AMD Ryzen Threadripper PRO 3995WX vs Intel Xeon 674X Comparison

Where Each One Wins

The benchmark split between these two workstation processors is unusually clean. The AMD Ryzen Threadripper PRO 3995WX takes 6 of the 16 head-to-head tests, while the Intel Xeon 674X wins 10. But the wins are not evenly distributed across workload types. The AMD part dominates memory-bandwidth-sensitive and integer-heavy tasks, while the Intel part wins everything that rewards high clock speeds and modern instruction efficiency.

The AMD Ryzen Threadripper PRO 3995WX is the clear choice for data compression and encryption workloads. It beats the Intel Xeon 674X by 48.9% in Passmark data compression (1,841,292 vs 1,236,272) and by an enormous 103.3% in data encryption (124,433 vs 61,195). These are workloads that scale with core count and memory bandwidth, and the 64-core AMD part simply has more resources to throw at them. Integer math also strongly favors AMD: 490,116 vs 308,968, a 58.6% advantage. Random string sorting goes AMD's way by 47.9% (188,574 vs 127,529), and floating point math shows a 13.9% lead (277,715 vs 243,877). Extended instructions round out the AMD wins at 9.3% (106,423 vs 97,373).

The Intel Xeon 674X, meanwhile, wins every Cinebench test by a small margin. In Cinebench R15 multicore it scores 7,213 vs 7,190 (0.3% ahead), and in R23 multicore it scores 71,566 vs 71,338, again 0.3% ahead. Single-core Cinebench results are similarly close: R15 single-core 1,018 vs 1,014 (0.4% ahead), R20 single-core 4,243 vs 4,229 (0.3% ahead). The Intel part also wins Passmark multithread with 84,196 vs 83,928, a 0.3% margin. The bigger Intel wins come in single-threaded performance: Passmark single-thread 3,933 vs 2,595, a 34% advantage. Physics simulations heavily favor Intel: 7,586 vs 5,498, a 27.5% lead. Prime number finding also goes Intel's way: 693 vs 566, an 18.3% margin.

The pattern is clear. AMD wins where core count and parallel throughput matter. Intel wins where per-thread speed matters and in mixed workloads where the margin is slim. If your work involves heavy encryption, compression, or integer processing, the AMD part is dramatically faster. If your work is single-threaded or physics-based, the Intel part has a decisive edge.

The Verdict

Pick the AMD Ryzen Threadripper PRO 3995WX if your workloads are massively parallel and memory-bandwidth hungry. The data shows it is more than twice as fast in encryption, 58.6% faster in integer math, and 48.9% faster in compression. These are not small margins; they represent real time savings in production environments. The 64 cores and 128 threads, combined with 256 MB of L3 cache, make it the superior choice for database work, compression pipelines, and cryptographic operations.

Pick the Intel Xeon 674X if your workloads are single-threaded, physics-based, or require the fastest possible per-core response. The 34% lead in Passmark single-thread performance is substantial, as is the 27.5% advantage in physics. The Intel part also edges out the AMD part in every Cinebench test, though by only 0.3% to 0.4%. For rendering tasks measured by Cinebench, the two are effectively tied. The Intel part achieves this with 28 cores and 56 threads, less than half the core count of the AMD part, which speaks to the efficiency of its Granite Rapids architecture and its 4.90 GHz boost clock.

For most mixed workloads, the database suggests a near tie. The average benchmark score for the AMD part is 171,748, while the Intel part averages 143,103. However, the Intel part's nearest rivals include the Intel Xeon 6732P (0.2% behind), AMD Ryzen 9 PRO 9965X3D (0.4% behind), and Intel Xeon w9-3575X (0.8% behind), showing it sits in a competitive cluster. The AMD part sits among EPYC-class parts: 2.4% ahead of the EPYC 7C13, 5.7% ahead of the EPYC 9375F, but 2.5% behind the Xeon 6740P and 4.5% behind the EPYC 7763.

Head-to-Head Benchmarks

The largest single victory belongs to the AMD part in data encryption. The 103.3% delta means the Threadripper PRO 3995WX delivers more than double the encryption throughput of the Xeon 674X. This is the kind of result that changes hardware purchasing decisions for security-focused workloads.

The second-largest AMD win is integer math at 58.6%. The 490,116 score versus 308,968 reflects the raw parallel integer throughput of 64 Zen 2 cores. Data compression follows at 48.9%, and random string sorting at 47.9%. These four workloads alone justify the AMD part for data-heavy server roles.

The Intel part's biggest win is Passmark single-thread at 34% (3,933 vs 2,595). This is a massive per-core advantage, driven by the 4.90 GHz boost clock versus 4.20 GHz on the AMD part, and the newer 5 nm process. Physics shows a 27.5% Intel advantage (7,586 vs 5,498), and prime number finding shows 18.3% (693 vs 566). The extended instructions test goes to AMD by 9.3%, and floating point math to AMD by 13.9%.

The Cinebench results are notable for their closeness. Across R15, R20, and R23, both multicore and single-core, the Intel part wins by between 0.3% and 0.4%. These are within run-to-run variation and should be treated as statistical ties. The same applies to Passmark multithread, where Intel leads by 0.3% (84,196 vs 83,928). The AMD part achieves parity in these tests despite having 64 cores versus 28, which indicates the Intel cores are doing significantly more work per thread.

FAQ

Q: Which processor is faster in Cinebench R23 multicore?

A: The Intel Xeon 674X scores 71,566 versus 71,338 for the AMD Ryzen Threadripper PRO 3995WX, a 0.3% difference. For practical purposes, these are tied in this workload.

Q: How much faster is the AMD part in encryption?

A: The AMD Ryzen Threadripper PRO 3995WX scores 124,433 in Passmark data encryption versus 61,195 for the Intel Xeon 674X. That is a 103.3% advantage, meaning the AMD part is more than twice as fast.

Q: What is the single-thread performance gap?

A: The Intel Xeon 674X scores 3,933 in Passmark single-thread versus 2,595 for the AMD part, a 34% lead. The Intel part also leads in Cinebench R20 single-core with 4,243 versus 4,229, a 0.3% margin.

Q: Which processor has more cores and threads?

A: The AMD Ryzen Threadripper PRO 3995WX has 64 cores and 128 threads. The Intel Xeon 674X has 28 cores and 56 threads.

Q: How do their memory systems differ?

A: The AMD part supports DDR4 with eight-channel memory and 204.8 GB/s bandwidth. The Intel part supports DDR5 with eight-channel memory and 409.6 GB/s bandwidth, double the bandwidth.

Q: Which CPU has a higher boost clock?

A: The Intel Xeon 674X boosts to 4.90 GHz, while the AMD Ryzen Threadripper PRO 3995WX boosts to 4.20 GHz.

Architecture Differences

The two processors come from different design philosophies. The AMD Ryzen Threadripper PRO 3995WX uses the Zen 2 architecture, codenamed Castle Peak, built on a 7 nm process at TSMC. It packs 30,400 million transistors across 8 dies, each 74 mm². The design is chiplet-based, with a 256 MB L3 cache that is shared across the entire processor. Each core gets 64 KB of L1 and 512 KB of L2. The AMD part uses the AMD Socket WRX8 and is classified as a Desktop processor, though it is clearly workstation-oriented.

The Intel Xeon 674X uses the Granite Rapids architecture, also its codename, built on a 5 nm process at Intel. The die size is 2x 598 mm², indicating a monolithic dual-die design. L1 cache is 112 KB per core, L2 is 2 MB per core, and L3 is 144 MB shared. The Intel part uses Intel Socket 4710 and is classified as a Server/Workstation processor. It has an integrated graphics controller listed as N/A, while the AMD part has no integrated graphics at all.

Memory architecture differs significantly. The AMD part uses DDR4 with eight-channel support and 204.8 GB/s bandwidth. The Intel part uses DDR5 with eight-channel support and 409.6 GB/s bandwidth. Both support ECC memory. PCIe connectivity is the same generation count: the AMD part is Gen 4 with 128 lanes (CPU only), and the Intel part is Gen 5 with 128 lanes (CPU only).

The production status for both is Active. The AMD part was released on 2020-07-13, while the Intel part has a release date of 2026-02-01. The AMD part has a locked multiplier, while the Intel part has an unlocked multiplier. The AMD part's launch MSRP is $5489, and the Intel part's launch MSRP is $2199.

Specification Differences

| Specification | AMD Ryzen Threadripper PRO 3995WX | Intel Xeon 674X |

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

| Manufacturer | AMD | Intel |

| Cores | 64 | 28 |

| Threads | 128 | 56 |

| Base clock | 2.70 GHz | 3.00 GHz |

| Boost clock | 4.20 GHz | 4.90 GHz |

| TDP | 280 W | 270 W |

| Socket | AMD Socket WRX8 | Intel Socket 4710 |

| Architecture | Zen 2 | Granite Rapids |

| Codename | Castle Peak | Granite Rapids |

| Process node | 7 nm | 5 nm |

| Foundry | TSMC | Intel |

| Transistors | 30,400 million | Not disclosed |

| Die size | 8x 74 mm² | 2x 598 mm² |

| L1 cache | 64 KB per core | 112 KB per core |

| L2 cache | 512 KB per core | 2 MB per core |

| L3 cache | 256 MB | 144 MB shared |

| Memory support | DDR4 | DDR5 |

| Memory bus | Eight-channel | Eight-channel |

| Memory bandwidth | 204.8 GB/s | 409.6 GB/s |

| ECC memory | Yes | Yes |

| PCIe | Gen 4, 128 lanes (CPU only) | Gen 5, 128 lanes (CPU only) |

| Integrated graphics | None | N/A |

| Market segment | Desktop | Server/Workstation |

| Multiplier unlocked | No | Yes |

| Release date | 2020-07-13 | 2026-02-01 |

| Launch MSRP | $5489 | $2199 |

The core count difference is the most obvious: 64 versus 28. The Intel part compensates with higher clocks (3.00 GHz base, 4.90 GHz boost versus 2.70 GHz and 4.20 GHz) and a newer process node. The AMD part has a larger total L3 cache (256 MB versus 144 MB), but the Intel part has more L2 per core (2 MB versus 512 KB). The Intel part has double the memory bandwidth (409.6 GB/s versus 204.8 GB/s) despite using the same eight-channel configuration, due to DDR5 versus DDR4. PCIe generation also differs: Gen 5 on Intel versus Gen 4 on AMD, both with 128 lanes. The Intel part is unlocked for overclocking, while the AMD part is not.