AMD Ryzen Threadripper PRO 3995WX vs Intel Xeon 6740E Comparison

FAQ

Q: Which processor has more cores, and how does that affect multithreaded performance?

A: The Intel Xeon 6740E has 96 cores and 96 threads, while the AMD Ryzen Threadripper PRO 3995WX has 64 cores and 128 threads. Despite the core deficit, the Threadripper wins 13 of 17 head-to-head benchmarks, including all Cinebench versions, where it leads by exactly 9.2% in each test.

Q: What is the single-thread performance gap between the two?

A: The AMD part is substantially ahead in single-core work. In Cinebench R23 single-core, it scores 10071 vs 9140, a 9.2% lead. In PassMark single-thread, the gap widens to 23% (2595 vs 1997). The Threadripper’s 4.20 GHz boost clock versus 3.20 GHz on the Xeon largely explains this.

Q: Are there any workloads where the Intel Xeon 6740E wins decisively?

A: Yes, in four specific areas. The Xeon wins PassMark physics by 38.4% (7608 vs 5498), random string sorting by 17% (220684 vs 188574), floating point math by 11.4% (309333 vs 277715), and data encryption by 10.2% (137106 vs 124433).

Q: How do the two compare in memory bandwidth?

A: The Xeon 6740E supports DDR5 with eight-channel memory and 409.6 GB/s bandwidth. The Threadripper PRO 3995WX uses DDR4, also eight-channel, but with 204.8 GB/s — exactly half the Xeon’s bandwidth.

Q: Which processor has a higher average benchmark score?

A: The Xeon 6740E has an average benchmark score of 187718, ranking in the 99th percentile of all CPUs. The Threadripper PRO 3995WX averages 171748, placing in the 98th percentile. The Xeon’s nearest rival data shows it is 1.4% ahead of the AMD EPYC 8534P and 2.7% ahead of the Threadripper PRO 9975WX.

Q: What are the release dates and launch MSRPs?

A: The Intel Xeon 6740E was released on 2024-06-02 with a launch MSRP of $5265. The AMD Ryzen Threadripper PRO 3995WX was released on 2020-07-13 with a launch MSRP of $5489.

Architecture Differences

The Intel Xeon 6740E is built on Sierra Forest architecture, Intel’s Xeon 6 generation, fabricated on a 5 nm process at Intel’s foundry. It uses a die size of 578 mm². The AMD Ryzen Threadripper PRO 3995WX is based on Zen 2 architecture (Castle Peak codename), produced on TSMC’s 7 nm process, using eight chiplets of 74 mm² each, totaling 30,400 million transistors.

The core organization differs fundamentally. The Xeon 6740E has 96 cores with 96 threads—no simultaneous multithreading—and each core has 96 KB of L1 cache, with 4 MB L2 per module and 96 MB shared L3. The Threadripper PRO 3995WX has 64 cores but 128 threads, with 64 KB L1 per core, 512 KB L2 per core, and a larger 256 MB L3 cache.

Memory architecture is a major differentiator. The Xeon uses DDR5 with eight-channel support and 409.6 GB/s bandwidth. The Threadripper uses DDR4, also eight-channel, but halves the bandwidth to 204.8 GB/s. Both support ECC memory. PCIe lanes also differ: the Xeon provides Gen 5 with 88 lanes (CPU only), while the Threadripper offers Gen 4 with 128 lanes (CPU only).

The Xeon targets the server/workstation segment with a 250 W TDP and Intel Socket 4710. The Threadripper is a desktop part with a 280 W TDP on AMD Socket WRX8. The Xeon has no integrated graphics; the Threadripper’s integrated graphics field is null. Neither has an unlocked multiplier. The Xeon’s base clock is 2.40 GHz with a 3.20 GHz boost; the Threadripper runs at 2.70 GHz base and 4.20 GHz boost.

Head-to-Head Benchmarks

The AMD Ryzen Threadripper PRO 3995WX dominates the Cinebench suite with perfect consistency. Across all six Cinebench tests—R15, R20, and R23, both multi-core and single-core—it beats the Xeon 6740E by exactly 9.2%. For instance, in Cinebench R23 multi-core, the Threadripper scores 71338 versus 64741, and in R23 single-core it scores 10071 versus 9140. This uniform delta suggests a clock-speed advantage that scales evenly, not a workload-specific edge.

The Threadripper also wins PassMark multithread by 9.2% (83928 vs 76167), mirroring the Cinebench results. In integer math, it leads 490116 vs 457614, a 6.6% margin. Extended instructions show the largest AMD win: 106423 vs 78968, a 25.8% gap. The Threadripper also edges ahead in find prime numbers (566 vs 526, +7.1%), data compression (1841292 vs 1786845, +3%), and single-thread (2595 vs 1997, +23%).

The Intel Xeon 6740E’s wins are fewer but sharply defined. PassMark physics is the standout: 7608 vs 5498, a massive 38.4% advantage. Random string sorting goes to Intel by 17% (220684 vs 188574). Floating point math favors the Xeon at 309333 vs 277715, an 11.4% lead. Data encryption also goes Intel’s way, 137106 vs 124433, a 10.2% margin.

The win count is lopsided—AMD takes 13 benchmarks, Intel takes 4—but the nature of the wins matters. The Xeon’s victories are in specialized workloads where its architecture or memory bandwidth provides a clear edge, while the Threadripper’s wins are broad-based across rendering, encryption (where it loses), and general compute.

Specification Differences

| Specification | Intel Xeon 6740E | AMD Ryzen Threadripper PRO 3995WX |

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

| Cores | 96 | 64 |

| Threads | 96 | 128 |

| Base Clock | 2.40 GHz | 2.70 GHz |

| Boost Clock | 3.20 GHz | 4.20 GHz |

| TDP | 250 W | 280 W |

| Socket | Intel Socket 4710 | AMD Socket WRX8 |

| Architecture | Sierra Forest | Zen 2 |

| Process Node | 5 nm | 7 nm |

| Foundry | Intel | TSMC |

| Die Size | 578 mm² | 8x 74 mm² |

| L1 Cache | 96 KB (per core) | 64 KB (per core) |

| L2 Cache | 4 MB (per module) | 512 KB (per core) |

| L3 Cache | 96 MB (shared) | 256 MB |

| Memory Support | DDR5 | DDR4 |

| Memory Bandwidth | 409.6 GB/s | 204.8 GB/s |

| PCIe | Gen 5, 88 Lanes | Gen 4, 128 Lanes |

| Release Date | 2024-06-02 | 2020-07-13 |

| Market Segment | Server/Workstation | Desktop |

The Xeon has more cores but fewer threads, a higher process node (5 nm vs 7 nm), and double the memory bandwidth. The Threadripper has higher clocks, a larger L3 cache (256 MB vs 96 MB), and more PCIe lanes (128 vs 88), albeit on Gen 4 instead of Gen 5.

Where Each One Wins

AMD Ryzen Threadripper PRO 3995WX wins for general multithreaded productivity. The data is unambiguous: it leads in every Cinebench test by 9.2%, in PassMark multithread by 9.2%, and in integer math by 6.6%. For rendering, video encoding, and CPU-bound compilation workloads that scale with threads and clock speed, the Threadripper’s combination of 128 threads and a 4.20 GHz boost clock outperforms the Xeon’s 96 threads at 3.20 GHz. Its 25.8% lead in extended instructions also suggests an advantage in SIMD-heavy and specialized instruction workloads.

Intel Xeon 6740E wins for memory-bandwidth-sensitive and floating-point-heavy workloads. The Xeon’s 409.6 GB/s memory bandwidth, double the Threadripper’s, is the likely driver behind its 38.4% win in physics and 17% win in random string sorting—both of which are memory-latency and bandwidth-sensitive. Its 11.4% win in floating point math indicates strong FP throughput. Data encryption also goes to Intel by 10.2%, which may reflect the newer architecture’s cryptographic instruction efficiency.

For single-threaded performance, the Threadripper is the clear choice. The 23% lead in PassMark single-thread and 9.2% lead in Cinebench R23 single-core show that applications relying on one or two fast cores will run noticeably better on the AMD part. The Xeon’s lower boost clock of 3.20 GHz simply cannot compete.

For server deployments with high memory demands, the Xeon is better positioned. The eight-channel DDR5 support with 409.6 GB/s bandwidth, combined with 96 cores, makes it suited for data-intensive server workloads like in-memory databases or scientific simulations that saturate memory interfaces. The Threadripper’s DDR4 bandwidth at 204.8 GB/s is a structural disadvantage in such scenarios.

For PCIe expansion, the Threadripper offers more lanes. With 128 Gen 4 lanes versus the Xeon’s 88 Gen 5 lanes, the AMD part supports more add-in cards or storage devices simultaneously, though the Xeon’s lanes are faster per lane. The tradeoff depends on whether quantity or per-lane speed matters more for the intended workload.

The overall benchmark record favors the Threadripper in raw performance terms, but the Xeon’s wins are in areas where its platform advantages—memory bandwidth and floating-point throughput—are architecturally deliberate. The right choice depends entirely on whether the workload is general compute (AMD) or memory-bound and FP-heavy (Intel).