Intel Xeon 6732P vs Intel Xeon 674X Comparison

The Intel Xeon 6732P and Intel Xeon 674X are two Granite Rapids server processors that occupy nearly the same average benchmark territory, yet achieve it through very different configurations. The 6732P fields 32 cores at a 3.80 GHz base clock, while the 674X counters with 28 cores but a much higher 4.90 GHz boost. Their average benchmark scores are almost identical—143444 for the 6732P versus 143103 for the 674X, a margin of just 0.2%—but the distribution of wins across workloads tells a clear story of divergent strengths.

Head-to-Head Benchmarks

The 674X dominates the Cinebench suite outright. In Cinebench R15 multicore, it scores 7213 against the 6732P’s 6412, a 11.1% advantage. The same 11.1% delta repeats in R15 singlecore (1018 vs 905), R20 multicore (30057 vs 26720), R20 singlecore (4243 vs 3772), and R23 multicore (71566 vs 63621). This consistency across rendering workloads suggests the 674X’s higher boost clock—4.90 GHz versus 4.10 GHz—provides a decisive edge in lightly-threaded and moderately-threaded tasks alike, despite having four fewer cores.

The 6732P strikes back in PassMark’s computational tests. Data compression favors it heavily: 1339480 versus 1236272, an 8.3% win. Extended instructions show a 9.6% gap (106697 vs 97373), and floating-point math lands at 7.3% ahead (261703 vs 243877). Integer math adds another 8.2% victory (334340 vs 308968), while physics yields 6.9% (8109 vs 7586). The pattern is clear—the 6732P’s extra four cores and 64 threads (versus 56) translate directly into throughput for parallel arithmetic and compression tasks.

The 674X reclaims ground in single-threaded PassMark testing, and here the margin is enormous. Its 3933 single-thread score dwarfs the 6732P’s 2506, a 36.3% gap. That is the single largest delta in the entire comparison, underscoring how much the 674X sacrifices core count for raw clock speed. PassMark multithread also goes to the 674X at 84196 versus 74849 (11.1%), and find prime numbers favors it at 693 versus 628 (9.4%). Random string sorting, however, goes to the 6732P at 133467 versus 127529 (4.7%), and data encryption follows at 63848 versus 61195 (4.3%).

Overall, the 674X takes 9 wins to the 6732P’s 7. Yet the average benchmark scores remain within 0.2% of each other, meaning the wins and losses cancel out almost perfectly across the full test battery. The 6732P’s wins are concentrated in multi-threaded math and compression, while the 674X’s are spread across rendering, single-thread, and prime-number workloads.

Architecture Differences

Both processors share the Granite Rapids architecture and are built on the same Intel 5 nm process node. They also share the same socket, Intel Socket 4710, and identical cache layouts: 112 KB of L1 per core, 2 MB of L2 per core, and 144 MB of shared L3. Memory support is likewise identical—DDR5 over an eight-channel bus with 409.6 GB/s of bandwidth and ECC enabled.

The core counts diverge significantly. The 6732P packs 32 cores and 64 threads, while the 674X offers 28 cores and 56 threads. That difference is partially offset by clock speeds: the 6732P runs at a 3.80 GHz base and 4.10 GHz boost, whereas the 674X starts lower at 3.00 GHz base but rockets to 4.90 GHz boost. The TDP reflects this tradeoff—the 6732P draws 350 watts to power its extra cores, while the 674X is rated at 270 watts.

The 674X also carries a notable physical distinction: its die size is listed as 2x 598 mm², indicating a dual-die package. The 6732P’s die size is not specified. PCIe lane counts differ as well, with the 6732P providing 136 Gen 5 lanes versus 128 on the 674X. The 674X is multiplier-unlocked, a feature the 6732P lacks, and its part number (SA2CZ) differs from the 6732P’s SRVP2. Neither chip includes integrated graphics.

Release timing separates the pair further. The 6732P launched on 2025-05-21 with a launch MSRP of $5295. The 674X arrived later, on 2026-02-01, at a launch MSRP of $2199. Both are listed as Active production and target the Server/Workstation market segment.

FAQ

Q: Which processor has the higher average benchmark score?

A: The Intel Xeon 6732P posts an average benchmark score of 143444, while the Intel Xeon 674X scores 143103. The 6732P leads by 0.2%, and both sit at the 98th percentile of all CPUs.

Q: How do the two compare in Cinebench R23 multicore?

A: The Intel Xeon 674X wins decisively, scoring 71566 against the 6732P’s 63621. That represents an 11.1% advantage for the 674X in this rendering workload.

Q: What is the biggest single benchmark gap between them?

A: PassMark single-thread testing shows the largest delta, with the 674X scoring 3933 versus the 6732P’s 2506. That is a 36.3% difference in favor of the 674X.

Q: Which chip is better for data compression workloads?

A: The Intel Xeon 6732P leads in PassMark data compression with a score of 1339480, compared to 1236272 for the 674X—an 8.3% advantage for the 6732P.

Q: Do both processors support the same memory configuration?

A: Yes. Both support DDR5 memory over an eight-channel bus, with 409.6 GB/s of bandwidth and ECC memory support.

Q: What is the core and thread count difference?

A: The 6732P has 32 cores and 64 threads, while the 674X has 28 cores and 56 threads. The 6732P also carries a higher TDP of 350 watts versus 270 watts for the 674X.

Specification Differences

The two processors differ in several key specification fields. Core count is 32 on the 6732P versus 28 on the 674X, with threads at 64 versus 56. Base clock speeds are 3.80 GHz versus 3.00 GHz, while boost clocks reverse the order: 4.10 GHz on the 6732P versus 4.90 GHz on the 674X. TDP is 350 watts versus 270 watts.

The 674X features a die size of 2x 598 mm², while the 6732P has no listed die size. PCIe lane counts differ at 136 for the 6732P and 128 for the 674X, both Gen 5. The 674X is multiplier-unlocked; the 6732P is not. Generation labels also differ: the 6732P is listed as Xeon 6 (Granite Rapids-SP), while the 674X is Xeon 600 (Granite Rapids-WS).

Release dates and launch MSRP values are distinct. The 6732P launched on 2025-05-21 at $5295, while the 674X launched on 2026-02-01 at $2199. Part numbers are SRVP2 for the 6732P and SA2CZ for the 674X. All other major specifications—including L1, L2, and L3 cache sizes, memory support, memory bus width, memory bandwidth, ECC support, integrated graphics, socket, architecture, codename, process node, foundry, and market segment—are identical between the two.

Where Each One Wins

The Intel Xeon 674X is the clear choice for single-threaded performance and rendering workloads. Its 36.3% lead in PassMark single-thread testing is unmatched, and it extends that advantage across every Cinebench test—R15, R20, and R23—with a uniform 11.1% margin. It also wins PassMark multithread (84196 vs 74849) and find prime numbers (693 vs 628). Workloads like 3D rendering, frequency-sensitive simulations, and any application that rewards high boost clocks will favor the 674X. Its lower 270-watt TDP and unlocked multiplier add operational flexibility, though its 28-core count limits raw parallel throughput.

The Intel Xeon 6732P is the better fit for data-heavy, parallel compute tasks. It wins PassMark data compression by 8.3%, extended instructions by 9.6%, floating-point math by 7.3%, integer math by 8.2%, physics by 6.9%, data encryption by 4.3%, and random string sorting by 4.7%. These are precisely the workloads that scale with core count and thread count—the 6732P’s 32 cores and 64 threads outmuscle the 674X’s 28 cores and 56 threads. Database operations, compression pipelines, scientific computing, and encryption-heavy services will see measurably better throughput on the 6732P.

The average benchmark scores tell the final story: 143444 for the 6732P and 143103 for the 674X, a 0.2% separation that is statistically negligible. The 674X wins more individual tests (9 versus 7), but the 6732P’s wins are larger in aggregate on the PassMark side, balancing the ledger. In practice, the decision comes down to workload type. Frequency-bound tasks belong to the 674X; thread-bound tasks belong to the 6732P. Neither chip is a universal victor, and both hold the 98th percentile ranking among all CPUs.