Intel Xeon 638 vs Intel Xeon w5-3535X Comparison
Intel Xeon 638
Xeon w5-3535X
PERFORMANCE BENCHMARKS
Analysis: Intel Xeon 638 vs Intel Xeon w5-3535X
The Intel Xeon w5-3535X and Intel Xeon 638 are remarkably close in overall benchmark performance, with the w5-3535X holding a negligible 0.5% average score advantage (81115 vs 80723). The data shows the Xeon 638 wins a majority of individual tests (12 of 17), but the w5-3535X counter-punches decisively in specific workloads. The Xeon 638 is the better all-rounder for general compute, physics simulation, and prime-number work, while the w5-3535X is the specialist for data compression, encryption, and extended instruction sets. For most server/workstation buyers, the Xeon 638’s higher single-thread scores and lower TDP make it the pragmatic choice; the w5-3535X only makes sense if your workload heavily favors its specific wins.
The Verdict
The benchmark results paint a picture of two processors that trade blows within a narrow band, but the Xeon 638 edges out the w5-3535X in the majority of tests. The Xeon 638 wins 12 of 17 head-to-head benchmarks, including all six Cinebench tests (R15, R20, R23, both multi- and single-core), with deltas ranging from -2.5% to -2.6% in favor of the 638. It also dominates in PassMark physics (4704 vs 3547, a 24.6% lead) and prime number finding (381 vs 269, a 29.4% lead).
The w5-3535X, however, is not without its strongholds. It wins PassMark data encryption by 2.1%, extended instructions by 6.5%, and narrowly edges out the 638 in data compression (0.8%), floating-point math (0.8%), and integer math (0.7%). The 638’s wins in PassMark multithread (55651 vs 54088, a 2.8% lead) and single-thread (3670 vs 3602, a 1.9% lead) reinforce its general-purpose superiority.
The verdict is clear: the Xeon 638 is the safer choice for diverse workloads, particularly those involving physics, prime-number calculation, or any Cinebench-style rendering. The w5-3535X is the pick only for specialized tasks like encryption, compression, and heavy extended-instruction use, where its wins are substantial but narrower in scope.
Architecture Differences
The two processors differ fundamentally in their underlying design. The w5-3535X is built on Sapphire Rapids architecture using a 10 nm process node, with a die size of 4x 477 mm². The Xeon 638, in contrast, uses Granite Rapids architecture on a 5 nm node, with a single 598 mm² die. This process advantage likely contributes to the 638’s higher efficiency, reflected in its 180 W TDP versus the w5-3535X’s 300 W TDP.
Core counts favor the w5-3535X: it has 20 cores and 40 threads, compared to the 638’s 16 cores and 32 threads. Despite this 25% core disadvantage, the 638 still wins most benchmarks, suggesting per-core efficiency gains from the newer process and architecture. Cache layouts also differ: the w5-3535X has 80 KB L1 per core, 2 MB L2 per core, and a 52.5 MB L3 cache. The 638 has 112 KB L1 per core, 2 MB L2 per core, and a larger 72 MB shared L3 cache.
Memory subsystems diverge significantly. The w5-3535X supports eight-channel DDR5 memory with 307.2 GB/s bandwidth, while the 638 uses quad-channel DDR5 with 204.8 GB/s. The w5-3535X also offers more PCIe lanes: 112 Gen 5 lanes (CPU only) versus the 638’s 80 Gen 5 lanes. Both support ECC memory and have no integrated graphics. Clock speeds are identical for boost (4.80 GHz), but the 638 has a higher base clock (3.20 GHz vs 2.90 GHz). Both have unlocked multipliers, but they use different sockets (4677 for the w5-3535X, 4710 for the 638).
Where Each One Wins
The Xeon 638 wins in all rendering and general compute benchmarks. Its Cinebench R23 multicore score of 47202 beats the w5-3535X’s 45974 by 2.6%, a consistent margin across R15 (4757 vs 4634), R20 (19824 vs 19309), and all single-core variants. This makes the 638 the clear choice for 3D rendering, video encoding, and any workload that scales with multicore Cinebench-style performance.
The 638’s dominance is most pronounced in physics and prime-number workloads. Its PassMark physics score of 4704 is 24.6% higher than the w5-3535X’s 3547, and its prime number score of 381 is 29.4% higher (269 for the w5-3535X). These are massive margins that indicate the 638’s per-core strength and possibly better integer execution. The 638 also wins PassMark multithread (55651 vs 54088) and random string sorting (74318 vs 73618), further cementing its general workload advantage.
The w5-3535X wins in data-centric tasks. Its PassMark data encryption score of 36784 is 2.1% higher than the 638’s 36030, and extended instructions score of 60183 is 6.5% higher (56498 for the 638). It also narrowly wins data compression (731388 vs 725818, 0.8%), floating-point math (145924 vs 144757, 0.8%), and integer math (186158 vs 184884, 0.7%). These wins, while smaller in percentage, reveal the w5-3535X’s strength in cryptographic and SIMD-heavy workloads.
FAQ
Q: Which processor has more cores and threads?
A: The Intel Xeon w5-3535X has 20 cores and 40 threads, while the Intel Xeon 638 has 16 cores and 32 threads.
Q: Which processor has higher memory bandwidth?
A: The w5-3535X offers 307.2 GB/s via eight-channel DDR5, whereas the 638 offers 204.8 GB/s via quad-channel DDR5.
Q: Does the Xeon 638 win in single-threaded performance?
A: Yes, the 638 wins all single-core Cinebench tests (R15: 671 vs 654, R20: 2798 vs 2725, R23: 6663 vs 6490) and PassMark single-thread (3670 vs 3602).
Q: What is the biggest benchmark margin between the two?
A: The largest delta is in PassMark find prime numbers, where the 638 scores 381 versus the w5-3535X’s 269, a 29.4% advantage.
Q: Which processor has a larger L3 cache?
A: The Intel Xeon 638 has 72 MB of shared L3 cache, while the w5-3535X has 52.5 MB.
Q: Are both processors unlocked for overclocking?
A: Yes, both have an unlocked multiplier, though they use different sockets (4677 for the w5-3535X, 4710 for the 638).
Head-to-Head Benchmarks
The Cinebench suite shows a consistent pattern: the Xeon 638 leads every test by roughly 2.5-2.6%. In Cinebench R15 multicore, the 638 scores 4757 against 4634, a -2.6% delta. The single-core R15 test shows 671 vs 654, a -2.5% delta. This pattern holds in R20 multicore (19824 vs 19309, -2.6%) and R23 multicore (47202 vs 45974, -2.6%). The 638’s wins are systematic, not isolated, indicating a per-core clock-for-clock advantage.
The PassMark physics test delivers one of the most lopsided results: 4704 for the 638 versus 3547 for the w5-3535X, a 24.6% margin. This is a massive gap in a test that often correlates with gaming and real-time simulation performance. Similarly, in PassMark find prime numbers, the 638’s 381 crushes the w5-3535X’s 269, a 29.4% delta. These two tests alone show the 638’s superior integer and physics throughput, likely stemming from its newer Granite Rapids architecture.
The w5-3535X’s biggest win is in PassMark extended instructions, where it scores 60183 versus 56498, a 6.5% advantage. This is its largest percentage win and highlights its edge in AVX-512-style workloads. It also wins data encryption (36784 vs 36030, 2.1%) and narrowly takes data compression (731388 vs 725818, 0.8%), floating-point math (145924 vs 144757, 0.8%), and integer math (186158 vs 184884, 0.7%). These wins are smaller but show the w5-3535X is not without merit in specific niches.
The PassMark multithread test goes to the 638 by 2.8% (55651 vs 54088), while random string sorting favors the 638 by 0.9% (74318 vs 73618). The single-thread tests are close, with the 638 leading by 1.9% (3670 vs 3602). In aggregate, the data shows a 12-5 win count for the 638, but the average benchmark scores are nearly identical (81115 vs 80723), meaning the choice between them depends entirely on workload profile rather than raw capability.