AMD EPYC 9754 vs Intel Xeon 6980P Comparison

FAQ

Q: Which processor has more cores and threads?

A: Both the AMD EPYC 9754 and the Intel Xeon 6980P feature 128 cores and 256 threads, so thread counts are identical.

Q: How do their single-threaded performances compare?

A: In Cinebench R23 single-core, the EPYC 9754 scores 11850 versus the Xeon 6980P's 8918, a 32.9% advantage for AMD. PassMark single-thread testing shows the same result: 2328 for AMD versus 1681 for Intel, a 38.5% lead.

Q: Which chip has a larger L3 cache?

A: The Intel Xeon 6980P has a shared 504 MB L3 cache, while the AMD EPYC 9754 has a shared 256 MB L3 cache. Intel's L3 is nearly double the size.

Q: What is the power draw difference?

A: The Intel Xeon 6980P has a TDP of 500 watts, while the AMD EPYC 9754 is rated at 360 watts. AMD draws significantly less power under its rated specification.

Q: Which processor supports more PCIe lanes?

A: The AMD EPYC 9754 supports Gen 5 with 128 lanes (CPU only), while the Intel Xeon 6980P supports Gen 5 with 96 lanes (CPU only). AMD provides 32 additional lanes.

Q: What does the average benchmark score say about overall performance?

A: The AMD EPYC 9754 has an average benchmark score of 364371, while the Intel Xeon 6980P averages 251516. The AMD part also sits at the 100th percentile versus all CPUs, while the Xeon sits at the 99th.

Architecture Differences

The AMD EPYC 9754 is built on the Zen 4c architecture under the Bergamo codename, part of the EPYC 9004 series, and manufactured on a 5 nm process at TSMC. The Intel Xeon 6980P uses the Granite Rapids architecture, part of the Xeon 6 family, also on a 5 nm node but fabricated by Intel itself.

The physical designs diverge sharply. The EPYC 9754 uses eight separate chiplets, each 73 mm², totaling 71,000 million transistors. The Xeon 6980P uses three large dies of 598 mm² each, and the database records no transistor count for it. That difference in chiplet strategy reflects different approaches to scaling: AMD spreads across many small dies, Intel concentrates into fewer, larger ones.

Cache hierarchies differ per core. The EPYC 9754 has 64 KB of L1 per core and 1 MB of L2 per core, while the Xeon 6980P has 112 KB of L1 per core and 2 MB of L2 per core. Intel's per-core L1 and L2 are both larger. However, the shared L3 tells a different story: AMD has 256 MB shared, Intel has 504 MB shared, so Intel's aggregate L3 capacity is nearly double.

Both platforms use DDR5 memory with a twelve-channel memory bus, but the bandwidth figures differ meaningfully. The Xeon 6980P records 614.4 GB/s of memory bandwidth versus 460.8 GB/s for the EPYC 9754, a 33% advantage for Intel. Both support ECC memory, and neither has an unlocked multiplier. The Xeon reports integrated graphics as "N/A", while the EPYC has no integrated graphics listed.

Socket and platform requirements are entirely different. AMD uses Socket SP5, Intel uses Socket 7529. The EPYC 9754 was released on 2023-06-12, while the Xeon 6980P arrived on 2024-09-23, so Intel's part is over a year newer.

Head-to-Head Benchmarks

The head-to-head results are one-sided. Across all 17 recorded benchmark comparisons, the AMD EPYC 9754 wins every single one. The Xeon 6980P does not claim a single victory in the database's head-to-head set.

The Cinebench suite shows a consistent pattern. In Cinebench R15 multicore, AMD scores 8460 versus Intel's 6367, a 32.9% delta. Single-core R15 shows 1194 versus 898, a 33% edge. R20 multicore repeats the 32.9% delta at 35254 versus 26533, and R20 single-core also lands at 32.9% with 4977 versus 3745. R23 multicore follows suit: 83939 versus 63175, again 32.9%. R23 single-core: 11850 versus 8918, 32.9%. The consistency across every Cinebench generation indicates a steady architectural advantage rather than a workload-specific quirk.

PassMark results widen the gap in several areas. The biggest delta is in physics, where AMD scores 8793 versus Intel's 3350, a 162.5% advantage. Data encryption shows an 85.1% lead for AMD: 231891 versus 125246. Integer math is also lopsided at 61.1%: 1026896 versus 637476. Data compression comes in at 50.5% with 3558043 versus 2364519.

Some workloads are closer. Extended instructions show only a 4.4% lead for AMD: 224322 versus 214794. Prime number finding is an 8.8% edge: 604 versus 555. Floating-point math is a 17.2% lead: 588187 versus 501720. Random string sorting shows a 27.3% gap: 306481 versus 240792. Multithreaded PassMark is 32.9%: 98752 versus 74324.

The single-thread PassMark tests repeat the Cinebench story. AMD scores 2328 versus Intel's 1681, a 38.5% lead, recorded twice in the database under two test names. This means AMD's advantage is not merely about core scaling; even on a per-thread basis, the EPYC 9754 is clearly ahead in these measurements.

The closest competitor data also provides context. The EPYC 9754's nearest rival is the Intel Xeon 6960P, with a delta of -0.2%, meaning the 9754 is essentially tied with that chip on average score. The EPYC 9655 sits 2.4% higher, the EPYC 9535 sits 4% higher, and the EPYC 9475F sits 3.8% lower. For the Xeon 6980P, its nearest rival is the AMD EPYC 9634, which is 3% higher, and the Intel Xeon 6747P is 5.6% lower. The AMD EPYC 9684X is 5.8% higher, and the AMD Ryzen Threadripper 9970X is 10.1% higher.

The Verdict

The data points to a clear winner for compute-heavy workloads. The AMD EPYC 9754 wins all 17 head-to-head benchmark comparisons and holds a 30% or greater lead in the majority of them. Its average benchmark score of 364371 versus 251516 places it at the 100th percentile, while the Xeon 6980P sits at the 99th.

The Xeon 6980P does hold advantages in specific hardware capabilities. It has more L3 cache (504 MB versus 256 MB), higher memory bandwidth (614.4 GB/s versus 460.8 GB/s), larger per-core L1 and L2 caches, and a higher boost clock (3.90 GHz versus 3.10 GHz). The EPYC 9754 counters with a higher base clock (2.25 GHz versus 2.00 GHz), more PCIe lanes (128 versus 96), and a substantially lower TDP (360 watts versus 500 watts).

None of Intel's hardware advantages translate into benchmark wins in the recorded data. The higher boost clock and larger caches do not overcome AMD's per-thread and multi-thread execution efficiency in these tests. The Xeon's memory bandwidth advantage may matter in specific memory-bound scenarios, but the available benchmark suite does not show it.

The verdict from the database is straightforward: for anyone choosing between these two on measured performance alone, the AMD EPYC 9754 is the superior part. The Xeon 6980P is not uncompetitive in every dimension, but it loses every recorded benchmark, often by large margins.

Specification Differences

The two processors differ on several recorded specifications.

• Base clock: AMD EPYC 9754 at 2.25 GHz, Intel Xeon 6980P at 2.00 GHz
• Boost clock: AMD at 3.10 GHz, Intel at 3.90 GHz
• TDP: AMD at 360 watts, Intel at 500 watts
• Socket: AMD Socket SP5 versus Intel Socket 7529
• Architecture: Zen 4c (Bergamo) versus Granite Rapids
• Process node: both 5 nm, but AMD at TSMC and Intel at Intel
• Transistors: AMD at 71,000 million, Intel not recorded
• Die size: AMD at 8x 73 mm², Intel at 3x 598 mm²
• L1 cache per core: AMD at 64 KB, Intel at 112 KB
• L2 cache per core: AMD at 1 MB, Intel at 2 MB
• L3 cache shared: AMD at 256 MB, Intel at 504 MB
• Memory bandwidth: AMD at 460.8 GB/s, Intel at 614.4 GB/s
• PCIe lanes: AMD at Gen 5, 128 lanes (CPU only), Intel at Gen 5, 96 lanes (CPU only)
• Integrated graphics: AMD none listed, Intel N/A
• Release date: AMD 2023-06-12, Intel 2024-09-23
• Part number: AMD 100-000001234, Intel SRPL2Q5SM
• Launch MSRP: AMD $11900, Intel $12460

Both have 128 cores, 256 threads, DDR5 memory support, a twelve-channel memory bus, ECC memory support, and are locked multipliers. Both are classified as Server/Workstation parts and are currently Active in production.

Where Each One Wins

The AMD EPYC 9754 wins in every recorded benchmark category. The largest margins are in physics (162.5% ahead), data encryption (85.1%), and integer math (61.1%). These are workloads that stress raw arithmetic throughput and parallel execution efficiency. Data compression also favors AMD heavily at 50.5%, which matters for database workloads, file servers, and analytics pipelines that move large volumes of data.

The Cinebench results, consistent at roughly 33% across all versions and both single and multi-threaded tests, indicate that AMD has a general-purpose execution advantage that shows up in rendering and content creation workloads as well. Even extended instructions, the closest category at 4.4%, still favors AMD.

The Intel Xeon 6980P has no benchmark wins in the database, but its hardware specifications point to where it could plausibly excel in unmeasured scenarios. The 614.4 GB/s memory bandwidth is 33% higher than AMD's 460.8 GB/s, which can benefit memory-bound workloads such as large in-memory databases, high-frequency trading, or certain scientific simulations. The 504 MB L3 cache, nearly double AMD's 256 MB, could help workloads with large working sets that fit in cache. The higher boost clock of 3.90 GHz versus 3.10 GHz gives Intel a clock-speed ceiling that could matter in lightly threaded tasks, though the recorded single-thread benchmarks do not reflect that advantage.

The practical guidance from the data is this: if the workload is represented by the benchmark suite, choose the EPYC 9754. If the workload is heavily dependent on memory bandwidth or massive cache capacity, the Xeon 6980P's specifications justify consideration, but the measured results do not confirm an advantage. For general compute, virtualization, encryption, compression, and rendering, the AMD part is the pick.