AMD Ryzen 9 PRO 9965 vs AMD Ryzen 9 PRO 9965X3D Comparison

The AMD Ryzen 9 PRO 9965 and the AMD Ryzen 9 PRO 9965X3D are near-identical processors on paper, sharing the same 16-core/32-thread configuration, 4.30 GHz base clock, 5.50 GHz boost clock, and 170 W TDP. However, benchmark data reveals a clear division of labor: the standard 9965 wins the majority of compute-heavy workloads, while the X3D variant dominates in specific latency-sensitive tasks. With 8 wins for the 9965 against 3 for the 9965X3D in head-to-head tests, the data points to the non-X3D part as the more consistently versatile performer, though the X3D’s victories are dramatic enough to warrant serious consideration for specialized use cases. Both chips sit in the 98th percentile of all CPUs, indicating top-tier performance overall.

Head-to-Head Benchmarks

The most striking result in the comparison is the passmark_find_prime_numbers test, where the 9965X3D scores 599 against the 9965’s 364. This represents a 39.2% advantage for the X3D part, the single largest margin in any test. This workload is highly sensitive to cache latency and capacity, and the X3D’s doubled L3 cache (128 MB vs 64 MB) provides an enormous boost. Similarly, the passmark_physics test shows the X3D ahead with 4801 points versus 3256, a 32.2% lead. Physics simulations often rely on repeated access to working sets that fit within larger caches, explaining this substantial gap.

The 9965X3D also takes the passmark_multithread test, scoring 68225 against the 9965’s 66655, a 2.3% advantage. While modest, this result indicates that the larger cache helps even in highly parallel workloads, likely by reducing memory traffic stalls. These three wins for the X3D are consistent: they all involve either prime number generation, physics calculations, or general multithreaded throughput where cache residency matters.

Conversely, the standard 9965 wins every other head-to-head test, with margins ranging from 0.6% to 2.3%. The passmark_integer_math test shows the 9965 ahead at 243280 versus 237955, a 2.2% lead. passmark_floating_point_math also favors the 9965, with 160746 points against 157123, another 2.3% margin. In passmark_data_encryption, the 9965 scores 44920 versus 43905, a 2.3% advantage, and in passmark_data_compression, it leads with 908293 versus 895563, a 1.4% edge. The 9965 also wins passmark_random_string_sorting (94915 vs 92886, 2.2%), passmark_single_thread (4682 vs 4614, 1.5%), and passmark_extended_instructions (71210 vs 70797, 0.6%).

What the numbers reveal is that the X3D’s cache advantage is not universal. In arithmetic, encryption, compression, and sorting workloads, the standard 9965 consistently outperforms by small but repeatable margins. This suggests that while the larger L3 cache helps in specific access patterns, it may introduce slightly higher latency or power characteristics that disadvantage other workloads. The 9965’s wins are narrower than the X3D’s biggest victories, but they cover far more test categories, making the standard part the safer choice for mixed workloads.

The Verdict

The data supports a clear verdict: choose the AMD Ryzen 9 PRO 9965 for general-purpose workstation and server tasks, and the AMD Ryzen 9 PRO 9965X3D only if your primary workloads resemble prime-number sieving, physics simulation, or other cache-resident algorithms. The 9965 wins 8 of 11 head-to-head benchmarks, including all single-thread tests, all math tests, and all data-processing tests. Its 1.5% single-thread advantage and 2.2-2.3% margins in integer and floating-point math indicate that for most software, the standard part delivers equal or better performance without requiring any cache-specific optimization.

The 9965X3D, however, is the undisputed champion in its niche. The 39.2% lead in prime-number finding and 32.2% lead in physics are not incremental improvements; they are transformative for those specific tasks. If your workflow consists of running the same bounded dataset repeatedly, the X3D’s 128 MB L3 cache will yield outsized gains. But for mixed, unpredictable workloads, the 9965’s consistent wins across 8 different tests make it the more reliable choice. The X3D’s multithread win (2.3%) is notable, but it does not offset the 9965’s broader superiority.

FAQ

Q: Which CPU has a higher single-thread score?

A: The AMD Ryzen 9 PRO 9965 scores 4682 in passmark_single_thread, while the 9965X3D scores 4614. The 9965 leads by 1.5%.

Q: How much faster is the 9965X3D in the prime numbers test?

A: The 9965X3D scores 599 versus the 9965’s 364, giving the X3D a 39.2% advantage in passmark_find_prime_numbers.

Q: What is the L3 cache difference between the two?

A: The 9965 has 64 MB of L3 cache, while the 9965X3D has 128 MB. Both have 80 KB of L1 and 1 MB of L2 per core.

Q: Do both CPUs have the same clock speeds?

A: Yes, both have a base clock of 4.30 GHz and a boost clock of 5.50 GHz, with a TDP of 170 W.

Q: Which CPU wins the multithread benchmark?

A: The 9965X3D wins passmark_multithread with 68225 points versus 66655 for the 9965, a 2.3% margin.

Q: How many benchmark tests does each CPU win?

A: The 9965 wins 8 head-to-head tests, while the 9965X3D wins 3 tests.

Specification Differences

The two processors share nearly every specification. Both are built on TSMC’s 4 nm process, use the Granite Ridge codename, and belong to the Ryzen 9 generation with Zen 5 architecture. They have identical transistor counts (16,630 million), die sizes (2x 70.6 mm²), and memory support (DDR5 dual-channel with 89.6 GB/s bandwidth). Both support ECC memory and feature Radeon Graphics integrated, with Gen 5 PCIe (24 lanes CPU only). The socket is AM5 for both, and both have a production status of Active with a release date of 2026-06-29.

The only specification difference is the L3 cache size: the 9965 has 64 MB, while the 9965X3D has 128 MB. This doubles the total cache capacity for the X3D part. The part numbers differ as well: 100-000002001 for the 9965 and 100-000001999 for the 9965X3D. No launch MSRP is listed for either. Both are locked multipliers, and both target the Server/Workstation market segment.

Architecture Differences

Architecturally, the two chips are identical in core design, node, and foundry. Both use 16 cores and 32 threads with a 4 nm process from TSMC, comprising 16,630 million transistors across two 70.6 mm² dies. The L1 cache is 80 KB per core and L2 is 1 MB per core for both. The sole architectural divergence is the L3 cache: the 9965X3D features 128 MB, which is double the 9965’s 64 MB. This additional cache is the defining feature of the X3D part, enabling its superior performance in cache-sensitive workloads like prime-number generation and physics simulations.

Neither part has a listed vCache3d field, but the X3D naming and the doubled L3 capacity strongly indicate an additional cache die. The base architecture, including the memory bus (dual-channel DDR5) and PCIe lanes (Gen 5, 24 lanes), remains unchanged. The integrated Radeon Graphics and ECC support are also identical. The result is that the X3D’s performance advantage comes entirely from cache capacity, not from any core or clock differences.

Where Each One Wins

The AMD Ryzen 9 PRO 9965 wins in all general-purpose compute tasks. It leads in integer math (243280 vs 237955), floating-point math (160746 vs 157123), data encryption (44920 vs 43905), data compression (908293 vs 895563), and random string sorting (94915 vs 92886). It also wins single-thread performance (4682 vs 4614) and extended instructions (71210 vs 70797). This makes it the better choice for software development, database workloads, encryption-heavy applications, and any mixed office or server environment where workloads vary unpredictably. The 9965’s 1.5% single-thread lead also makes it preferable for lightly threaded applications.

The AMD Ryzen 9 PRO 9965X3D wins in three specific areas: prime-number finding (599 vs 364, +39.2%), physics calculations (4801 vs 3256, +32.2%), and multithreaded throughput (68225 vs 66655, +2.3%). The physics and prime-number results are so dominant that any workload involving repeated access to a bounded dataset will benefit enormously. This includes scientific simulation, computational geometry, and certain types of AI inference where model weights fit within the 128 MB cache. The multithread win suggests that even in broad parallel tasks, the larger cache reduces memory bottlenecks. For users running these specific applications, the X3D is the clear choice despite losing most head-to-head tests.