Intel Xeon 6980P vs Intel Xeon w9-3595X Comparison

The Intel Xeon 6980P and Intel Xeon w9-3595X represent two distinct interpretations of high-core-count x86 performance, and the benchmark data separates them cleanly by workload type. The Xeon 6980P, a Granite Rapids-AP part with 128 cores and 256 threads, dominates in data-heavy and encryption tasks, while the Xeon w9-3595X, a Sapphire Rapids workstation chip with 60 cores and 120 threads, wins decisively in single-threaded and latency-sensitive tests. The w9-3595X takes 10 of the 16 head-to-head benchmarks, but the 6980P claims the largest individual victories in raw compute throughput. Neither processor is a universal champion; the choice hinges entirely on whether the target workload scales across many cores or depends on per-core speed and memory latency. The data shows the w9-3595X is the better pick for interactive workstation use and lightly threaded applications, while the 6980P is the stronger option for server-side data processing and parallel floating-point workloads.

The Verdict

The Intel Xeon w9-3595X is the clear winner for single-threaded and lightly threaded performance. Its PassMark single-thread score of 3720 is 54.8% higher than the 6980P’s 1681, a margin that dwarfs any other difference between the two chips. This advantage extends to Cinebench single-core tests, where the w9-3595X leads by 25.1% across R15, R20, and R23. For users running CAD, simulation pre/post-processing, or any application that relies on a few fast cores, the w9-3595X is the data-backed choice. Its PassMark physics score of 5842 is also 42.7% ahead of the 6980P’s 3350, confirming stronger per-core execution in real-time physics calculations.

The Intel Xeon 6980P is the pick for throughput-oriented server workloads. Its PassMark data compression score of 2,364,519 is 29.1% higher than the w9-3595X’s 1,831,962, and its data encryption score of 125,246 beats 92,249 by 35.8%. The 6980P also wins extended instructions by 50.4%, floating-point math by 32.4%, and integer math by 34.6%. These are precisely the metrics that matter for database servers, virtualization hosts, and scientific computing clusters. The 6980P’s 128 cores and 504 MB of shared L3 cache provide the parallel resources needed to sustain these workloads, while the w9-3595X’s 60 cores and 112.5 MB L3 cache hit a ceiling.

The verdict is not subtle: the w9-3595X is a workstation part that excels at interactive and single-threaded tasks, while the 6980P is a server part built for massive parallel throughput. The w9-3595X’s average benchmark score of 209,881 sits below the 6980P’s 251,516, but that aggregate figure masks the w9-3595X’s dominance in the categories that matter most for desktop-class workflows. The 6980P’s 99th percentile ranking among all CPUs is matched by the w9-3595X, so both are elite parts; the differentiator is workload fit, not raw capability.

Where Each One Wins

The w9-3595X wins every Cinebench test, regardless of core count. In R15 multicore, it scores 8,497 against 6,367 for the 6980P, a 25.1% lead. The same delta appears in R20 multicore (35,407 vs 26,533) and R23 multicore (84,304 vs 63,175). This is counterintuitive given the 6980P has more than double the cores, but the w9-3595X’s higher boost clock of 4.80 GHz versus 3.90 GHz and its larger per-core L1 cache (80 KB vs 112 KB per core, though the 6980P has more total cache) contribute to better scaling in Cinebench’s rendering workload. The w9-3595X also wins PassMark multithread (99,576 vs 74,324, a 25.4% lead) and PassMark find prime numbers (580 vs 555, a 4.3% edge), suggesting its architecture handles certain parallel tasks more efficiently despite fewer cores.

The 6980P wins all the PassMark data-processing tests. Its data compression score of 2,364,519 is 29.1% higher, data encryption is 35.8% higher, and random string sorting is 26.2% higher (240,792 vs 190,745). The 6980P’s floating-point math score of 501,720 beats 379,008 by 32.4%, and its integer math score of 637,476 beats 473,507 by 34.6%. Extended instructions go to the 6980P by a massive 50.4% (214,794 vs 142,785). These wins are all attributable to the 6980P’s core count advantage and its 614.4 GB/s memory bandwidth versus the w9-3595X’s 307.2 GB/s. Twelve-channel memory versus eight-channel memory gives the 6980P a decisive edge in bandwidth-hungry operations.

The split is clean: the w9-3595X wins on single-thread, Cinebench, and physics, while the 6980P wins on encryption, compression, sorting, and math. The only overlap is PassMark multithread, where the w9-3595X wins despite having fewer cores, which points to its superior per-core efficiency in that specific benchmark.

Architecture Differences

The two processors share Intel’s manufacturing but diverge on process node and physical design. The 6980P uses a 5 nm process with a die size of 3x 598 mm², while the w9-3595X uses a 10 nm process with a 4x 477 mm² die. The 6980P’s newer node allows for 128 cores and 256 threads, double the w9-3595X’s 60 cores and 120 threads. Both have a 2.00 GHz base clock, but the w9-3595X boosts to 4.80 GHz versus the 6980P’s 3.90 GHz, explaining the single-thread gap.

Cache configurations differ substantially. The 6980P offers 112 KB of L1 per core, 2 MB of L2 per core, and 504 MB of shared L3. The w9-3595X has 80 KB of L1 per core, 2 MB of L2 per core, and 112.5 MB of L3. The 6980P’s 504 MB L3 is nearly 4.5 times larger, which helps in data-heavy workloads but does not compensate for the w9-3595X’s clock speed advantage in latency-sensitive tasks.

Memory and I/O also separate the parts. The 6980P uses twelve-channel DDR5 with 614.4 GB/s of bandwidth, while the w9-3595X uses eight-channel DDR5 with 307.2 GB/s. Both support ECC memory. PCIe lane counts differ: the 6980P provides 96 Gen 5 lanes, while the w9-3595X provides 112 Gen 5 lanes. The w9-3595X has an unlocked multiplier, making it overclockable, whereas the 6980P is locked. Sockets are incompatible: the 6980P uses Intel Socket 7529, and the w9-3595X uses Intel Socket 4677. The 6980P’s TDP is 500 W, and the w9-3595X’s is 385 W, though the 6980P’s higher core count and memory bandwidth justify the additional power draw.

The release timeline shows the w9-3595X launched on 2024-08-23, one month before the 6980P on 2024-09-23. Both are active production parts without integrated graphics. The 6980P’s part number is SRPL2Q5SM, and the w9-3595X’s is SRN71.

FAQ

Q: Which processor has higher single-thread performance?

A: The Intel Xeon w9-3595X. Its PassMark single-thread score of 3720 is 54.8% higher than the 6980P’s 1681. In Cinebench R23 single-core, the w9-3595X scores 8,918 against 8,918 for the 6980P, a 25.1% lead.

Q: Which processor wins in multi-threaded Cinebench tests?

A: The w9-3595X wins all Cinebench multicore tests. In R23 multicore, it scores 84,304 versus 63,175 for the 6980P, a 25.1% advantage. This occurs despite the 6980P having 128 cores versus 60, due to the w9-3595X’s higher boost clock.

Q: Is the 6980P better for data encryption workloads?

A: Yes. The 6980P scores 125,246 in PassMark data encryption, which is 35.8% higher than the w9-3595X’s 92,249. The 6980P also wins data compression by 29.1% and random string sorting by 26.2%.

Q: Do both processors support ECC memory?

A: Yes, both the 6980P and w9-3595X support ECC memory. However, the 6980P uses twelve-channel DDR5 with 614.4 GB/s bandwidth, while the w9-3595X uses eight-channel DDR5 with 307.2 GB/s.

Q: Which processor has more PCIe lanes?

A: The w9-3595X has 112 Gen 5 lanes, while the 6980P has 96 Gen 5 lanes. Both are CPU-only lane counts.

Q: How do the average benchmark scores compare?

A: The 6980P has an average benchmark score of 251,516, which is 19.8% higher than the w9-3595X’s 209,881. Both rank in the 99th percentile of all CPUs.

Head-to-Head Benchmarks

The w9-3595X’s biggest win is in PassMark single-thread, where it scores 3,720 against the 6980P’s 1,681, a delta of 54.8%. This is the largest percentage gap in the entire comparison and underscores the w9-3595X’s clear advantage in any workload that cannot parallelize. The same pattern appears in PassMark physics, where the w9-3595X scores 5,842 versus 3,350, a 42.7% lead. Both results align with the w9-3595X’s boost clock of 4.80 GHz, which is 23.1% higher than the 6980P’s 3.90 GHz.

The w9-3595X also wins every Cinebench test by a consistent 25.1%. In R15 multicore, it scores 8,497 versus 6,367; in R20 multicore, 35,407 versus 26,533; and in R23 multicore, 84,304 versus 63,175. The single-core Cinebench results follow the same pattern: R15 at 1,199 versus 898, R20 at 4,998 versus 3,745, and R23 at 8,918 versus 8,918. The consistency of the 25.1% delta across all Cinebench versions suggests a systematic per-core performance difference rather than a workload-specific artifact.

PassMark multithread goes to the w9-3595X with a score of 99,576 versus 74,324, a 25.4% advantage. This is notable because the 6980P has 128 cores versus 60, yet the w9-3595X still wins, indicating that the w9-3595X’s per-core efficiency in this benchmark outweighs the core count disparity. The w9-3595X also wins PassMark find prime numbers, scoring 580 versus 555, a modest 4.3% lead.

The 6980P’s largest victory is in PassMark extended instructions, where it scores 214,794 versus 142,785, a 50.4% lead. This test likely leverages the 6980P’s 128 cores and 504 MB L3 cache to process large instruction sets in parallel. The 6980P also wins PassMark integer math by 34.6% (637,476 vs 473,507) and floating-point math by 32.4% (501,720 vs 379,008). Data compression goes to the 6980P at 2,364,519 versus 1,831,962, a 29.1% lead, and data encryption at 125,246 versus 92,249, a 35.8% lead. Random string sorting is won by the 6980P at 240,792 versus 190,745, a 26.2% margin.

The results show a clear dichotomy: the w9-3595X wins all 10 benchmarks that favor single-thread speed and Cinebench’s rendering model, while the 6980P wins all 6 benchmarks that favor memory bandwidth and parallel data processing. The 6980P’s wins are generally larger in percentage terms (average 34.8% across its six wins) compared to the w9-3595X’s wins (average 26.1% across its ten wins), but the w9-3595X wins more tests overall. The data suggests the 6980P is optimized for throughput, while the w9-3595X is optimized for responsiveness and per-core performance.