AMD EPYC 9734 vs Intel Xeon 6980P Comparison

Head-to-Head Benchmarks

The benchmark data presents a remarkably one-sided picture. Across the recorded head-to-head tests, the AMD EPYC 9734 wins 16 of 17 comparisons, with the Intel Xeon 6980P taking only a single narrow victory. The margin in most workloads is substantial, often exceeding 37%.

In Cinebench, the entire R15, R20, and R23 suites show consistent dominance for the EPYC 9734. Multi-core scores run 37.6% higher across all three versions: 8763 vs 6367 in R15, 36516 vs 26533 in R20, and 86943 vs 63175 in R23. Single-core results follow the same pattern, with the EPYC 9734 ahead by 37.8% in R15 (1237 vs 898), 37.7% in R20 (5155 vs 3745), and 37.6% in R23 (12274 vs 8918). This uniformity across every Cinebench iteration suggests a fundamental per-thread performance advantage, not just a core-count effect.

PassMark tests reveal where the EPYC 9734 excels most dramatically. The physics test shows the largest gap: 6747 vs 3350, a 101.4% advantage. That is more than double the Xeon's result. Prime number finding also favors AMD heavily, with 829 vs 555, a 49.4% delta. Random string sorting follows at 48.5% (357638 vs 240792), and data encryption at 43.2% (179390 vs 125246). Integer math shows a 29.1% lead (823150 vs 637476), while data compression is 22.6% ahead (2900008 vs 2364519). Floating-point math is closer but still favors AMD: 549045 vs 501720, a 9.4% margin. Multithread and single-thread aggregate scores both land at 37.6% and 37.4% respectively, reinforcing the consistent per-core and per-thread gap.

The Intel Xeon 6980P's sole win comes in PassMark extended instructions, where it scores 214794 vs 205925, a 4.1% advantage. This is a narrow margin in a specialized workload category, and it does little to offset the broader trend.

Looking at aggregate standings, the EPYC 9734 posts an average benchmark score of 310619, placing it in the 99th percentile of all CPUs. Its nearest rivals include the AMD EPYC 9575F (311774, just 0.4% behind), the Intel Xeon 6781P (315524, 1.6% ahead), and the AMD Ryzen Threadripper PRO 9985WX (320749, 3.2% ahead). The Xeon 6980P, by contrast, averages 251516, also in the 99th percentile, with its closest competitor being the AMD EPYC 9634 at 244274 (3% behind) and the Intel Xeon 6747P at 238263 (5.6% behind). The gap between the two processors in average score is roughly 19%, a meaningful separation at this performance tier.

Where Each One Wins

The EPYC 9734 dominates nearly every measured category, but the scale of its wins varies by workload type. In physics simulation, the advantage is overwhelming, doubling the Xeon's output. This suggests workloads that stress floating-point operations and memory latency benefit enormously from the AMD design. Prime number finding and random string sorting, both heavily dependent on raw integer throughput and cache behavior, also show massive leads above 48%.

Data encryption shows a 43.2% edge for AMD, which matters for database encryption, secure connections, and any always-on cryptographic workload. Integer math, compression, and multithread aggregate scores all sit in the 22% to 29% range, indicating broad competence across general server tasks. The Cinebench results, both multi-core and single-core, hover consistently around 37.6%, which points to a uniform architectural advantage in rendering and 3D workloads.

The Intel Xeon 6980P has exactly one niche: extended instruction sets. Its 4.1% lead in PassMark extended instructions suggests that AVX-512 or similar vectorized instruction paths execute slightly better on the Granite Rapids architecture. This could matter for specific scientific computing or media encoding pipelines that leverage those instructions heavily. But the margin is small, and it is the only area where Intel holds any measured advantage.

For everything else, from general server consolidation to high-performance computing simulation, the EPYC 9734 is the clear benchmark leader. The data shows no other category where the Xeon 6980P closes the gap to within single digits except floating-point math, where it still trails by 9.4%.

The Verdict

The benchmark data is unambiguous. The AMD EPYC 9734 outperforms the Intel Xeon 6980P in 16 of 17 recorded tests, with margins ranging from 9.4% to 101.4%. The average benchmark score difference is substantial: 310619 vs 251516, a gap of roughly 19%. If your workload falls into any of the tested categories except extended instruction execution, the EPYC 9734 is the stronger choice.

The Xeon 6980P should only be considered if your application specifically depends on extended instruction throughput, where it holds a 4.1% advantage. Even then, the rest of the performance profile is so heavily tilted toward AMD that the trade-off is difficult to justify on benchmark evidence alone. The EPYC 9734 also offers 112 cores and 224 threads, while the Xeon provides 128 cores and 256 threads, yet the AMD chip still wins most multithreaded tests. That means the per-core efficiency of the Zen 4c design is substantially higher than Granite Rapids in these measurements.

For server buyers prioritizing raw compute density, rendering throughput, encryption, compression, or physics simulation, the data points squarely to the EPYC 9734. The Xeon 6980P is not uncompetitive in absolute terms, it sits in the 99th percentile of all CPUs, but in a direct head-to-head it loses on nearly every metric.

FAQ

Q: Which processor wins in Cinebench R23 multi-core?

A: The AMD EPYC 9734 scores 86943 versus the Intel Xeon 6980P's 63175, a 37.6% advantage for AMD.

Q: Is there any workload where the Intel Xeon 6980P beats the AMD EPYC 9734?

A: Yes, in PassMark extended instructions, the Xeon scores 214794 versus 205925 for the EPYC, a 4.1% lead.

Q: How large is the gap in physics performance?

A: The EPYC 9734 scores 6747 in PassMark physics, more than double the Xeon's 3350, representing a 101.4% advantage.

Q: What are the aggregate benchmark scores for each CPU?

A: The EPYC 9734 averages 310619 across all benchmarks, while the Xeon 6980P averages 251516. Both sit in the 99th percentile of all CPUs.

Q: Does the Intel Xeon 6980P's higher core count help in multithreaded tests?

A: No. Despite having 128 cores versus 112 for the EPYC, the Xeon scores 74324 in PassMark multithread, well below the EPYC's 102286, a 37.6% deficit.

Q: How close is the EPYC 9734 to its nearest rival in the database?

A: The AMD EPYC 9575F averages 311774, just 0.4% higher, making it the closest competitor. The Intel Xeon 6781P is 1.6% ahead, and the Ryzen Threadripper PRO 9985WX is 3.2% ahead.

Architecture Differences

The two processors come from fundamentally different design philosophies. The AMD EPYC 9734 uses the Zen 4c architecture, codenamed Bergamo, built on a 5 nm process at TSMC. It packs 112 cores and 224 threads, with a transistor count of 71,000 million spread across 8 dies of 73 mm² each. The Intel Xeon 6980P uses the Granite Rapids architecture, also on a 5 nm process but fabricated by Intel, with 128 cores and 256 threads across 3 dies of 598 mm² each.

Cache hierarchies differ significantly. The EPYC 9734 provides 64 KB of L1 cache per core and 1 MB of L2 per core, with 256 MB of shared L3. The Xeon 6980P offers 112 KB of L1 per core and 2 MB of L2 per core, with a much larger 504 MB of shared L3. Despite the Xeon's larger cache, the benchmark results show the EPYC winning in cache-sensitive tests like prime number finding and random string sorting, suggesting cache capacity alone does not dictate performance.

Memory subsystems are both DDR5 with twelve-channel buses, but bandwidth differs. The EPYC 9734 reaches 460.8 GB/s, while the Xeon 6980P reaches 614.4 GB/s. The Xeon offers 33% more theoretical memory bandwidth, yet the EPYC still wins in memory-heavy workloads like physics and floating-point math. This indicates that memory bandwidth is not the limiting factor in these tests.

PCIe connectivity also differs. The EPYC 9734 supports Gen 5 with 128 lanes from the CPU, while the Xeon 6980P provides Gen 5 with 96 lanes. For systems requiring maximum expansion, the AMD part offers 32 additional lanes.

The Xeon 6980P has an integrated graphics designation of N/A, while the EPYC 9734 lists no integrated graphics at all. Both are server/workstation parts with active production status. The EPYC 9734 launched on 2023-06-12 with a launch MSRP of $9600, while the Xeon 6980P launched on 2024-09-23 with a launch MSRP of $12460.

Specification Differences

The core and thread counts favor Intel: 128 cores and 256 threads versus 112 cores and 224 threads. Clock speeds favor AMD in base frequency at 2.20 GHz versus 2.00 GHz, but Intel's boost clock is higher at 3.90 GHz versus 3.00 GHz. Thermal design power favors AMD at 340 W versus 500 W for Intel, a significant difference for cooling and power delivery requirements.

Socket compatibility separates the platforms completely. The EPYC 9734 uses AMD Socket SP5, while the Xeon 6980P uses Intel Socket 7529. These are not interchangeable, so motherboard selection is determined entirely by processor choice.

L1 and L2 cache per core differ, with Intel providing larger per-core caches: 112 KB L1 and 2 MB L2 versus AMD's 64 KB L1 and 1 MB L2. Shared L3 cache also favors Intel at 504 MB versus 256 MB. Memory bandwidth favors Intel at 614.4 GB/s versus 460.8 GB/s. PCIe lanes favor AMD at 128 versus 96. Both support DDR5 with ECC and twelve-channel memory buses.

The EPYC 9734 uses 8 dies of 73 mm² each, while the Xeon 6980P uses 3 dies of 598 mm² each. The AMD part lists a transistor count of 71,000 million, while the Intel part does not report a transistor figure in the database. The EPYC 9734 is built at TSMC, the Xeon 6980P at Intel's own fabs, both on 5 nm processes.