Intel Xeon 6745P vs Intel Xeon 678X Comparison

Head-to-Head Benchmarks

The benchmark data paints an unequivocal picture: the Intel Xeon 678X wins every single head-to-head comparison against the Intel Xeon 6745P, taking all 16 recorded tests. The margin of victory varies significantly by workload, ranging from a modest 8.9% in single-threaded tasks to a commanding 48.3% in prime number finding.

Starting with the Cinebench suite, the 678X demonstrates consistent multi-core dominance. In Cinebench R15 multi-core, it scores 8444 against the 6745P's 7214, a 17.1% advantage. The single-core R15 test shows the same 17.1% delta, with scores of 1192 and 1018 respectively. Moving to Cinebench R20, the multi-core gap narrows slightly to 17% (35185 vs 30062), while single-core also lands at 17% (4967 vs 4244). Cinebench R23 multi-core continues the pattern with another 17% lead, scoring 83775 versus 71578.

The Passmark suite reveals where the 678X truly stretches its legs. The largest margin comes in the find prime numbers test, where the 678X scores 1010 against just 681 for the 6745P, a staggering 48.3% delta. Floating point math also shows a substantial gap at 35.4%, with scores of 362070 and 267438. Extended instructions deliver a 30.6% advantage (141431 vs 108326), while physics testing shows a 27.1% lead (7809 vs 6144).

Data-centric workloads further reinforce the 678X's position. Data compression scores 1690896 against 1352801, a 25% advantage. Data encryption follows closely at 25.4%, with 83598 versus 66665. Random string sorting shows a 22.9% delta (164102 vs 133528), while integer math lands at 20.2% (405075 vs 336926). The Passmark multithread test shows a 17% lead (98559 vs 84210).

The narrowest margin comes in single-threaded performance. The Passmark single thread test shows the 678X at 3758 versus 3450 for the 6745P, an 8.9% advantage. This is notable because the 6745P actually has a higher base clock (3.10 GHz vs 2.40 GHz), yet the 678X still pulls ahead, likely due to its higher boost clock (4.90 GHz vs 4.30 GHz).

FAQ

Q: Which processor wins more benchmark tests?

A: The Intel Xeon 678X wins all 16 head-to-head benchmark comparisons against the Intel Xeon 6745P. The 6745P does not win a single recorded test.

Q: What is the biggest performance gap between the two?

A: The largest delta is in the Passmark find prime numbers test, where the 678X leads by 48.3%. The second-largest gap is in floating point math at 35.4%.

Q: How close is single-threaded performance?

A: Single-threaded performance is the closest category. The Passmark single thread test shows only an 8.9% advantage for the 678X, with scores of 3758 and 3450.

Q: Do both processors have the same power envelope?

A: Yes, both the Intel Xeon 678X and the Intel Xeon 6745P have a TDP of 300 watts, despite their differing core counts and clock speeds.

Q: How do these processors compare to their nearest rivals in the database?

A: The 678X has an average benchmark score of 193477, placing it 0.4% behind the AMD EPYC 9335 (194228) and 0.7% behind the Intel Xeon 6741P (194901), while leading the Intel Xeon 6740E by 3.1% and the AMD EPYC 8534P by 4.5%. The 6745P averages 154858, sitting 2.3% behind the Intel Xeon 676X (158540), 3.4% behind the AMD EPYC 9355P (160358), 4.4% behind the AMD EPYC 7663 (161973), and 4.7% behind the AMD EPYC 9375F (162497).

Q: What are the core and thread counts for each?

A: The Intel Xeon 678X features 48 cores and 96 threads, while the Intel Xeon 6745P offers 32 cores and 64 threads.

The Verdict

The data supports a clear separation between these two Granite Rapids parts. The Intel Xeon 678X is the outright performance leader, winning every benchmark in the comparison set. Its 48 cores and 96 threads provide a substantial multi-threaded advantage, while its higher boost clock of 4.90 GHz also secures single-threaded wins despite a lower base clock.

For workloads that scale with core count, such as Cinebench multi-core rendering or Passmark multithread tests, the 678X delivers a consistent 17% improvement. For tasks that stress specific instruction paths, like prime number calculation or floating point math, the advantage grows to between 35% and 48%. Data compression and encryption workloads also favor the 678X by roughly 25%.

The Intel Xeon 6745P, with 32 cores and 64 threads, remains a capable server processor in its own right. Its average benchmark score of 154858 places it in the 98th percentile of all CPUs in the database, and it holds its own against several AMD EPYC rivals, trailing the EPYC 9355P by only 3.4%. However, in a direct comparison, it cannot match the 678X's raw throughput.

The choice comes down to workload density. If a deployment requires maximum parallel processing, the 678X is the stronger option based on benchmark evidence. If a lower core count is acceptable and the 6745P's launch MSRP of $5250 fits the budget, it still delivers competitive performance for server and workstation tasks. The 678X carries a launch MSRP of $3749, which is lower than the 6745P's launch MSRP, though pricing considerations are separate from the performance data.

Specification Differences

The two processors differ in several key specification areas while sharing a common foundation. Both use the Intel Socket 4710, feature 5 nm process technology, and have the same die size of 2x 598 mm². Both support DDR5 memory with an eight-channel memory bus and 409.6 GB/s memory bandwidth, and both include ECC memory support. Neither has integrated graphics, and both are classified for the Server/Workstation market segment.

The most significant difference is core count: the 678X has 48 cores and 96 threads, while the 6745P has 32 cores and 64 threads. Clock speeds also differ. The 678X has a base clock of 2.40 GHz and a boost clock of 4.90 GHz, whereas the 6745P has a higher base clock of 3.10 GHz but a lower boost clock of 4.30 GHz. TDP is identical at 300 watts.

Cache configurations differ notably. Both have 112 KB of L1 cache per core and 2 MB of L2 cache per core. However, the L3 cache is significantly larger on the 6745P at 336 MB shared, compared to 192 MB shared on the 678X. This is an interesting inversion, as the lower-core-count processor carries more shared L3 cache.

PCIe lane counts also vary. The 678X provides Gen 5 with 128 lanes (CPU only), while the 6745P offers Gen 5 with 88 lanes (CPU only). The multiplier is unlocked on the 678X but locked on the 6745P. Release dates differ as well, with the 6745P launching on 2025-02-23 and the 678X following on 2026-02-01. The part numbers are SA2CX for the 678X and SRWPAQ7L9 for the 6745P.

Architecture Differences

Both processors share the Granite Rapids architecture and codename, built on Intel's 5 nm process. They belong to different generations within the Xeon 6 family: the 678X is part of the Xeon 600 (Granite Rapids-WS) line, while the 6745P is part of the Xeon 6 (Granite Rapids-SP) line. This generation split likely accounts for some of the performance and feature differences.

The core microarchitecture is presumably identical given the shared codename, but the implementations differ. The 678X packs 48 cores versus 32 cores on the 6745P, a 50% increase in core count. This directly explains its substantial lead in multi-threaded benchmarks, though the 6745P's larger L3 cache (336 MB vs 192 MB) may help mitigate some memory-latency penalties in certain workloads.

The 678X's higher boost clock of 4.90 GHz versus 4.30 GHz on the 6745P is a clear architectural advantage for single-threaded performance. Despite the 6745P's higher base clock of 3.10 GHz, the 678X still wins single-threaded tests by 8.9%, indicating that boost behavior is more critical than base frequency in these benchmarks.

PCIe lane allocation also reflects their different positioning. The 678X offers 128 Gen 5 lanes, enabling more expansion and I/O bandwidth, while the 6745P provides 88 Gen 5 lanes. Both are CPU-only lane counts, meaning additional chipset lanes would be needed for full platform connectivity.

The unlocked multiplier on the 678X is a notable architectural feature absent from the 6745P. This allows for overclocking, which could further extend the 678X's performance lead in the right thermal and power conditions. The 6745P's locked multiplier restricts such tuning.

Both processors are actively in production and hold high percentiles in the database: the 678X sits in the 99th percentile of all CPUs, while the 6745P sits in the 98th percentile. Their average benchmark scores of 193477 and 154858 respectively reflect the 48-core part's overall superiority, but the 32-core part remains a strong contender in its own segment.