AMD Ryzen Threadripper PRO 9965WX vs Intel Xeon 676X Comparison

The Intel Xeon 676X and AMD Ryzen Threadripper PRO 9965WX are both flagship workstation processors, yet they approach performance from fundamentally different angles. The data reveals a clear split: the Threadripper dominates in single-threaded and rendering workloads, while the Xeon counters with decisive victories in floating-point math and physics simulations. With the AMD chip winning 11 of the 16 head-to-head benchmarks and the Intel chip taking 5, the choice between them hinges on whether your primary workloads favor raw core count and parallel throughput or specific compute patterns where Intel’s architecture holds a distinct edge.

Head-to-Head Benchmarks

The most striking margin in the entire dataset belongs to the Intel Xeon 676X in PassMark's floating-point math test, where it scores 283,570 against the Threadripper's 229,685. That is a 23.5% advantage, and it is not a minor outlier—it represents a fundamental strength in workloads that rely heavily on FPU throughput. Similarly, in PassMark physics, the Xeon posts 8,281 versus 7,529, a 10% lead that suggests better handling of simulation and physics-based calculations. These are not close calls; they are commanding wins that indicate the Xeon’s 32 cores are doing something different with floating-point data than the Threadripper's 24 cores.

However, the Threadripper PRO 9965WX answers back in the Cinebench suite with remarkable consistency. Across R15, R20, and R23, both single-core and multi-core tests show a uniform 4.4% delta in AMD's favor. The multi-core scores tell the story: 8,162 versus 7,806 in R15, 34,009 versus 32,527 in R20, and 80,976 versus 77,447 in R23. That consistent 4.4% gap across all six Cinebench tests suggests a clock-speed and IPC advantage that scales uniformly, rather than a workload-specific quirk. The Threadripper’s single-core wins are even more pronounced in PassMark, where it scores 4,551 versus 4,015—an 11.8% lead that is the largest single-thread delta in the comparison.

The remaining PassMark tests split the field. The Xeon wins data compression (1,355,807 versus 1,345,230, a 0.8% edge), data encryption (67,638 versus 66,155, a 2.2% lead), integer math (354,777 versus 349,195, a 1.6% margin), and physics as noted. The Threadripper counters with extended instructions (108,753 versus 105,231, a 3.2% win), find prime numbers (752 versus 738, a 1.9% edge), multithread (92,604 versus 91,115, a 1.6% lead), and random string sorting (149,617 versus 137,976, a substantial 7.8% win). The pattern is clear: AMD wins the synthetic and rendering benchmarks by small but consistent margins, while Intel wins the specialized compute tests by a wider spread.

Architecture Differences

The two processors could not be more different under the hood. The Intel Xeon 676X uses the Granite Rapids architecture, built on Intel's 5 nm process, and features a dual-die design with a combined die size of 2x 598 mm². It packs 32 cores and 64 threads, with a base clock of 2.80 GHz and a boost clock of 4.90 GHz. Its cache hierarchy is notably generous: 112 KB of L1 per core, 2 MB of L2 per core, and a massive 144 MB of shared L3 cache. The Threadripper PRO 9965WX, in contrast, uses AMD's Zen 5 architecture (codename Shimada Peak), fabricated on TSMC's 4 nm process. It has 24 cores and 48 threads, but compensates with much higher clocks: 4.20 GHz base and 5.40 GHz boost. Its cache is smaller per core—64 KB L1 and 1 MB L2—but the L3 is still substantial at 128 MB. The Threadripper also lists 33,260 million transistors across four 70.6 mm² chiplets, a stark contrast to Intel's monolithic-ish dual-die approach.

These architectural differences explain the benchmark results. The Threadripper’s higher boost clock (5.40 versus 4.90 GHz) and newer 4 nm process give it a clear single-thread advantage, which shows up in the 11.8% PassMark single-thread win and the 4.4% Cinebench single-core margins. The Xeon’s larger L3 cache (144 MB versus 128 MB) and higher core count (32 versus 24) do not translate into multi-core wins in Cinebench, suggesting that AMD’s Zen 5 IPC and clock speed more than offset the core deficit. However, the Xeon’s 23.5% floating-point lead hints at a different FPU design or better memory bandwidth utilization for those specific operations.

Both processors share several traits: eight-channel DDR5 memory support with 409.6 GB/s bandwidth, ECC memory, PCIe Gen 5 with 128 lanes (CPU only), no integrated graphics, and unlocked multipliers. They target the same server/workstation segment and are both active production parts. The Threadripper launched earlier (2025-07-22) with a launch MSRP of $2899, while the Xeon followed (2026-02-01) at $2499.

The Verdict

The benchmark data paints a nuanced picture that defies a simple "which is better" answer. If your workloads are dominated by Cinebench-style rendering, single-threaded performance, or general multi-threaded throughput, the AMD Ryzen Threadripper PRO 9965WX is the clear choice. It wins every Cinebench test by 4.4%, takes the single-thread crown by 11.8%, and also wins PassMark multithread (92,604 versus 91,115). For video editors, 3D artists, and developers who rely on these types of benchmarks, the Threadripper’s consistency is compelling.

On the other hand, the Intel Xeon 676X is the pick for specialized compute workloads. The 23.5% floating-point math lead and 10% physics win are not trivial—they suggest that simulations, scientific computing, and any FPU-heavy code will run significantly faster on the Xeon. The smaller wins in data compression (0.8%), encryption (2.2%), and integer math (1.6%) add to its profile as a general workload processor that excels at specific math operations. The Xeon also has 8 more cores and 16 more threads, which could matter for highly parallel tasks that scale perfectly with core count, even if Cinebench does not reflect that advantage.

The average benchmark scores tell a similar story: the Xeon averages 158,540 across all tests, while the Threadripper averages 147,009. That is a 7.3% difference in the Xeon’s favor, driven largely by its massive floating-point win. Yet the Threadripper holds a higher percentile in the specific tests it wins. The data suggests that for most users, the Threadripper is the safer all-rounder, but for those with FPU-intensive workloads, the Xeon is the specialized tool that delivers outsized gains.

Specification Differences

The following table highlights only the fields where the two processors differ:

| Specification | Intel Xeon 676X | AMD Ryzen Threadripper PRO 9965WX |

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

| Cores | 32 | 24 |

| Threads | 64 | 48 |

| Base Clock | 2.80 GHz | 4.20 GHz |

| Boost Clock | 4.90 GHz | 5.40 GHz |

| TDP | 275 W | 350 W |

| Socket | Intel Socket 4710 | AMD Socket sTR5 |

| Architecture | Granite Rapids | Zen 5 |

| Codename | Granite Rapids | Shimada Peak |

| Process Node | 5 nm | 4 nm |

| Foundry | Intel | TSMC |

| Transistors | N/A | 33,260 million |

| Die Size | 2x 598 mm² | 4x 70.6 mm² |

| L1 Cache | 112 KB (per core) | 64 KB (per core) |

| L2 Cache | 2 MB (per core) | 1 MB (per core) |

| L3 Cache | 144 MB (shared) | 128 MB |

| Launch MSRP | $2499 | $2899 |

| Part Number | SA2CY | 100-000000724 |

Both share DDR5 memory support, eight-channel memory bus, 409.6 GB/s bandwidth, ECC memory, PCIe Gen 5 (128 lanes), no integrated graphics, server/workstation market segment, active production status, and unlocked multipliers.

FAQ

Q: Which processor has more cores?

A: The Intel Xeon 676X has 32 cores and 64 threads, while the AMD Ryzen Threadripper PRO 9965WX has 24 cores and 48 threads.

Q: Does the higher core count translate into a multi-core benchmark win?

A: No. Despite having 8 more cores, the Xeon loses all three Cinebench multi-core tests by 4.4% (R15: 7,806 vs 8,162; R20: 32,527 vs 34,009; R23: 77,447 vs 80,976). The Threadripper also wins PassMark multithread by 1.6% (92,604 vs 91,115).

Q: What is the biggest single benchmark margin between the two?

A: The largest margin is in PassMark floating-point math, where the Intel Xeon 676X leads by 23.5% (283,570 vs 229,685).

Q: Which processor has the higher boost clock?

A: The AMD Ryzen Threadripper PRO 9965WX boosts to 5.40 GHz, compared to the Intel Xeon 676X's 4.90 GHz.

Q: Are both processors compatible with ECC memory?

A: Yes, both support ECC memory and use eight-channel DDR5 with the same 409.6 GB/s bandwidth.

Q: How do their average benchmark scores compare?

A: The Intel Xeon 676X has a higher average benchmark score of 158,540, while the Threadripper PRO 9965WX averages 147,009.

Where Each One Wins

The Intel Xeon 676X wins in five benchmark categories, and they cluster around specific compute types. The 23.5% floating-point math win is the headline, but the 10% physics lead and 2.2% encryption win are also notable. Data compression (0.8%) and integer math (1.6%) round out the Xeon’s wins with modest margins. This profile points to scientific computing, financial modeling, cryptography, and any simulation that relies heavily on floating-point arithmetic. The Xeon’s 144 MB L3 cache and 32 cores likely contribute to these wins, even if the overall thread throughput is not superior.

The AMD Ryzen Threadripper PRO 9965WX wins 11 benchmarks, covering a broader range of workloads. Its Cinebench sweep (all six tests by 4.4%) makes it the clear pick for 3D rendering and animation. The 11.8% single-thread win indicates better performance in lightly-threaded applications like legacy software or games. The 7.8% random string sorting win suggests an edge in data processing and database workloads, while extended instructions (3.2%) and prime number finding (1.9%) round out the list. The multithread win (1.6%) is less decisive but still favors AMD.

For a workstation that will handle a mix of rendering, general productivity, and single-threaded tasks, the Threadripper PRO 9965WX is the more versatile choice. Its wins are consistent and cover the most common professional workloads. However, for a system dedicated to FPU-heavy simulations or scientific computing, the Xeon 676X offers a 23.5% advantage that could significantly reduce job times. The data does not support a universal recommendation; it supports a workload-specific one.