AMD Ryzen 9 PRO 9945

The AMD Ryzen 9 PRO 9945 is a 12-core, 24-thread processor built on the Zen 5 (Granite Ridge) architecture, targeting the server and workstation segment with a 65 W TDP and AMD Socket AM5 compatibility. It achieves an average benchmark score of 96,083, placing it in the 96th percentile among all CPUs tested, with a boost clock of 5.40 GHz and a base clock of 3.40 GHz. This analysis examines its performance, power characteristics, platform support, and competitive positioning using only the supplied benchmark data.

Benchmark Performance

The Ryzen 9 PRO 9945 delivers a multi-threaded PassMark score of 48,664, which is the primary indicator of its overall throughput in heavily parallel workloads. This score, combined with a single-thread score of 4,619, produces an average benchmark score of 96,083 across all tested workloads. The processor’s standing relative to the broader market is strong, as its 96th percentile ranking indicates it outperforms the vast majority of CPUs in the benchmark database.

In specific compute-intensive tests, the processor shows distinct strengths. The floating-point math score of 107,340 and integer math score of 171,175 highlight robust arithmetic capabilities, which are essential for scientific computing, financial modeling, and engineering simulations. The extended instructions score of 44,374 suggests efficient handling of AVX-512 or similar vectorized workloads, while the data encryption score of 30,450 indicates solid cryptographic throughput for secure data processing tasks. The data compression score of 579,972 is particularly notable, reflecting strong performance in archiving, database compression, and file system operations.

When comparing its average score to the nearest rivals, the Ryzen 9 PRO 9945 sits in a tightly contested cluster. It leads the AMD EPYC 4565P by 0.3% (96,083 vs. 95,764), the AMD EPYC 9175F by 0.5% (96,083 vs. 95,615), and the AMD EPYC 4564P by 0.9% (96,083 vs. 95,183). However, it trails the AMD Ryzen 9 9955HX3D by 1.4% (96,083 vs. 97,453). These deltas are small, ranging from under one percent to just over one percent, indicating that the PRO 9945 is essentially performance-equivalent to its closest competitors in aggregate benchmarking. The differences in average score are minor enough that real-world application performance would likely be indistinguishable without controlled workload-specific testing.

In single-threaded performance, the 4,619 score is a key metric for responsiveness in lightly threaded applications. While the data does not provide direct rival single-thread comparisons, the overall percentile and average score suggest that this processor is competitive in the upper echelon of desktop and workstation CPUs. The single-thread score represents roughly 9.5% of the multi-thread score, which is a typical ratio for a 12-core processor with balanced per-core efficiency.

Power and Thermals

The Ryzen 9 PRO 9945 carries a TDP of 65 W, which classifies it as a power-efficient processor within the high-core-count segment. This TDP figure is notably low for a 12-core, 24-thread chip, especially when compared to typical workstation processors that often require significantly more power to sustain similar core counts. The 65 W TDP implies that the processor can be adequately cooled by a capable air cooler or a modest liquid cooling solution, without the need for extreme cooling infrastructure. For system integrators and workstation builders, this thermal profile simplifies chassis design and reduces overall system noise, as the cooling solution does not need to handle excessive heat dissipation.

The low TDP also has implications for sustained performance. Benchmarks such as the multi-thread score of 48,664 and the floating-point math score of 107,340 are achieved within this power envelope, suggesting that the processor does not throttle aggressively under load. The absence of an unlocked multiplier (multiplierUnlocked: false) means that users cannot overclock to push beyond the rated 5.40 GHz boost clock, but the stock performance is already well-optimized for the 65 W thermal budget. For dense server deployments or multi-socket workstations, the 65 W TDP allows for higher core density per chassis and lower cooling costs, which are critical factors in data center environments.

The production status is listed as Active, and the release date of 2025-09-15 indicates this is a current-generation part. The 4 nm process node from TSMC contributes to the favorable power-to-performance ratio, as smaller geometries typically reduce leakage current and improve switching efficiency. The die size of 2x 70.6 mm² (a dual-chiplet design) and total transistor count of 16,630 million are consistent with a modern high-core-count processor, but the 65 W TDP is the defining characteristic for thermal management.

Platform and Compatibility

The Ryzen 9 PRO 9945 uses the AMD Socket AM5 platform, which is the current mainstream socket for AMD desktop and workstation processors. This socket provides a clear upgrade path for users who may want to move to higher-core-count or higher-clock processors in the future, as AM5 is designed to support multiple generations of AMD CPUs. The processor supports DDR5 memory with a dual-channel bus, offering a memory bandwidth of 89.6 GB/s. This bandwidth is sufficient for feeding the 12 cores in memory-intensive applications, and the support for ECC memory is a critical feature for server and workstation reliability, where data integrity is paramount.

PCIe connectivity is provided via Gen 5 with 24 lanes (CPU only). This high-speed interface supports the latest GPUs, NVMe storage devices, and network adapters without bottlenecking data transfer rates. The 24 lanes are allocated to CPU-attached devices, which is typical for workstation-class processors and allows for multiple high-bandwidth peripherals to operate concurrently. The integrated graphics, listed as Radeon Graphics, provide basic display output capabilities, eliminating the need for a discrete GPU in headless servers or simple workstation configurations where graphics performance is not a priority.

Memory support is limited to DDR5, which means users must adopt current-generation memory modules. The dual-channel configuration, as opposed to quad-channel, differentiates this processor from some server-specific chips like the EPYC series, which often feature higher memory channel counts. However, for a 12-core workstation processor, dual-channel DDR5 with 89.6 GB/s is adequate for most workloads, and the ECC support adds a layer of fault tolerance that is essential for long-running computations or database servers. The market segment is defined as Server/Workstation, and the part number 100-000001407 confirms this is a distinct PRO-series SKU, likely with enhanced reliability and manageability features.

How It Compares

vs. AMD EPYC 4565P: The Ryzen 9 PRO 9945 holds a 0.3% average score advantage (96,083 vs. 95,764). This is a negligible difference, effectively placing the two processors on par in overall benchmark performance. The EPYC 4565P is a server-focused chip, while the PRO 9945 targets workstations, but the data shows no significant performance gap in aggregated workloads. Buyers choosing between these would base decisions on platform features (e.g., memory channels, PCIe lanes) rather than raw throughput.

vs. AMD EPYC 9175F: The PRO 9945 leads by 0.5% (96,083 vs. 95,615). The EPYC 9175F is likely clocked for high-frequency operation, but the PRO 9945’s 5.40 GHz boost clock and efficient Zen 5 architecture keep it competitive. The small delta suggests that in single-threaded or lightly threaded tasks, the PRO 9945 might edge ahead, while in heavily multi-threaded server workloads, the EPYC’s platform advantages (e.g., more memory bandwidth) could reverse the trend, but the benchmark data does not show this.

vs. AMD EPYC 4564P: The PRO 9945 outperforms this rival by 0.9% (96,083 vs. 95,183). This is the largest lead among the three EPYC rivals, but it remains under one percent. The 4564P is likely a lower-clocked variant, and the PRO 9945’s boost clock of 5.40 GHz contributes to its slight edge. In practical terms, this difference is within run-to-run variance for most benchmarks, so users should not expect a noticeable performance difference in typical applications.

vs. AMD Ryzen 9 9955HX3D: The PRO 9945 lags this rival by 1.4% (96,083 vs. 97,453). The 9955HX3D is a mobile-oriented processor with 3D V-Cache, which likely boosts performance in cache-sensitive workloads. The 1.4% deficit is modest, and the PRO 9945’s 65 W TDP is likely lower than the HX3D’s, making it more power-efficient. For users prioritizing power consumption over marginal performance, the PRO 9945 remains a strong choice, but the HX3D’s higher average score indicates it has a slight edge in aggregate throughput.

Single-Thread vs Multi-Thread Behavior

The Ryzen 9 PRO 9945 exhibits a clear split between single-thread and multi-thread performance, with a single-thread score of 4,619 and a multi-thread score of 48,664. The multi-thread score is approximately 10.5 times the single-thread score, which is slightly lower than the theoretical scaling of 12 cores (which would be 12x if perfectly linear). This indicates that the processor achieves about 88% scaling efficiency from its 12 cores, accounting for overheads such as memory contention and thread synchronization. This is a strong result for a 65 W TDP part, as power constraints often reduce scaling efficiency in higher-core-count processors.

For real workloads, this split means that the PRO 9945 excels in multi-threaded applications like video rendering, 3D simulation, and batch data processing, where all 24 threads can be utilized. The integer math score of 171,175 and floating-point math score of 107,340 corroborate this, as both are heavily parallel operations. Conversely, the single-thread score of 4,619 is critical for applications like legacy database queries, single-threaded scripting, or older games that do not scale across cores. A score of 4,619 is respectable, but not class-leading, meaning that users with workloads that are predominantly single-threaded may find faster options among higher-clock processors, though the 5.40 GHz boost clock mitigates this concern.

The find prime numbers score of 342 is notably lower than other metrics, which is typical for this test as it is highly dependent on integer division and branch prediction. This does not necessarily indicate a weakness, as the test is not representative of most real-world workloads. Similarly, the physics score of 2,902 reflects a specific computation pattern that may not align with the processor’s strengths. The random string sorting score of 62,458 and data compression score of 579,972 are more indicative of general data processing tasks, and both are strong, suggesting that the PRO 9945 handles memory-intensive and I/O-bound workloads efficiently.

FAQ

Q: What is the average benchmark score of the AMD Ryzen 9 PRO 9945?

A: The average benchmark score is 96,083, which places it in the 96th percentile among all CPUs tested.

Q: How does the Ryzen 9 PRO 9945 compare to the AMD EPYC 4565P?

A: The PRO 9945 has an average score of 96,083, which is 0.3% higher than the EPYC 4565P’s 95,764, making them effectively performance-equivalent.

Q: Does the Ryzen 9 PRO 9945 support ECC memory?

A: Yes, ECC memory support is listed as true, which is essential for server and workstation reliability.

Q: What is the TDP of the Ryzen 9 PRO 9945?

A: The TDP is 65 W, which is low for a 12-core processor and implies modest cooling requirements.

Q: What socket does the Ryzen 9 PRO 9945 use?

A: It uses AMD Socket AM5, which supports DDR5 memory and PCIe Gen 5 with 24 CPU-attached lanes.

Q: How does it perform in single-threaded tasks?

A: The single-thread score is 4,619, which is strong but not top-tier, indicating good responsiveness in lightly threaded applications.

Q: What is the boost clock of the Ryzen 9 PRO 9945?

A: The boost clock is 5.40 GHz, with a base clock of 3.40 GHz.

Who Should Consider It

The Ryzen 9 PRO 9945 is best suited for users whose workloads are predominantly multi-threaded and who prioritize power efficiency. The multi-thread score of 48,664 makes it an excellent choice for content creators working on video editing, 3D rendering, or software compilation, where the 12 cores and 24 threads can be fully utilized. The floating-point math score of 107,340 and integer math score of 171,175 further support its capability in scientific computing and engineering analysis, where heavy arithmetic operations are common. The 65 W TDP is a significant advantage for these users, as it allows for compact workstation builds with quiet cooling solutions, or for server deployments where power density is a concern.

For gamers, the single-thread score of 4,619 is adequate for modern titles, but the processor is not optimized for pure gaming performance. The integrated Radeon Graphics provide basic display output, but discrete GPUs would be needed for high-end gaming. The 96th percentile ranking indicates that it will not bottleneck a powerful GPU in most scenarios, but gaming-focused users might prefer processors with higher single-thread scores. Office and general productivity workloads, such as spreadsheet analysis, document processing, and web browsing, will be handled effortlessly, but the processor’s capabilities are overkill for these tasks, and a lower-core-count chip would suffice.

The Server/Workstation market segment and ECC memory support make it particularly attractive for small-scale server deployments, database servers, or network-attached storage systems where data integrity is critical. The PCIe Gen 5 with 24 lanes allows for multiple high-speed NVMe drives or network cards, and the 89.6 GB/s memory bandwidth supports concurrent data access from multiple applications. The 65 W TDP enables higher density in rack-mounted systems, and the AM5 socket provides a clear upgrade path to future AMD processors without changing the motherboard.

Architecture and Design

The Ryzen 9 PRO 9945 is built on the Granite Ridge codename, which is part of the Zen 5 architecture, the latest generation from AMD for the 9000 series. The manufacturing process is 4 nm, produced by TSMC, which is a leading-edge node that balances transistor density with power efficiency. The processor contains 16,630 million transistors across a die size of 2x 70.6 mm², indicating a dual-chiplet design where two CCDs (core complex dies) are used to house the 12 cores. This chiplet architecture allows AMD to scale core counts efficiently and improve yields, as smaller dies are easier to manufacture.

The cache hierarchy is structured with 80 KB of L1 cache per core, 1 MB of L2 cache per core, and a total of 64 MB of L3 cache. The L1 cache is split between instruction and data, though the pack does not specify the breakdown. The L2 cache is large at 1 MB per core, which helps reduce memory latency for frequently accessed data. The 64 MB L3 cache is shared across the 12 cores, providing a substantial pool of fast memory for workloads that exhibit data reuse, such as database queries or scientific simulations. There is no 3D V-Cache option listed, so this processor relies on the standard L3 cache configuration.

The processor supports DDR5 memory with a dual-channel bus, achieving a memory bandwidth of 89.6 GB/s. This is a critical design decision, as dual-channel memory is typical for mainstream processors, while server processors often use eight channels. The 89.6 GB/s bandwidth is sufficient for the 12 cores, but users with memory-bandwidth-intensive workloads (e.g., large in-memory databases) may need to consider quad-channel platforms. The ECC memory support is integrated into the memory controller, ensuring error correction for critical applications. The PCIe Gen 5 interface with 24 lanes is split between the CPU and chipset, with the 24 CPU-attached lanes providing direct high-speed connectivity to GPUs and NVMe storage. The integrated Radeon Graphics is a basic GPU solution, suitable for display output but not for compute tasks, and it does not affect the processor’s core performance characteristics. The multiplier is locked, meaning users cannot overclock, which is consistent with a PRO-series processor aimed at stability and reliability in professional environments. The production status is Active, and the release date of September 15, 2025, marks it as a current-generation product.