Intel Xeon 6747P vs Intel Xeon 696X Comparison

Both processors in this comparison belong to Intel's Granite Rapids family, but they target noticeably different corners of the server and workstation market. The Intel Xeon 696X is a 64-core flagship with a 4.80 GHz boost clock, while the Intel Xeon 6747P is a 48-core part with a lower 3.90 GHz boost. The recorded benchmarks show the 696X winning 12 of 14 head-to-head tests, but the 6747P takes two decisive victories in workloads where its higher base clock and different core configuration shine. This is not a simple "bigger is better" story; the data reveals a clear split between raw throughput, where the 696X dominates, and latency-sensitive or physics-based tasks, where the 6747P is the better tool.

Where Each One Wins

The Intel Xeon 696X is the clear winner for any workload that scales with core count and shared cache. Its 64 cores and 128 threads give it a massive parallel advantage over the 48-core, 96-thread 6747P. The benchmark data confirms this: the 696X leads in every Cinebench multi-core test, all PassMark math and encryption tests, and the multithread score. For database compression, encryption, floating-point math, and integer math, the 696X posts wins between 21.3% and 23.9% over the 6747P. This makes it the obvious choice for heavy simulation, financial modeling, or any enterprise workload that can fully utilize 128 threads.

The 6747P wins in exactly two areas: prime number finding and the PassMark physics test. The physics score is the most dramatic result in the entire comparison. The 6747P scores 13398 versus 3382 for the 696X, a 74.8% advantage. That is not a small edge, it is a category-level difference. The prime number test also favors the 6747P by 25.9%, with scores of 1151 versus 853. These tests are more sensitive to per-core efficiency and clock speed at lower thread counts, and the 6747P's 2.70 GHz base clock versus 2.40 GHz on the 696X helps explain the outcome. For single-threaded integer workloads, the 696X still wins by 15.6% thanks to its higher boost clock, but the 6747P's sustained base frequency gives it an edge in certain deterministic loops.

Architecture Differences

Both chips are built on the same Granite Rapids architecture, use a 5 nm process, and share the same dual-die design with a combined die size of 2x 598 mm². They also share the same per-core L1 cache at 112 KB and L2 cache at 2 MB. The differences start with core count and cache size. The 696X has 64 cores and a 336 MB shared L3 cache, while the 6747P has 48 cores and a 288 MB shared L3 cache. That is a 48 MB difference in L3, which matters for working sets that exceed the smaller chip's capacity.

The sockets are identical, both use Intel Socket 4710, but the PCIe lane allocation differs significantly. The 696X provides Gen 5 with 128 lanes, while the 6747P offers Gen 5 with 88 lanes. This is a 40-lane gap that can impact systems with multiple GPUs or high-speed storage arrays. Both support DDR5 memory with eight-channel buses and identical 409.6 GB/s memory bandwidth, so memory throughput is not a differentiator. ECC memory is supported on both, as expected for server parts.

The 696X has an unlocked multiplier, making it a more flexible part for overclocking or fine-tuning, while the 6747P is locked. The 696X also carries a higher TDP of 350 watts versus 330 watts for the 6747P, reflecting the extra cores and higher boost clock. The release dates differ by roughly a year: the 6747P launched on 2025-02-23, while the 696X arrived on 2026-02-01. Both are listed as active production parts.

Head-to-Head Benchmarks

The Cinebench suite shows a consistent pattern. In R15 multi-core, the 696X scores 8994 against 8712 for the 6747P, a 3.2% lead. R20 shows 37475 versus 36301, again 3.2%. R23 follows the exact same trajectory: 89227 versus 86432, 3.2%. The consistency across all three Cinebench versions suggests a fixed multi-core scaling advantage rather than workload-specific variance. The PassMark multithread score also lands at 3.2% in favor of the 696X, with 104974 against 101685.

The gap widens dramatically in data-centric workloads. Data compression sees the 696X at 2264907 versus 1833378, a 23.5% win. Data encryption is similar: 112529 versus 90789, a 23.9% edge. Extended instructions show 172975 versus 142557, a 21.3% advantage. Floating-point math goes 450164 to 365904, a 23% lead. Integer math is 572072 versus 468518, a 22.1% margin. These are the workloads where the extra 16 cores and larger L3 cache translate directly into throughput.

Single-thread performance favors the 696X by 15.6%, with scores of 3742 versus 3236. This reflects its 4.80 GHz boost clock versus 3.90 GHz on the 6747P. Random string sorting is a near-tie: 180392 versus 180382, with a 0% delta in favor of the 696X. That is effectively identical performance, so any difference in this workload is negligible.

The two 6747P wins are outliers in the otherwise lopsided data. The physics test is the standout: 13398 versus 3382, a 74.8% margin. This is not a small win, it is a dominant one. Prime number finding also goes to the 6747P: 1151 versus 853, a 25.9% edge. These results suggest that the 6747P has a per-core advantage in certain sequential or semi-parallel workloads, likely tied to its higher base clock and lower thread count reducing contention.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon 696X has 64 cores and 128 threads, while the Intel Xeon 6747P has 48 cores and 96 threads.

Q: What is the difference in boost clock speeds?

A: The 696X boosts to 4.80 GHz, while the 6747P boosts to 3.90 GHz. The 6747P has a higher base clock at 2.70 GHz versus 2.40 GHz on the 696X.

Q: How do the processors compare in single-thread performance?

A: The 696X wins by 15.6% in the PassMark single-thread test, scoring 3742 versus 3236 for the 6747P.

Q: Which processor performs better in Cinebench R23 multi-core?

A: The 696X scores 89227, which is 3.2% higher than the 6747P's 86432.

Q: Why does the 6747P win the physics test?

A: The 6747P scores 13398 versus 3382 for the 696X, a 74.8% advantage. This likely stems from its higher base clock and lower thread count, which can reduce contention in this specific workload.

Q: Do both processors support the same memory configuration?

A: Yes, both support DDR5 with eight-channel memory buses and 409.6 GB/s bandwidth. Both also support ECC memory.

The Verdict

The data makes a straightforward recommendation for most buyers. The Intel Xeon 696X is the superior all-around processor. It wins 12 of 14 benchmarks, posts double-digit leads in encryption, compression, floating-point math, and integer math, and delivers a 15.6% single-thread advantage. For any workload that benefits from 64 cores, 128 threads, and a 336 MB L3 cache, this is the clear pick. Its unlocked multiplier also makes it more attractive to users who want to extract additional performance beyond stock settings.

The Intel Xeon 6747P is not a bad processor, it is just a narrower one. Its 74.8% win in the physics test is remarkable, and its 25.9% lead in prime number finding shows that it handles certain per-core tasks more efficiently. For users running physics simulations or similar workloads, the 6747P could be the better choice despite having fewer cores. However, those wins are isolated. In every other measured workload, the 696X is ahead, often by a wide margin.

The average benchmark scores tell the same story. The 696X sits at 286102 average, while the 6747P averages 238263. The 696X sits within 0.3% of the AMD EPYC 9555P and 0.2% ahead of the AMD EPYC 9565, placing it at the top of the performance pyramid. The 6747P, meanwhile, is 2.5% behind the AMD EPYC 9634 and 5.3% behind the Intel Xeon 6980P. The 696X is the flagship, and the benchmark data confirms it. The 6747P is a capable mid-range server part with a specific niche, but it is not a replacement for the 696X in most scenarios. Choose the 696X for maximum throughput, choose the 6747P only if the physics and prime number workloads are your primary concern.