AMD Ryzen Threadripper 3960X vs Intel Xeon Gold 6346 Comparison

FAQ

Q: Which processor has more cores and threads?

A: The AMD Ryzen Threadripper 3960X has 24 cores and 48 threads, while the Intel Xeon Gold 6346 has 16 cores and 32 threads. The Threadripper therefore offers 50% more cores and 50% more threads.

Q: How big is the performance gap in multi-core workloads?

A: Across all six head-to-head Cinebench tests, the AMD Ryzen Threadripper 3960X wins by a consistent 47.1%. For example, in Cinebench R23 multi-core, the Threadripper scores 46557 versus the Xeon's 31650.

Q: What about single-core performance?

A: The Threadripper 3960X also dominates single-core tests. In Cinebench R23 single-core, it scores 6572 versus the Xeon Gold 6346's 4468, a 47.1% advantage. This is larger than the typical generational gap you would see between two mainstream desktop CPUs.

Q: Do these CPUs support ECC memory?

A: Only the Intel Xeon Gold 6346 supports ECC memory. The AMD Ryzen Threadripper 3960X does not list ECC as a supported feature.

Q: Which processor has higher memory bandwidth?

A: The Intel Xeon Gold 6346 has a memory bandwidth of 204.8 GB/s through its eight-channel memory bus. The AMD Ryzen Threadripper 3960X has 102.4 GB/s via quad-channel memory, exactly half the Xeon's bandwidth.

Q: Are both processors still in production?

A: Yes, both the AMD Ryzen Threadripper 3960X (release date 2019-11-24) and the Intel Xeon Gold 6346 (release date 2021-04-05) carry a production status of "Active."

Architecture Differences

The two processors come from fundamentally different design philosophies. The AMD Ryzen Threadripper 3960X is built on the Zen 2 architecture (codename Castle Peak) using a 7 nm process from TSMC. It packages 15,200 million transistors across a 4x 74 mm² die configuration. This chiplet-based design allows AMD to scale core counts aggressively while keeping manufacturing yields manageable.

The Intel Xeon Gold 6346, in contrast, uses the Ice Lake-SP architecture (codename Ice Lake) on Intel's 10 nm process. Intel does not list transistor count or die size in the data, but the monolithic design approach differs from AMD's chiplet strategy. The Xeon has a simpler cache hierarchy: 64 KB L1 per core, 1 MB L2 per core, and a shared 36 MB L3. The Threadripper has 64 KB L1 and 512 KB L2 per core, but a massive 128 MB L3 cache, which is more than three times the Xeon's total.

Memory architecture also diverges sharply. The Threadripper uses quad-channel DDR4 with 102.4 GB/s bandwidth, while the Xeon Gold 6346 uses eight-channel DDR4 with 204.8 GB/s. The Xeon also supports ECC memory, which the Threadripper does not. Both support PCIe Gen 4, but the Xeon specifies 64 lanes (CPU only), whereas the Threadripper's lane count is not listed in the data.

The socket platforms are incompatible: AMD Socket TRX4 for the Threadripper versus Intel Socket 4189 for the Xeon. The Threadripper has an unlocked multiplier, making it overclockable, while the Xeon Gold 6346 is locked. The Threadripper also targets the desktop market segment, whereas the Xeon is explicitly a server/workstation part.

Where Each One Wins

The benchmark data is unambiguous: the AMD Ryzen Threadripper 3960X wins every single head-to-head test, totaling 6 wins to 0. This sweep covers both multi-core and single-core workloads across Cinebench R15, R20, and R23.

In multi-core rendering and compute-heavy tasks, the Threadripper's extra 8 cores and 16 threads translate directly into a 47.1% lead. Cinebench R23 multi-core shows 46557 versus 31650, which is a massive margin for anyone doing CPU-based rendering, video encoding, or scientific simulation.

The Threadripper also wins in single-core performance, which is more surprising given the Xeon's newer 2021 release date. The Threadripper's 4.50 GHz boost clock versus the Xeon's 3.60 GHz boost clock explains this advantage. For lightly-threaded applications like legacy software, spreadsheet macros, or general desktop responsiveness, the Threadripper is clearly superior.

The Intel Xeon Gold 6346 has no benchmark wins in this dataset. Its strengths lie elsewhere: ECC memory support and double the memory bandwidth (204.8 GB/s versus 102.4 GB/s). These features matter for mission-critical servers where data integrity and memory throughput are paramount, even if raw compute performance lags.

The Xeon also has a lower TDP of 205 watts versus the Threadripper's 280 watts, which could matter in dense server environments where thermal management is a constraint. However, the performance-per-watt picture is not favorable to Intel given the 47.1% performance deficit.

Specification Differences

| Specification | AMD Ryzen Threadripper 3960X | Intel Xeon Gold 6346 |

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

| Cores | 24 | 16 |

| Threads | 48 | 32 |

| Base Clock | 3.80 GHz | 3.10 GHz |

| Boost Clock | 4.50 GHz | 3.60 GHz |

| TDP | 280 W | 205 W |

| Socket | AMD Socket TRX4 | Intel Socket 4189 |

| Architecture | Zen 2 | Ice Lake |

| Codename | Castle Peak | Ice Lake-SP |

| Process Node | 7 nm | 10 nm |

| Foundry | TSMC | Intel |

| Transistors | 15,200 million | Not listed |

| Die Size | 4x 74 mm² | Not listed |

| L1 Cache | 64 KB (per core) | 64 KB (per core) |

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

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

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

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

| ECC Memory | No | Yes |

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

| Market Segment | Desktop | Server/Workstation |

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

| Multiplier Unlocked | Yes | No |

| Part Number | 100-000000010100-100000010WOF | SRKHNCD8068904570201 |

Head-to-Head Benchmarks

The head-to-head data shows a complete sweep for the AMD Ryzen Threadripper 3960X. Every single test, whether multi-core or single-core, in Cinebench R15, R20, or R23, yields a 47.1% advantage for AMD. This consistency is remarkable — it suggests the performance gap is structural rather than workload-specific.

Starting with Cinebench R15 multi-core, the Threadripper scores 4692 versus the Xeon's 3190. That is a 1502-point gap, which represents roughly 47% more rendering throughput. The single-core R15 test shows 662 versus 450, a 212-point difference at the same 47.1% delta.

Cinebench R20 follows the identical pattern. Multi-core: 19553 versus 13293, a 6260-point gap. Single-core: 2760 versus 1876, an 884-point gap. Again, exactly 47.1%.

Cinebench R23, the most modern test in the dataset, confirms the trend. Multi-core: 46557 versus 31650, a 14907-point difference. Single-core: 6572 versus 4468, a 2104-point difference. The 47.1% delta persists across every generation of the Cinebench benchmark.

Looking at the broader benchmark context, the Threadripper 3960X also has data for 3DMark and Geekbench tests that the Xeon Gold 6346 lacks. The Threadripper scores 13978 in 3DMark max threads, 9989 in 3DMark 16 threads, and 13326 in Geekbench multi-core. The Xeon has no comparable scores in this dataset, so direct comparison is not possible for those workloads.

The average benchmark scores tell a similar story. The Threadripper averages 9284 across its benchmark suite, while the Xeon averages 9155. Both sit at the 65th percentile of all CPUs, but the Threadripper's nearest rivals include the Intel Xeon Platinum 8280M (deltaPct 0.3) and Intel Core i5-10300H (deltaPct 0.4), while the Xeon's closest competitor is the Intel Xeon Gold 5320 (deltaPct -0.9). The Threadripper's average is higher, but both are competitive within their respective peer groups.

The Verdict

The data makes a decisive case: the AMD Ryzen Threadripper 3960X is the faster processor in every benchmark measured. With a 47.1% lead across all six Cinebench tests, it is not a close contest. Anyone prioritizing raw compute performance — rendering, encoding, compilation, or simulation — should choose the Threadripper without hesitation.

The Threadripper's advantages go beyond core count. Its 4.50 GHz boost clock versus the Xeon's 3.60 GHz gives it a single-core edge that is rare for a 24-core part. Its 128 MB L3 cache dwarfs the Xeon's 36 MB, which helps in cache-sensitive workloads. And its unlocked multiplier means overclocking is possible, which the locked Xeon cannot offer.

The Intel Xeon Gold 6346 is not without merit, but its strengths are not visible in the benchmark data. It supports ECC memory, which the Threadripper does not. Its eight-channel memory bus delivers 204.8 GB/s of bandwidth, exactly double the Threadripper's 102.4 GB/s. For workloads that are memory-bandwidth-bound rather than compute-bound — certain database workloads, large-scale data analytics, or high-throughput networking — the Xeon's memory subsystem could be the deciding factor.

The Xeon also consumes less power on paper, with a 205 W TDP versus the Threadripper's 280 W. In a dense server environment with many sockets, that 75-watt difference per CPU could reduce cooling requirements and power bills. The Xeon's server/workstation market segment and ECC support make it the safer choice for enterprise deployment where data corruption is unacceptable.

However, for a workstation builder who wants maximum performance per socket, the Threadripper 3960X is the clear winner. It was released earlier (2019 versus 2021) yet still outperforms the Xeon in every test. The 24-core/48-thread configuration offers more parallel throughput, and the higher clock speeds ensure snappy single-threaded performance. The Threadripper is the better choice for content creators, engineers, and researchers who need raw compute power. The Xeon Gold 6346 is the better choice for enterprise buyers who need ECC reliability, higher memory bandwidth, and lower power draw in a server chassis. Choose based on your workload's bottleneck: compute or memory integrity.