AMD EPYC 9684X vs Intel Xeon 6781P Comparison

The Verdict

The data splits these two server processors into clear roles. The Intel Xeon 6781P wins 4 of the 14 head-to-head benchmarks, while the AMD EPYC 9684X takes 10. The AMD part is the default choice for raw multi-threaded throughput, leading in every Cinebench multi-core test and in the PassMark multithread score. The Intel part is the pick for workloads that reward single-thread speed, extended instruction throughput, and floating-point math. The AMD EPYC 9684X also holds the overall average benchmark score advantage, though the database places both in the 99th percentile of all CPUs. For a buyer who must choose one, the AMD EPYC 9684X is the safer bet for general server consolidation, while the Intel Xeon 6781P is the specialist for single-thread-bound and FP-heavy tasks.

The nearest-rival data adds context. The Intel Xeon 6781P sits 1.2% above the AMD EPYC 9575F and 1.6% above the AMD EPYC 9734 in average score, while trailing the AMD Ryzen Threadripper PRO 9985WX by 1.6% and the 9980X by 1.9%. The AMD EPYC 9684X, by contrast, sits 6.1% above the Intel Xeon 6980P but 4.6% below the AMD Ryzen Threadripper 9970X and 4.8% below the Intel Xeon 6780E. This shows the 9684X is not universally dominant; its lead over the 6781P in the head-to-head is real but not overwhelming, and specific benchmark categories swing heavily in either direction.

Where Each One Wins

The AMD EPYC 9684X dominates the multi-core and integer-heavy landscape. In Cinebench R15, R20, and R23 multi-core, it wins by 3% each time, with scores of 10418 versus 10105, 43409 versus 42106, and 103355 versus 100254. The PassMark multithread test shows a similar 3% edge, 121595 versus 117946. The gap widens dramatically in integer math, where the 9684X scores 844145 against 584834, a 30.7% advantage. Data compression favors the 9684X by 9.5% (2698807 versus 2441690), and random string sorting goes to AMD by 23.1% (349126 versus 268573). Encryption is the biggest AMD win: 178453 versus 119623, a 33% lead. Physics simulation also favors the 9684X by 28.1% (24686 versus 17753), and prime number finding goes to AMD by 16.5% (2020 versus 1687).

The Intel Xeon 6781P wins where the workload is narrower. Single-thread performance goes to Intel by 9%, with a PassMark single-thread score of 3152 versus 2891. Extended instructions (AVX-512 style workloads) favor Intel by 11.9%, 199048 versus 177956. Floating-point math also goes to Intel by 7.5%, 507406 versus 472174. These are not minor wins; they show Intel's architecture retains a clear edge in vectorized and scalar FP work despite losing the integer and multi-core battles.

Architecture Differences

The two chips come from different design philosophies. The Intel Xeon 6781P uses the Granite Rapids architecture on a 5 nm process fabricated by Intel, with an 80-core, 160-thread configuration. The AMD EPYC 9684X uses Zen 4 on a 5 nm process fabricated by TSMC, with 96 cores and 192 threads. The core count difference (16 more cores on AMD) explains much of the multi-core gap, but not all of it. The Intel part has a higher boost clock at 3.80 GHz versus 3.70 GHz, which aligns with its single-thread win. The AMD part has a higher base clock at 2.55 GHz versus 2.00 GHz.

Cache is a major differentiator. The AMD EPYC 9684X features 1152 MB of shared L3 cache, a massive pool enabled by its Genoa-X design. The Intel Xeon 6781P has 336 MB of shared L3. Per-core L1 and L2 also differ: Intel has 112 KB L1 and 2 MB L2 per core, while AMD has 64 KB L1 and 1 MB L2 per core. The larger L3 on AMD is the likely driver behind its wins in data compression, encryption, and random string sorting, all of which are cache-sensitive. The Intel part compensates with a larger die (2x 598 mm² versus 12x 72 mm²) and a higher TDP of 350 watts versus 400 watts, though the AMD part consumes more power overall.

Memory architecture also differs. The Intel part uses eight-channel DDR5 with 409.6 GB/s bandwidth, while the AMD part uses twelve-channel DDR5 with 460.8 GB/s. The AMD part has more memory channels and higher bandwidth, which supports its multi-threaded and integer-heavy workloads. PCIe lanes are close: Intel has 136 Gen 5 lanes, AMD has 128 Gen 5 lanes. Both support ECC memory. The Intel part uses Socket 4710, the AMD part uses Socket SP5.

FAQ

Q: Which CPU is faster in single-threaded tasks?

A: The Intel Xeon 6781P wins the PassMark single-thread test by 9%, scoring 3152 versus the AMD EPYC 9684X's 2891. This aligns with Intel's higher boost clock of 3.80 GHz versus 3.70 GHz.

Q: Why does the AMD EPYC 9684X win so many multi-core benchmarks?

A: The AMD part has 96 cores and 192 threads versus Intel's 80 cores and 160 threads. It also has a larger shared L3 cache (1152 MB versus 336 MB) and higher memory bandwidth (460.8 GB/s versus 409.6 GB/s), which helps in multi-threaded workloads.

Q: Is the Intel Xeon 6781P better at any math workload?

A: Yes, the Intel part wins floating-point math by 7.5% (507406 versus 472174) and extended instructions by 11.9% (199048 versus 177956). This suggests a stronger vector FP unit despite losing integer math by 30.7%.

Q: How do the two compare in data encryption?

A: The AMD EPYC 9684X dominates, scoring 178453 versus 119623, a 33% advantage. This is the largest single benchmark gap in the head-to-head data.

Q: Which CPU has more cache per core?

A: The Intel Xeon 6781P has more L1 and L2 per core (112 KB and 2 MB respectively) versus AMD's 64 KB and 1 MB. However, AMD has far more total L3 cache (1152 MB versus 336 MB), which is shared across all cores.

Q: What is the launch MSRP for each?

A: The Intel Xeon 6781P has a launch MSRP of $8960, while the AMD EPYC 9684X has a launch MSRP of $14756.

Head-to-Head Benchmarks

The biggest wins for the AMD EPYC 9684X are in encryption and integer math. The encryption gap of 33% (178453 versus 119623) is the largest delta in the dataset. Integer math shows a 30.7% lead (844145 versus 584834), which is a massive difference for any integer-heavy database or application workload. Physics follows at 28.1% (24686 versus 17753), and random string sorting at 23.1% (349126 versus 268573). Prime number finding goes to AMD by 16.5% (2020 versus 1687). Data compression is a 9.5% AMD win (2698807 versus 2441690). The Cinebench multi-core tests and PassMark multithread all show a consistent 3% AMD edge, indicating a stable but modest multi-thread advantage.

The Intel Xeon 6781P's wins are fewer but sharp. Extended instructions show an 11.9% lead (199048 versus 177956), which is the second-largest delta in the entire comparison. Single-thread performance is a 9% Intel win (3152 versus 2891). Floating-point math is a 7.5% Intel win (507406 versus 472174). These three wins reveal a clear pattern: Intel's architecture is more efficient per clock for scalar and vector FP work, even with fewer cores. The data suggests that if a workload is dominated by AVX-512 or FP operations, the Intel part is the better choice despite its lower core count.

Specification Differences

The two processors differ in nearly every major specification. The Intel Xeon 6781P has 80 cores and 160 threads, while the AMD EPYC 9684X has 96 cores and 192 threads. Base clocks are 2.00 GHz for Intel and 2.55 GHz for AMD; boost clocks are 3.80 GHz for Intel and 3.70 GHz for AMD. TDP is 350 watts for Intel and 400 watts for AMD. The Intel part uses Intel Socket 4710, the AMD part uses AMD Socket SP5. Process nodes are both 5 nm, but the foundries differ: Intel fabricates its own chip, while TSMC fabricates the AMD part. The Intel die is 2x 598 mm², while the AMD die is 12x 72 mm². The AMD part has 135,240 million transistors, while the Intel part's transistor count is not recorded.

Cache layouts are distinctly different. Intel provides 112 KB L1 and 2 MB L2 per core, with 336 MB shared L3. AMD provides 64 KB L1 and 1 MB L2 per core, with 1152 MB shared L3. Memory support is DDR5 for both, but Intel uses eight channels with 409.6 GB/s bandwidth, while AMD uses twelve channels with 460.8 GB/s. PCIe is Gen 5 for both, with Intel offering 136 lanes and AMD offering 128 lanes. Both support ECC memory. The Intel part has no integrated graphics, and the AMD part also has no integrated graphics listed. Both are active production parts, with the Intel Xeon 6781P released on 2025-02-23 and the AMD EPYC 9684X released on 2023-06-12. The Intel part number is SRV5J, the AMD part number is 100-100000892. Neither has an unlocked multiplier. The AMD part belongs to the EPYC 9004 series and uses the Zen 4 architecture with the Genoa-X codename, while the Intel part uses Granite Rapids and belongs to the Xeon 6 generation.