Intel Xeon 696X vs Intel Xeon 6980P Comparison

The Intel Xeon 696X and Intel Xeon 6980P are both high-end Granite Rapids server processors, but the data shows they are engineered for different extremes. The 696X combines a lower core count with a far higher boost clock and unlocked multiplier, while the 6980P doubles the core count and expands memory channels but at a much lower boost clock and a locked multiplier. Benchmark results reveal a clear division: the 696X dominates multi-threaded rendering and single-thread tests, while the 6980P leads in specific memory-bandwidth-sensitive and integer-heavy workloads.

FAQ

Q: Which processor has more cores?

A: The Intel Xeon 6980P has 128 cores and 256 threads, exactly double the 64 cores and 128 threads of the Intel Xeon 696X.

Q: How do they compare in single-thread performance?

A: The Intel Xeon 696X scores 3742 in PassMark single-thread, which is 122.6% higher than the 1681 scored by the Intel Xeon 6980P.

Q: Which chip wins in Cinebench multi-core benchmarks?

A: The Intel Xeon 696X wins all three Cinebench multi-core tests, with a 41.3% lead in R15, 41.2% in R20, and 41.2% in R23 over the Intel Xeon 6980P.

Q: Does the 6980P win any benchmark categories?

A: Yes, the 6980P wins 6 of 14 head-to-head tests, including PassMark data compression (2364519 vs 2264907), data encryption (125246 vs 112529), and integer math (637476 vs 572072).

Q: Are both processors on the same process node?

A: Yes, both are built on Intel's 5 nm process node, but the 696X uses a 2x 598 mm² die setup while the 6980P uses a 3x 598 mm² die setup.

Q: What is the memory bandwidth difference?

A: The Intel Xeon 6980P supports twelve-channel DDR5 memory with 614.4 GB/s bandwidth, while the 696X has eight-channel memory at 409.6 GB/s.

Architecture Differences

The two processors share the Granite Rapids architecture and 5 nm process node, but their physical and feature layouts diverge significantly. The Intel Xeon 696X is part of the Granite Rapids-WS generation and uses a dual-die design with 2x 598 mm², whereas the Intel Xeon 6980P belongs to the Granite Rapids-AP generation and uses three dies totaling 3x 598 mm². This extra die in the 6980P enables its 128-core configuration and 504 MB of shared L3 cache, compared to 64 cores and 336 MB of L3 on the 696X. Per-core L1 and L2 caches are identical at 112 KB and 2 MB respectively, but the total L3 capacity is 50% larger on the 6980P.

Clock speeds present a stark contrast. The 696X has a base clock of 2.40 GHz and a boost clock of 4.80 GHz, while the 6980P runs at 2.00 GHz base and 3.90 GHz boost. The 696X also features an unlocked multiplier, allowing overclocking, while the 6980P is locked. This clock advantage explains the 696X's dominance in single-threaded workloads, as the data shows a 122.6% lead in PassMark single-thread performance.

Memory architecture differs as well. The 696X uses an eight-channel DDR5 bus with a theoretical bandwidth of 409.6 GB/s, while the 6980P expands to a twelve-channel configuration delivering 614.4 GB/s. Both support ECC memory. PCIe lanes are another differentiator: the 696X provides 128 Gen 5 lanes (CPU only), while the 6980P offers 96 Gen 5 lanes. Socket compatibility is not shared, as the 696X uses Intel Socket 4710 and the 6980P uses Intel Socket 7529.

Power and physical specifications reflect the core-count gap. The 696X has a TDP of 350 W, while the 6980P draws 500 W. The release dates are also distinct, with the 6980P launching on 2024-09-23 and the 696X arriving later on 2026-02-01. The part numbers differ (SRWQ7 for the 696X, SRPL2Q5SM for the 6980P), confirming they are separate SKUs despite the shared architecture.

The Verdict

The data suggests a clear choice based on workload type. For multi-threaded rendering and general compute tasks, the Intel Xeon 696X is the stronger option. Its Cinebench R23 score of 89227 is 41.2% higher than the 6980P's 63175, and it also wins PassMark multithread by 41.2% (104974 vs 74324). The 696X also delivers superior single-thread performance with a 122.6% lead, making it more versatile for mixed workloads that include latency-sensitive threads.

The Intel Xeon 6980P is the pick for memory-bandwidth-bound or integer-heavy server workloads. Its twelve-channel memory bus and 504 MB L3 cache translate to wins in PassMark data compression (2364519 vs 2264907), data encryption (125246 vs 112529), and integer math (637476 vs 572072). It also leads in floating-point math by 10.3% and extended instructions by 19.5%. However, its lower clocks and locked multiplier make it less responsive in single-threaded tasks, as evidenced by the 1681 single-thread score versus 3742 for the 696X.

The 6980P also trails in prime number finding by a wide margin, scoring 555 versus 853 (a 53.7% deficit). Overall, the 696X wins 8 of 14 head-to-head benchmarks, while the 6980P wins 6. Neither chip is universally superior; the 696X suits high-frequency, rendering-focused deployments, while the 6980P targets scale-out, memory-heavy environments.

Specification Differences

| Specification | Intel Xeon 696X | Intel Xeon 6980P |

|----------------|-----------------|-------------------|

| Cores | 64 | 128 |

| Threads | 128 | 256 |

| Base Clock | 2.40 GHz | 2.00 GHz |

| Boost Clock | 4.80 GHz | 3.90 GHz |

| TDP | 350 W | 500 W |

| Socket | Intel Socket 4710 | Intel Socket 7529 |

| Generation | Xeon 600 (Granite Rapids-WS) | Xeon 6 (Granite Rapids-AP) |

| Die Size | 2x 598 mm² | 3x 598 mm² |

| L3 Cache | 336 MB (shared) | 504 MB (shared) |

| Memory Bus | Eight-channel | Twelve-channel |

| Memory Bandwidth | 409.6 GB/s | 614.4 GB/s |

| PCIe | Gen 5, 128 Lanes (CPU only) | Gen 5, 96 Lanes (CPU only) |

| Release Date | 2026-02-01 | 2024-09-23 |

| Multiplier Unlocked | Yes | No |

| Part Number | SRWQ7 | SRPL2Q5SM |

Head-to-Head Benchmarks

The most decisive victory for the Intel Xeon 696X is in PassMark single-thread, where it scores 3742 against the 6980P's 1681, a 122.6% advantage. This margin is reflected in the Cinebench single-core tests as well, with the 696X posting 898 in R15, 3745 in R20, and 8918 in R23, though those scores are only available for the 6980P in the data. The 696X also dominates multi-threaded Cinebench workloads: 8994 vs 6367 in R15, 37475 vs 26533 in R20, and 89227 vs 63175 in R23, each with a 41.2-41.3% delta. PassMark multithread follows the same pattern, with the 696X scoring 104974 versus 74324 for a 41.2% win. Prime number finding is another strong suit for the 696X, at 853 versus 555, a 53.7% lead. Physics performance is nearly identical, with the 696X edging out a 1% win (3382 vs 3350).

The Intel Xeon 6980P takes its biggest win in PassMark random string sorting, scoring 240792 against 180392 for a 25.1% advantage. Extended instructions show a 19.5% lead for the 6980P (214794 vs 172975). Data encryption favors the 6980P by 10.2% (125246 vs 112529), and both floating-point math and integer math show the 6980P ahead by 10.3% (501720 vs 450164 and 637476 vs 572072, respectively). Data compression is a narrower win for the 6980P, at 2364519 versus 2264907 for a 4.2% margin.

The average benchmark scores reflect the overall balance: the 696X has an average score of 286102, placing it 0.2% above the AMD EPYC 9565 and 2.3% above the AMD Ryzen Threadripper 9970X. The 6980P averages 251516, which is 3% above the AMD EPYC 9634 and 5.6% above the Intel Xeon 6747P, but 5.8% below the AMD EPYC 9684X and 10.1% below the Threadripper 9970X. Both chips sit in the 99th percentile of all CPUs, but their nearest rivals and deltaPct values show the 696X competing in a higher performance tier.