal design power is 240 W for the AMD chip and 340 W for the Intel chip. The EPYC 7663 uses AMD Socket SP3, while the Xeon w9-3575X uses Intel Socket 4677.

The cache structures differ substantially. The EPYC 7663 has 64 KB of L1 per core, 512 KB of L2 per core, and 256 MB of shared L3. The Xeon w9-3575X has 80 KB of L1 per core, 2 MB of L2 per core, and 97.5 MB of L3. The process nodes also differ: the EPYC is built on a 7 nm process at TSMC, while the Xeon uses a 10 nm process at Intel. The EPYC's die is composed of 8x 81 mm² sections with 33,200 million transistors; the Xeon's die is 4x 477 mm² with no transistor count recorded.

Memory support is another dividing line. The EPYC 7663 supports DDR4 with an eight-channel memory bus and 204.8 GB/s of bandwidth. The Xeon w9-3575X supports DDR5, also with an eight-channel bus, but offers 307.2 GB/s of bandwidth. PCIe connectivity differs as well: the EPYC provides Gen 4 with 128 CPU-only lanes, while the Xeon provides Gen 5 with 112 CPU-only lanes. The Xeon has an unlocked multiplier; the EPYC does not. The integrated graphics field is listed as N/A for the Xeon and null for the EPYC.

Architecture Differences

The AMD EPYC 7663 is built on the Zen 3 architecture with the codename Milan, belonging to the EPYC 7003 series. It was released in March 2021. The Intel Xeon w9-3575X uses the Sapphire Rapids codename and belongs to the Xeon W generation, released in August 2024. The process technology differs by node: the EPYC uses TSMC's 7 nm process, while the Xeon uses Intel's 10 nm process. The transistor count is recorded only for the AMD part at 33,200 million, spread across 8 dies of 81 mm² each. The Intel part has 4 dies of 477 mm² each, with no transistor figure in the database.

Both processors support ECC memory, and both target the server and workstation market segment. The EPYC 7663's memory bandwidth is 204.8 GB/s over DDR4, whereas the Xeon w9-3575X reaches 307.2 GB/s over DDR5. The L3 cache is a major architectural difference: the EPYC 7663 has 256 MB of shared L3, which is more than 2.5 times the Xeon's 97.5 MB. The per-core L1 and L2 caches are larger on the Xeon (80 KB and 2 MB per core versus 64 KB and 512 KB per core), but the EPYC's massive shared L3 pool appears to drive its advantages in data-heavy workloads.

The PCIe generations differ, with the EPYC at Gen 4 and the Xeon at Gen 5, though the EPYC offers 128 lanes versus 112 lanes on the Xeon. The Xeon's launch MSRP is $3789, while the EPYC 7663 has a launch MSRP of $6366. The production status for both is active. The Xeon w9-3575X has an unlocked multiplier, a feature absent from the EPYC 7663, and its part number is SRN72, while the EPYC's part number is 100-000000318100-100000318WOF.