AMD Ryzen Threadripper 3960X vs Intel Xeon Gold 5320 Comparison

The AMD Ryzen Threadripper 3960X and Intel Xeon Gold 5320 occupy the same percentile rank among all CPUs (65th), yet their benchmark profiles could hardly be more distinct. The data shows a decisive sweep in head-to-head testing, with the AMD part winning all six shared benchmarks by a nearly uniform margin, while the Intel part counters with architectural advantages in platform capabilities rather than raw compute speed.

Head-to-Head Benchmarks

The head-to-head results are unambiguous: the AMD Ryzen Threadripper 3960X wins all six shared tests. In Cinebench R15 multicore, the AMD scores 4692 against Intel’s 3217, a delta of 45.9%. The single-core R15 test tells the same story: 662 versus 454, a 45.8% advantage. Moving to Cinebench R20, the AMD part posts 19553 in multicore and 2760 in single-core, while the Xeon Gold 5320 manages 13408 and 1892 respectively — again deltas of 45.8% and 45.9%. The pattern holds in Cinebench R23: AMD leads 46557 to 31924 in multicore and 6572 to 4506 in single-core, both at 45.8% margins.

These deltas are remarkably consistent, hovering within 0.1 percentage points of 45.8% across every test. That uniformity suggests the performance gap is structural — rooted in per-core efficiency and clock rates — rather than workload-specific. The AMD processor’s base clock of 3.80 GHz and boost of 4.50 GHz dwarf the Intel part’s 2.20 GHz base and 3.40 GHz boost. Even though the Xeon Gold 5320 has more cores (26 versus 24) and more threads (52 versus 48), it cannot overcome the clock disadvantage. In every multicore test, the AMD’s 24 cores outproduce Intel’s 26 cores, which means the per-core throughput advantage is substantial.

Interestingly, the average benchmark scores place these two very close overall. The AMD Ryzen Threadripper 3960X has an avgBenchmarkScore of 9284, while the Intel Xeon Gold 5320 sits at 9234 — a difference of just 0.5%. This narrow gap in aggregate metrics contrasts sharply with the lopsided head-to-head results. The explanation lies in the benchmark sets: the AMD part’s average includes 3DMark tests (single-thread 725, 2-thread 1436, 4-thread 2775, 8-thread 5267, 16-thread 9989, max-threads 13978) and Geekbench scores (single-core 1681, multicore 13326), while the Intel part only lists Cinebench results. The 3DMark and Geekbench numbers pull the AMD average up, masking the fact that in shared workloads, it wins every time.

Where Each One Wins

The AMD Ryzen Threadripper 3960X wins in every compute-heavy scenario captured by the benchmark suite. Cinebench R15, R20, and R23 — both multicore and single-core — all go to the AMD part. For users running rendering, simulation, or any workload that scales with Cinebench-style parallelism, the data is clear: the AMD delivers 45.8% higher throughput in multicore R23 (46557 versus 31924) and 45.8% higher single-core performance (6572 versus 4506). The 3DMark results, available only for the AMD, further demonstrate its strength in threaded graphics workloads — max-threads score of 13978 and 16-thread score of 9989 indicate solid scaling across its 48 threads.

The Intel Xeon Gold 5320 wins zero head-to-head benchmarks. Its advantages are not in raw compute speed but in platform specifications. It supports ECC memory (the AMD does not), which is critical for error-sensitive server and workstation environments. It also offers eight-channel memory with 191.6 GB/s bandwidth versus the AMD’s quad-channel 102.4 GB/s — a 87.1% bandwidth advantage that can benefit memory-bound workloads even if those workloads are not represented in the shared benchmark list. The Intel part also has more PCIe Gen 4 lanes (64 CPU-only) compared to the AMD’s unspecified lane count, which matters for multi-GPU or high-speed storage configurations. Additionally, the Xeon’s 26 cores and 52 threads provide more raw parallelism than the AMD’s 24 and 48, even though the AMD uses its fewer resources more efficiently.

Architecture Differences

The two processors come from fundamentally different design philosophies. The AMD Ryzen Threadripper 3960X uses Zen 2 architecture on a 7 nm process from TSMC, with a 15,200 million transistor count spread across a 4x 74 mm² die configuration. It is built on the Castle Peak codename and fits an AMD Socket TRX4. The Intel Xeon Gold 5320 uses Ice Lake architecture (Ice Lake-SP codename) on Intel’s 10 nm process, fitting an Intel Socket 4189. The AMD is a desktop-market part with an unlocked multiplier, while the Intel is a server/workstation part with a locked multiplier.

Cache configurations differ notably. Both have 64 KB of L1 per core. The L2 cache diverges: the AMD provides 512 KB per core, while the Intel provides 1 MB per core — double the per-core L2. The L3 cache is where the AMD dominates: 128 MB total versus the Intel’s 39 MB shared. This 128 MB L3 is a hallmark of the Threadripper design, giving the AMD a massive caching advantage for working sets that fit within it. The Intel’s smaller L3 is partially offset by its larger L2, but the AMD’s total cache hierarchy is far larger.

Memory support also reveals different priorities. The AMD supports DDR4 with a quad-channel bus and 102.4 GB/s bandwidth, with no ECC support. The Intel supports DDR4 with an eight-channel bus and 191.6 GB/s bandwidth, with ECC support. The Intel’s memory bandwidth is nearly double, which is a standard feature for server platforms handling large data sets. Both support PCIe Gen 4, but the Intel explicitly lists 64 lanes (CPU only) while the AMD does not specify a lane count in the data.

Release dates are separated by roughly 17 months: the AMD launched on 2019-11-24, and the Intel on 2021-04-05. The AMD’s launch MSRP was $1399, while the Intel has no listed launch MSRP. The AMD’s production status is Active, as is the Intel’s.

FAQ

Q: Which processor has more cores?

A: The Intel Xeon Gold 5320 has 26 cores and 52 threads, while the AMD Ryzen Threadripper 3960X has 24 cores and 48 threads.

Q: Does the AMD processor support ECC memory?

A: No. The AMD Ryzen Threadripper 3960X has eccMemory set to false, while the Intel Xeon Gold 5320 supports ECC memory.

Q: What is the difference in memory bandwidth?

A: The Intel Xeon Gold 5320 offers 191.6 GB/s over an eight-channel bus, compared to the AMD’s 102.4 GB/s over a quad-channel bus.

Q: Which CPU wins in single-core Cinebench R23?

A: The AMD Ryzen Threadripper 3960X scores 6572 versus the Intel’s 4506, a 45.8% advantage.

Q: Are these CPUs in the same performance percentile?

A: Yes, both have a percentileVsAllCpus of 65, and their average benchmark scores are within 0.5% of each other (9284 for AMD, 9234 for Intel).

Q: What is the process node for each?

A: The AMD uses a 7 nm process from TSMC, while the Intel uses a 10 nm process from Intel.

The Verdict

The data supports a straightforward conclusion for compute-bound workloads: the AMD Ryzen Threadripper 3960X is the faster processor. It wins every shared benchmark by roughly 45.8%, regardless of thread count or test generation. Its higher clock speeds (3.80 GHz base, 4.50 GHz boost versus 2.20 GHz and 3.40 GHz) and larger L3 cache (128 MB versus 39 MB) give it a decisive edge in both single-threaded and multi-threaded performance. For rendering, simulation, or any workload that scales with Cinebench-style parallelism, the AMD delivers nearly half again as much performance.

The Intel Xeon Gold 5320 is the choice when platform features outweigh raw speed. It provides ECC memory support, which the AMD lacks, and its eight-channel memory bus delivers 191.6 GB/s — 87.1% more bandwidth than the AMD’s 102.4 GB/s. Its 64 PCIe Gen 4 lanes (CPU only) exceed the AMD’s unspecified count, and its 26 cores provide more parallelism than the AMD’s 24, even if the AMD uses its cores more efficiently. For memory-bandwidth-sensitive server workloads, database operations, or high-availability systems requiring ECC, the Intel part offers capabilities the AMD cannot match.

The average benchmark scores (9284 for AMD, 9234 for Intel) suggest near-parity overall, but that metric is misleading because the AMD’s average includes 3DMark and Geekbench tests absent from the Intel’s record. In direct comparison, the AMD wins 6 out of 6 tests. The verdict: pick the AMD Ryzen Threadripper 3960X for maximum compute throughput and per-core performance; pick the Intel Xeon Gold 5320 for ECC support, higher memory bandwidth, and server-class platform features.

Specification Differences

| Field | AMD Ryzen Threadripper 3960X | Intel Xeon Gold 5320 |

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

| Cores | 24 | 26 |

| Threads | 48 | 52 |

| Base Clock | 3.80 GHz | 2.20 GHz |

| Boost Clock | 4.50 GHz | 3.40 GHz |

| TDP | 280 W | 185 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 specified |

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

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

| L3 Cache | 128 MB | 39 MB (shared) |

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

| Memory Bandwidth | 102.4 GB/s | 191.6 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 |

| Launch MSRP | $1399 | Not specified |

| Multiplier Unlocked | Yes | No |