AMD Ryzen Threadripper 3970X vs Intel Xeon Gold 6336Y Comparison

Head-to-Head Benchmarks

The benchmark data paints a remarkably one-sided picture in this matchup. Across all six Cinebench comparisons, the AMD Ryzen Threadripper 3970X wins decisively, with the Intel Xeon Gold 6336Y trailing by a consistent margin of roughly 27.8-27.9% in every test. This uniformity is striking; it suggests the performance gap stems from fundamental architectural differences rather than workload-specific quirks.

In Cinebench R15 multicore, the Threadripper 3970X scores 5405 against the Xeon's 3899, a delta of -27.9% from the Intel part's perspective. The single-core R15 test tells a similar story: 762 versus 550, a -27.8% gap. Moving to Cinebench R20, the AMD part delivers 22524 multicore and 3179 single-core, while the Intel chip manages 16249 and 2294 respectively. The deltas remain locked at -27.9% and -27.8%.

Cinebench R23 confirms the pattern. The Threadripper 3970X posts 53630 multicore and 7571 single-core, against the Xeon Gold 6336Y's 38689 and 5462. Again, the delta sits at -27.9% for multicore and -27.9% for single-core. The consistency across every generation of Cinebench — from R15 through R23 — indicates a stable, predictable performance hierarchy.

The average benchmark scores reflect this overall trend, though the gap narrows when including the AMD part's additional 3DMark and GeekBench results. The Xeon Gold 6336Y averages 11191 across its six Cinebench scores, while the Threadripper 3970X averages 10841 across its fourteen benchmark results. This discrepancy arises because the AMD chip's 3DMark threaded tests and GeekBench scores pull its average down relative to its dominant Cinebench showing.

Looking at the nearest rivals for context, the Xeon Gold 6336Y sits at the 66th percentile among all CPUs, with its closest competitor being the Intel Core i3-1305U at a -0.1% delta. The Threadripper 3970X also occupies the 66th percentile, but its nearest rivals are quite different: the Intel Core i7-6700 trails by -2.1%, while the AMD EPYC 7D12 sits 2.8% behind. These percentile rankings show both processors landing in similar overall performance tiers, despite the head-to-head dominance of the AMD part.

Architecture Differences

The two processors diverge sharply in nearly every architectural dimension. The Xeon Gold 6336Y is built on Intel's Ice Lake-SP architecture at a 10 nm process node, fabricated by Intel itself. It features 24 cores and 48 threads, with a base clock of 2.40 GHz and a boost clock of 3.60 GHz. The Threadripper 3970X, in contrast, uses AMD's Zen 2 architecture codenamed Castle Peak, built on a 7 nm process at TSMC. It packs 32 cores and 64 threads, with a substantially higher base clock of 3.70 GHz and boost clock of 4.50 GHz.

Cache configurations highlight another fundamental split. The Xeon provides 64 KB of L1 and 1 MB of L2 per core, with 36 MB of shared L3 cache. The Threadripper offers the same 64 KB L1 per core, but only 512 KB of L2 per core, while its L3 cache is a massive 128 MB. The AMD part's transistor count of 15,200 million across four 74 mm² dies explains its larger cache footprint. These cache differences likely contribute to the AMD processor's consistent single-core advantage, as the larger L3 pool can hold more working data.

Memory architecture diverges as well. The Xeon Gold 6336Y supports eight-channel DDR4 memory with a theoretical bandwidth of 204.8 GB/s, while the Threadripper 3970X uses quad-channel DDR4 with 102.4 GB/s bandwidth. The Xeon also supports ECC memory, whereas the Threadripper does not. Despite having half the memory channels and bandwidth on paper, the AMD part still manages to outperform the Intel chip in every benchmark tested — proof of its core count and clock speed advantages.

PCIe connectivity differs: the Xeon provides Gen 4 with 64 lanes from the CPU, while the Threadripper also supports Gen 4 but without a specified lane count in the data. The Xeon targets the server/workstation segment on Intel Socket 4189, while the Threadripper is a desktop part on AMD Socket TRX4. The Intel chip's multiplier is locked, while the AMD part is unlocked for overclocking. The Threadripper's part number is listed as 100-000000011100-100000011WOF, while the Xeon's part number is not provided.

The release timeline also separates them: the Xeon Gold 6336Y launched on April 5, 2021, while the Threadripper 3970X arrived on November 24, 2019. Despite launching roughly a year and a half later, the Intel part cannot close the performance gap against the older AMD design.

Where Each One Wins

The Threadripper 3970X wins every benchmark in this comparison, so the "where each one wins" analysis is largely about the margin and nature of those wins. The AMD part's advantages are most pronounced in multicore workloads, where its 32 cores versus the Xeon's 24 cores deliver a 27.9% lead in Cinebench R15, R20, and R23 multicore tests. This makes the Threadripper the clear choice for heavily threaded rendering, simulation, and content creation tasks.

Single-core performance also favors the Threadripper, with a 27.8-27.9% lead across all three Cinebench single-core tests. This advantage stems from its higher base and boost clocks — 3.70 GHz and 4.50 GHz versus 2.40 GHz and 3.60 GHz — rather than any architectural superiority in instruction handling. For applications that rely on lightly threaded performance, such as legacy software or interactive workloads, the AMD part remains ahead.

The Xeon Gold 6336Y's strengths lie outside raw benchmark performance. Its eight-channel memory subsystem with 204.8 GB/s bandwidth provides substantially higher theoretical memory throughput than the Threadripper's quad-channel 102.4 GB/s configuration. In memory-bandwidth-sensitive server workloads — such as large-scale database operations or high-performance computing tasks that stream data — the Xeon's memory architecture could offer advantages not captured in the Cinebench suite. Additionally, ECC memory support makes the Xeon suitable for environments requiring error correction, a feature the Threadripper lacks.

The Xeon also targets a different market segment: server/workstation rather than desktop. Its Intel Socket 4189 platform and 64 PCIe Gen 4 lanes support multi-socket configurations and extensive I/O expansion, which may be more relevant for enterprise deployments than raw CPU benchmark scores. The Threadripper, with its desktop orientation and unlocked multiplier, suits enthusiasts and professionals who prioritize maximum single-socket performance and overclocking flexibility.

For users choosing between these two, the decision hinges on workload nature. Pure computational throughput favors the Threadripper 3970X in every measured category. The Xeon's value proposition rests on platform features — memory bandwidth, ECC, PCIe lanes — that matter in specific server contexts. The benchmark data alone cannot capture those infrastructure-level differences.

FAQ

Q: Which processor has more cores?

A: The AMD Ryzen Threadripper 3970X has 32 cores and 64 threads, while the Intel Xeon Gold 6336Y has 24 cores and 48 threads.

Q: What is the performance difference in Cinebench R23?

A: In Cinebench R23 multicore, the Threadripper 3970X scores 53630 versus the Xeon's 38689, a -27.9% delta. In single-core, the AMD part scores 7571 against 5462, a -27.9% delta.

Q: Does the Xeon support ECC memory?

A: Yes, the Intel Xeon Gold 6336Y supports ECC memory. The AMD Ryzen Threadripper 3970X does not support ECC.

Q: What are the memory bandwidth specifications?

A: The Xeon Gold 6336Y has eight-channel memory with 204.8 GB/s bandwidth, while the Threadripper 3970X has quad-channel memory with 102.4 GB/s bandwidth.

Q: Which processor has a higher boost clock?

A: The AMD Ryzen Threadripper 3970X boosts to 4.50 GHz, while the Intel Xeon Gold 6336Y boosts to 3.60 GHz.

Q: What is the process node for each processor?

A: The Xeon Gold 6336Y uses Intel's 10 nm process, while the Threadripper 3970X uses TSMC's 7 nm process.

Specification Differences

| Specification | Intel Xeon Gold 6336Y | AMD Ryzen Threadripper 3970X |

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

| Cores | 24 | 32 |

| Threads | 48 | 64 |

| Base Clock | 2.40 GHz | 3.70 GHz |

| Boost Clock | 3.60 GHz | 4.50 GHz |

| TDP | 185 W | 280 W |

| Socket | Intel Socket 4189 | AMD Socket TRX4 |

| Architecture | Ice Lake | Zen 2 |

| Codename | Ice Lake-SP | Castle Peak |

| Process Node | 10 nm | 7 nm |

| Foundry | Intel | TSMC |

| Transistors | Not specified | 15,200 million |

| Die Size | Not specified | 4x 74 mm² |

| L2 Cache | 1 MB (per core) | 512 KB (per core) |

| L3 Cache | 36 MB (shared) | 128 MB |

| Memory Bus | Eight-channel | Quad-channel |

| Memory Bandwidth | 204.8 GB/s | 102.4 GB/s |

| ECC Memory | Yes | No |

| PCIe | Gen 4, 64 Lanes (CPU only) | Gen 4 |

| Market Segment | Server/Workstation | Desktop |

| Release Date | 2021-04-05 | 2019-11-24 |

| Launch MSRP | Not specified | $1999 |

| Multiplier Unlocked | No | Yes |