AMD EPYC 9754 vs Intel Xeon 6960P Comparison

The Intel Xeon 6960P and AMD EPYC 9754 are both flagship server processors, but the benchmark data reveals they are optimized for entirely different workloads. The Xeon 6960P wins 9 of 14 head-to-head tests, including every Cinebench multi-core test and several PassMark CPU tests, while the EPYC 9754 dominates in data compression, encryption, and math-heavy integer workloads. The overall average benchmark scores are nearly identical, with the Intel part at 365,194 and the AMD part at 364,371, a negligible 0.2% difference. However, the distribution of wins is stark: the Xeon 6960P is the choice for single-threaded performance, physics simulations, and prime number finding, while the EPYC 9754 is superior for data processing and encryption tasks. The EPYC 9754 offers more cores (128 vs 72) and threads (256 vs 144), but the Xeon 6960P compensates with higher clock speeds (2.70 GHz base and 3.90 GHz boost vs 2.25 GHz and 3.10 GHz) and a larger L3 cache (432 MB vs 256 MB). The verdict is clear: pick the Xeon 6960P for latency-sensitive, single-threaded tasks and physics workloads, and pick the EPYC 9754 for massive parallel data throughput, encryption, and integer math.

The Verdict

The data points to a clear split: the Intel Xeon 6960P is the superior part for tasks that depend on high clock speeds and strong single-core execution. Its PassMark single-thread score of 3,287 is 41.2% higher than the EPYC 9754’s 2,328, and it leads by 32.3% in all three Cinebench multicore tests (R15: 11,194 vs 8,460; R20: 46,645 vs 35,254; R23: 111,060 vs 83,939). This advantage is amplified in the PassMark physics test, where the Xeon scores 24,937 versus just 8,793 for the AMD part, a 183.6% margin. The Xeon also achieves a 145.7% lead in the find prime numbers test (1,484 vs 604), showing a massive edge in single-threaded integer loops.

The AMD EPYC 9754 is the throughput king for data-centric workloads. It wins the PassMark data compression test with 3,558,043 points versus 2,797,724 for Intel, a 21.4% advantage. Its data encryption score of 231,891 beats the Xeon’s 162,013 by 30.1%. The EPYC also leads in integer math (1,026,896 vs 727,750, a 29.1% margin) and floating-point math (588,187 vs 527,473, a 10.3% margin). For users running database compression, secure data handling, or heavy integer computation, the EPYC 9754’s 128 cores and 256 threads provide a clear performance benefit that the Xeon’s higher clocks cannot overcome.

The 0.2% overall average score difference between the two (365,194 vs 364,371) is statistically meaningless. The real story is workload specificity. The Xeon 6960P is the winner for engineering simulation, physics modeling, and any application where a single thread must run fast. The EPYC 9754 is the winner for big-data processing, encryption, and parallel integer crunching.

FAQ

Q: Which processor has more cores and threads?

A: The AMD EPYC 9754 has 128 cores and 256 threads, compared to the Intel Xeon 6960P’s 72 cores and 144 threads.

Q: What is the single-thread performance difference?

A: The Intel Xeon 6960P scores 3,287 in PassMark single-thread, which is 41.2% higher than the EPYC 9754’s 2,328.

Q: Which CPU is faster in Cinebench R23 multi-core?

A: The Intel Xeon 6960P scores 111,060, which is 32.3% higher than the EPYC 9754’s 83,939.

Q: Which CPU handles encryption better?

A: The AMD EPYC 9754 wins PassMark data encryption with 231,891 points, beating the Xeon 6960P’s 162,013 by 30.1%.

Q: What is the memory bandwidth difference?

A: The Intel Xeon 6960P has a twelve-channel memory bus with 614.4 GB/s bandwidth, while the AMD EPYC 9754 also uses twelve channels but offers 460.8 GB/s.

Q: Which CPU has a larger L3 cache?

A: The Intel Xeon 6960P has 432 MB of shared L3 cache, while the AMD EPYC 9754 has 256 MB.

Architecture Differences

The two processors are built on fundamentally different architectural philosophies. The Intel Xeon 6960P uses the Granite Rapids architecture, fabricated on Intel’s 5 nm process with a die size of 3x 598 mm². Each of its 72 cores has 112 KB of L1 cache and 2 MB of L2 cache, with a shared 432 MB L3 cache. The AMD EPYC 9754 uses the Zen 4c architecture (codenamed Bergamo), built on TSMC’s 5 nm process with 8x 73 mm² chiplets and 71,000 million transistors. Its 128 cores each have 64 KB of L1 and 1 MB of L2, with a smaller shared 256 MB L3 cache.

The Xeon 6960P features per-core L1 and L2 caches that are 75% and 100% larger than the EPYC’s, respectively (112 KB vs 64 KB L1, 2 MB vs 1 MB L2). This larger per-core cache, combined with 432 MB of L3, gives Intel a significant latency advantage for workloads that reuse data. The EPYC 9754, by contrast, uses a denser core design (Zen 4c) that allows 128 cores in the same power envelope, but with reduced cache per core. The Xeon also has a higher TDP of 500 watts versus the EPYC’s 360 watts, reflecting its higher clock speeds and larger cache.

Both processors support DDR5 memory with twelve-channel buses and ECC, but the Xeon 6960P offers 614.4 GB/s of memory bandwidth versus 460.8 GB/s for the EPYC. The EPYC 9754 provides 128 PCIe Gen 5 lanes (CPU only), while the Xeon 6960P offers 96 lanes. The Xeon 6960P was released on 2024-09-23, while the EPYC 9754 launched earlier on 2023-06-12.

Specification Differences

The core count is the most obvious differentiator: the EPYC 9754 has 128 cores and 256 threads, while the Xeon 6960P has 72 cores and 144 threads. Clock speeds favor Intel: the Xeon has a 2.70 GHz base and 3.90 GHz boost, versus the EPYC’s 2.25 GHz base and 3.10 GHz boost. The Xeon’s TDP is 500 watts, considerably higher than the EPYC’s 360 watts.

Cache distribution differs significantly. The Xeon allocates 112 KB L1 and 2 MB L2 per core, while the EPYC uses 64 KB L1 and 1 MB L2 per core. Total L3 cache is 432 MB for Intel and 256 MB for AMD. Memory bandwidth is 614.4 GB/s for the Xeon versus 460.8 GB/s for the EPYC, despite both using twelve-channel DDR5. PCIe lanes favor AMD with 128 Gen 5 lanes versus Intel’s 96. The Xeon uses Intel Socket 7529, while the EPYC uses AMD Socket SP5. The Xeon’s launch MSRP is $9625, while the EPYC’s is $11900.

Head-to-Head Benchmarks

The Intel Xeon 6960P delivers a clean sweep in Cinebench, winning all three multi-core tests by an identical 32.3% margin. In Cinebench R15, it scores 11,194 against the EPYC’s 8,460. In R20, the scores are 46,645 versus 35,254, and in R23, 111,060 versus 83,939. This consistency suggests a fundamental architectural advantage in rendering workloads.

The PassMark suite reveals the workload split. The Xeon wins the multithread test with 130,659 versus 98,752, again by 32.3%. Its most dramatic win is in physics, where it scores 24,937 versus 8,793 for the EPYC, a 183.6% advantage. The Xeon also wins random string sorting (371,795 vs 306,481, a 21.3% lead) and find prime numbers (1,484 vs 604, a 145.7% lead). Single-thread performance goes to Intel with 3,287 versus 2,328, a 41.2% margin.

The AMD EPYC 9754 fights back in data-centric benchmarks. It wins data compression with 3,558,043 versus 2,797,724, a 21.4% lead. Data encryption goes to AMD with 231,891 versus 162,013, a 30.1% margin. The EPYC also wins integer math (1,026,896 vs 727,750, 29.1% ahead) and floating-point math (588,187 vs 527,473, 10.3% ahead). Extended instructions also favor AMD: 224,322 versus 193,404, a 13.8% lead.

Where Each One Wins

The Intel Xeon 6960P wins in scenarios requiring high clock speeds and per-core cache. Its 41.2% single-thread lead and 183.6% physics performance advantage make it ideal for scientific simulation, real-time physics engines, and any workload that cannot be easily parallelized. The 145.7% win in prime number finding indicates strong integer loop performance. The 32.3% Cinebench lead shows it is superior for 3D rendering tasks that rely on multi-core scaling with high per-core efficiency. The Xeon’s larger L2 cache (2 MB per core) and 432 MB L3 cache, plus 614.4 GB/s memory bandwidth, make it the choice for latency-sensitive data retrieval.

The AMD EPYC 9754 wins in high-throughput, parallel data processing. Its 30.1% encryption lead and 21.4% compression lead make it the superior choice for database workloads, secure data transfer, and archive management. The 29.1% integer math win means it excels in financial modeling, scientific computing with large integer arrays, and code compilation. The 10.3% floating-point win and 13.8% extended instructions win indicate strength in scientific computation that uses vectorized operations. With 128 cores and 256 threads, the EPYC 9754 is the better processor for virtualized environments and massive multi-tenant cloud workloads, despite its lower clock speeds and smaller cache. The 128 PCIe lanes also give it an edge in systems with many high-speed peripherals or GPUs.