Intel Core Ultra 9 285K vs Intel Xeon 6736P Comparison

The Intel Xeon 6736P and Intel Core Ultra 9 285K represent two distinct extremes of Intel’s current lineup: a 36-core Granite Rapids server processor aimed at multi-socket workstations and a 24-core Arrow Lake desktop flagship. The benchmark data reveals a surprisingly close overall contest, with the Core Ultra 9 285K taking 10 of 17 head-to-head tests, yet the Xeon 6736P securing the most dramatic single-core victory. Average benchmark scores place the Xeon at 87,864 and the Core Ultra 9 at 83,807, a 4.8% gap that undersells how differently these chips perform across workloads.

Head-to-Head Benchmarks

The most striking result in the entire comparison is the Cinebench R23 single-core test, where the Xeon 6736P scores 6,008 against the Core Ultra 9’s 2,377. That is a 152.8% advantage for the server part, an extraordinary margin that completely inverts the usual desktop-versus-server relationship. The same pattern appears in Cinebench R15 single-core, with the Xeon at 605 and the Core Ultra 9 at 359, a 68.5% lead. These results suggest the Xeon’s Granite Rapids cores are exceptionally strong in legacy single-threaded rendering workloads, despite the Core Ultra 9’s much higher boost clock of 5.70 GHz versus 4.10 GHz.

Multi-core rendering tells a different story. The Core Ultra 9 wins Cinebench R15 multicore by 33.9% (6,494 vs 4,290) and Cinebench R20 multicore by 25.5% (24,003 vs 17,875). Yet in Cinebench R23 multicore, the two are effectively tied: the Xeon scores 42,561 and the Core Ultra 9 scores 42,522, a 0.1% difference. This inconsistency across Cinebench versions indicates that the Xeon’s 72 threads scale better as the workload becomes more demanding, while the Core Ultra 9’s 24 threads hit diminishing returns earlier.

PassMark results further split the field. The Core Ultra 9 dominates floating-point math, scoring 224,324 versus 149,770, a 33.2% lead, and also wins integer-heavy tasks like find prime numbers (541 vs 392, a 27.5% advantage). The Xeon fights back in integer math, scoring 206,833 against 172,379, a 20% win, and takes physics with 6,531 versus 3,938, a 65.8% margin. Data compression is nearly a draw, with the Xeon at 796,658 and the Core Ultra 9 at 790,052, a 0.8% edge.

Single-threaded PassMark strongly favors the desktop chip: 5,087 versus 2,024, a 60.2% deficit for the Xeon. Data encryption also goes to the Core Ultra 9 (57,745 vs 46,236, a 19.9% lead), as does extended instructions (62,277 vs 55,563, a 10.8% win). Random string sorting is a Xeon victory at 103,723 versus 94,927, a 9.3% margin. Overall, the Core Ultra 9 wins 10 tests, the Xeon wins 7, but the Xeon’s wins are often by larger percentages—particularly the 152.8% R23 single-core blowout.

Architecture Differences

The two processors are built on fundamentally different silicon. The Xeon 6736P uses Granite Rapids architecture on a 5 nm process from Intel’s own foundry, with a die size of 598 mm². The Core Ultra 9 285K uses Arrow Lake on a 3 nm process from TSMC, with a die size of 243 mm² and 17,800 million transistors. The process node advantage is clear in the Core Ultra 9’s lower 125 W TDP versus the Xeon’s 205 W, despite the Xeon having 36 cores and 72 threads—50% more cores and 200% more threads than the Core Ultra 9’s 24 cores and 24 threads.

Cache hierarchies diverge sharply. The Xeon offers 144 MB of shared L3 cache, four times the Core Ultra 9’s 36 MB. Per-core L1 and L2 caches, however, favor the desktop part: 192 KB L1 and 3 MB L2 per core versus the Xeon’s 112 KB L1 and 2 MB L2. The Xeon compensates with massive aggregate L2 capacity across 36 cores, but the Core Ultra 9’s larger per-core caches likely contribute to its single-thread PassMark advantage.

Memory subsystems reflect their intended roles. The Xeon supports eight-channel DDR5 with 409.6 GB/s of bandwidth, while the Core Ultra 9 is dual-channel with 102.4 GB/s—a fourfold difference. Both support ECC memory, but the Xeon’s 88 PCIe Gen 5 lanes dwarf the Core Ultra 9’s 20. The Xeon has no integrated graphics; the Core Ultra 9 includes Arc Xe-LPG Graphics 64EU. Sockets are incompatible, with the Xeon on Intel Socket 4710 and the Core Ultra 9 on Intel Socket 1851.

The Verdict

Data from the benchmarks indicates that the Core Ultra 9 285K is the better all-around processor for most workloads, winning 10 of 17 tests. Its advantages in floating-point math, encryption, and single-threaded PassMark scores make it the stronger choice for general desktop computing, content creation, and any task that benefits from high clock speeds. The 5.70 GHz boost clock and 3 nm process deliver tangible wins in latency-sensitive applications, as shown by the 60.2% lead in single-thread PassMark.

The Xeon 6736P is the pick when parallel throughput and memory bandwidth dominate. Its 72 threads and 144 MB L3 cache drive a 65.8% win in PassMark physics and a 20% win in integer math, while the eight-channel memory subsystem provides 409.6 GB/s bandwidth—essential for server workloads that stream data. The 152.8% R23 single-core victory is anomalous but real, suggesting certain legacy rendering workloads will favor the Xeon unexpectedly.

For a desktop user, the Core Ultra 9’s 24 cores are ample, and its 125 W TDP makes it far easier to cool. For a workstation running database or scientific workloads, the Xeon’s 36 cores and ECC memory support justify its higher power draw. The 0.8% data compression win for the Xeon hints that file-server roles slightly prefer it, while the Core Ultra 9’s 19.9% encryption lead suits security-focused tasks. Neither chip is a clear winner; the choice depends entirely on whether the workload scales across 72 threads.

Specification Differences

| Specification | Intel Xeon 6736P | Intel Core Ultra 9 285K |

|---|---|---|

| Cores | 36 | 24 |

| Threads | 72 | 24 |

| Base Clock | 2.00 GHz | 3.70 GHz |

| Boost Clock | 4.10 GHz | 5.70 GHz |

| TDP | 205 W | 125 W |

| Socket | Intel Socket 4710 | Intel Socket 1851 |

| Architecture | Granite Rapids | Arrow Lake |

| Process Node | 5 nm | 3 nm |

| Foundry | Intel | TSMC |

| Transistors | Not listed | 17,800 million |

| Die Size | 598 mm² | 243 mm² |

| L1 Cache | 112 KB (per core) | 192 KB (per core) |

| L2 Cache | 2 MB (per core) | 3 MB (per core) |

| L3 Cache | 144 MB (shared) | 36 MB (shared) |

| Memory Bus | Eight-channel | Dual-channel |

| Memory Bandwidth | 409.6 GB/s | 102.4 GB/s |

| PCIe Lanes | Gen 5, 88 Lanes | Gen 5, 20 Lanes |

| Integrated Graphics | N/A | Arc Xe-LPG Graphics 64EU |

| Market Segment | Server/Workstation | Desktop |

| Release Date | 2025-02-23 | 2024-10-23 |

| Launch MSRP | $3351 | $589 |

| Multiplier Unlocked | No | Yes |

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon 6736P has 36 cores and 72 threads, while the Intel Core Ultra 9 285K has 24 cores and 24 threads.

Q: What is the largest benchmark margin between them?

A: The largest margin is in Cinebench R23 single-core, where the Xeon 6736P scores 6,008 versus the Core Ultra 9’s 2,377, a 152.8% advantage.

Q: Do both processors support ECC memory?

A: Yes, both the Xeon 6736P and the Core Ultra 9 285K list ECC memory support.

Q: How do their memory bandwidths compare?

A: The Xeon 6736P provides 409.6 GB/s over an eight-channel bus, while the Core Ultra 9 285K provides 102.4 GB/s over a dual-channel bus.

Q: Which chip has a higher boost clock?

A: The Core Ultra 9 285K boosts to 5.70 GHz, compared to the Xeon 6736P’s 4.10 GHz.

Q: How many benchmark tests does each processor win?

A: The Core Ultra 9 285K wins 10 of 17 head-to-head tests, while the Xeon 6736P wins 7.

Where Each One Wins

The Core Ultra 9 285K is the clear choice for floating-point and encryption workloads. Its 33.2% lead in PassMark floating-point math (224,324 vs 149,770) and 19.9% lead in data encryption (57,745 vs 46,236) make it superior for scientific computing, financial modeling, and secure communications. The 27.5% win in find prime numbers (541 vs 392) and 10.8% lead in extended instructions (62,277 vs 55,563) further cement its position for algorithm-heavy tasks. Single-threaded applications are overwhelmingly its domain, with a 60.2% PassMark single-thread advantage.

The Xeon 6736P wins where thread count and cache size matter. Its 65.8% lead in PassMark physics (6,531 vs 3,938) indicates strong performance in simulation and dynamics workloads. The 20% win in integer math (206,833 vs 172,379) suits database indexing and financial integer calculations. Random string sorting favors the Xeon by 9.3%, useful for text processing and log analysis. The 0.8% data compression win makes it marginally better for archival and file-server tasks, and its 144 MB L3 cache provides a massive buffer for repeated data access patterns.

For multi-core rendering, the results are mixed. The Core Ultra 9 leads in R15 and R20 multicore, but the Xeon ties in R23 multicore and wins single-core by 152.8%. Users running older Cinebench versions should expect desktop dominance, while newer benchmarks show parity. The Xeon’s 88 PCIe Gen 5 lanes and eight-channel memory make it the only choice for high-bandwidth peripherals and large memory arrays, while the Core Ultra 9’s integrated graphics and unlocked multiplier serve desktop users who want a single-chip solution with overclocking headroom.