Winning at Every Core Count: Maximizing Performance, Minimizing Software Licensing Costs
Oct 31, 2025
In my recent blog, “Another Generation of Leadership: AMD EPYC 9005 vs. Intel Xeon 6,” I explored how the 5th Gen AMD EPYC™ 9005 Series Processors are setting a new benchmark for innovation and performance across a wide range of workloads—comparing top-of-stack SKUs head-to-head.
This follow-up takes a deeper look at performance comparisons between AMD EPYC 9005 (“Turin”) and Intel® Xeon® 6 (“Granite Rapids SP”), focusing on 32-core vs. 32-core configurations — the most popular setups in the heart of the enterprise market. While top-tier SKU performance and efficiency are critical for cloud service providers aiming to maximize VM density, core-to-core comparisons are especially relevant for enterprise environments, where software stacks are often licensed per core.
This blog highlights workloads that represent a broad spectrum of enterprise applications, offering insights into how these processors perform under real-world conditions. To round out the analysis, I’ve also included 64-core vs. 64-core comparisons at the end of the blog.
General Purpose Computing
The SPEC CPU® 2017 benchmark is a highly respected, regulated, and independently audited industry standard for evaluating compute-intensive workloads. It rigorously tests processors, memory subsystems, and compilers across a wide range of systems, offering valuable insights into performance and efficiency. As a critical tool for assessing modern computing architectures, SPEC CPU® 2017 includes 43 benchmarks organized into four suites. This blog focuses on two of the most foundational and frequently requested suites by customers: SPECrate® 2017 Integer and SPECrate® 2017 Floating Point. These suites provide essential points of comparison for understanding system performance in real-world enterprise environments.
Figure 1 illustrates that with GCC compiler: The 5th Gen AMD EPYC 9375F achieved an estimated 1.44x performance uplift over the 6th Gen Intel Xeon 6737P on SPECrate® 2017_int_base, and 1.65x (est.) on SPECrate® 2017_fp_base[1].
Figure 1: General purpose computing (GCC Compiler)
Server-Side Java
Java® is a platform-independent programming language that enables seamless application deployment across diverse environments. To evaluate server-side Java performance, we used a benchmark which simulates corporate IT workloads such as point-of-sale transactions, online purchases, and data mining. This benchmark provides two key metrics: m-jOPS, measuring “maximum sustained throughput” for Java e-commerce transactions, and c-jOPS, reflecting “throughput under p99 response time” constraints for latency-sensitive operations representing typical production Java enterprise deployments .
Figure 2 shows the 5th Gen AMD EPYC 9375F achieved a ~1.45x performance uplift over the 6th Gen Intel Xeon 6737P on m-jOPS, and ~1.28x on c-jOPS [2].
Figure 2: Server-side Java
Relational Databases
MySQL™ is one of the most widely adopted open-source database management systems, commonly used in enterprise and cloud-native environments for both decision support and transaction processing. To evaluate its performance, AMD leveraged TPROC-H—based on the popular TPC-H benchmark—for decision support workloads, and TPROC-C—based on the TPC-C benchmark®—for transaction processing systems. These benchmarks provide targeted insights into how MySQL handles diverse enterprise demands.
Figure 3 shows the 5th Gen AMD EPYC 9375F achieved a ~1.19x performance uplift over the 6th Gen Xeon 6737P on MySQL TPROC-H, and 1.59x on MySQL TPROC-C [3].
Figure 3: Relational Database
NoSQL Database
MongoDB is a document-based, open-source NoSQL database built to efficiently manage large volumes of unstructured or semi-structured data. It’s widely adopted in modern applications such as social media platforms, content management systems, mobile apps, and e-commerce sites—where flexibility, scalability, and performance are critical. For competitive analysis, we used the Yahoo! Cloud Serving Benchmark (YCSB), a popular open-source framework that evaluates and compares the performance of cloud databases and storage systems under various workload conditions.
Figure 4 shows the 5th Gen AMD EPYC 9375F achieved a ~1.66x MongoDB NoSQL Database performance uplift over the 6th Gen Intel Xeon 6737P [5].
Figure 4: NoSQL Database
Memcached® High Speed Data Caching
Memcached™ is a high-performance, distributed, in-memory caching system designed to store key-value pairs for small pieces of arbitrary data—such as strings and objects—retrieved from database queries, APIs, or rendered pages. Widely used in cloud environments, Memcached excels at quickly serving cached items while offering scalability and cost-efficiency under heavy loads. It’s commonly employed to cache database query results, session data, web pages, APIs, and objects like images, files, and metadata. Built for simplicity and speed, Memcached supports rapid development and deployment to meet the demands of large-scale data caching. To evaluate its performance, AMD used the Memtier Benchmark, which measures throughput and latency across various configurations.
Figure 5 shows the 5th Gen AMD EPYC 9375F achieved a 1.39x Memcached High Speed Data Caching performance uplift over the 6th Gen Intel Xeon 6737P [6].
Figure 5: High Speed Data Caching
Web Server
Web server is foundational to modern infrastructure, and NGINX™ stands out for its versatility—serving as a reverse proxy, load balancer, mail proxy, and HTTP cache. Designed to efficiently handle client requests and deliver web content, NGINX can operate as a standalone server or enhance performance and security when deployed as a reverse proxy. To assess its performance, AMD used the widely adopted WRK tool, which generates substantial HTTP loads to benchmark throughput and latency under real-world conditions.
Figure 6 shows the 5th Gen AMD EPYC 9375F achieved a 1.14x NGINX™ WRK performance uplift over the 6th Gen Intel Xeon 6737P [7].
Figure 6: Web Server NGINX™ wrk
In-Memory Analytics
Redis™ is a widely used in-memory data store that functions as a key-value database, cache, and message broker, with optional durability features. It’s well-suited for cloud environments and supports use cases such as streaming, microservices, and data analytics—helping developers build responsive and scalable applications. Redis handles various data types including strings, lists, maps, and more, and its in-memory architecture delivers exceptional performance. To evaluate Redis, AMD used the redis-benchmark tool, which simulates multiple client connections and measures request completion times, providing insights into the average number of requests the server can handle per second.
Figure 7 shows the 5th Gen AMD EPYC 9375F achieved a ~1.77x performance uplift over the 6th Gen Intel Xeon 6737P on Redis-Get, and ~1.66x on Redis-Set [4].
Figure 7: In-Memory Analytics Redis™ redis-Benchmark
Media Processing
Media processing performance is critical for a wide range of workloads—from video encoding and transcoding to collaboration platforms and streaming services. FFmpeg is a powerful, open-source multimedia framework that includes a rich set of libraries, codecs, and tools for handling video, audio, and other media formats. Known for its command-line efficiency, FFmpeg is widely used for tasks such as encoding, transcoding, editing, scaling, post-production, and ensuring standards compliance.
Figure 8 shows the 5th Gen AMD EPYC 9375F achieved FFmpeg performance uplifts of up to ~1.47x over the 6th Gen Intel Xeon 6737P [8].
Figure 8: FFmpeg Encoding and Transcoding
Summary
In the section above, I covered 32-core vs. 32-core comparisons. AMD EPYC™ continues to demonstrate a similar performance advantage over Intel “Granite Rapids” even at the 64-core level. See the chart below for a detailed comparison [9]-[16].
Figure 9: AMD EPYC 9575F 64-core performance gain over Intel Xeon 6767P 64-core
Wrapping Up
The lower core count middle stack of 5th Generation AMD EPYC processors marks a major advancement in server technology, building on AMD leadership in x86 architecture with higher core counts, advanced multi-threading, superior performance, and exceptional energy efficiency. Designed to meet the rapidly growing demands for performance, capacity, and efficiency in modern data centers, the AMD EPYC architecture provides outstanding value across a wide range of applications, including critical enterprise and cloud-native workloads. 5th gen EPYC processors feature a state-of-the-art design with increased core counts, improved instructions per cycle (IPC) generationally, and enhanced memory and I/O bandwidth, all while integrating cutting-edge confidential computing technologies. This innovative processor sets a new benchmark in server performance, made possible through the strong collaboration with our ecosystem partners.
Raghu Nambiar is a Corporate Vice President of Data Center Ecosystems and Solutions for AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites, and no endorsement is implied.
Footnotes
References
1. 9xx5-231: SPECrate®2017_int_base GCC 13 compiler comparison based on internal testing as of 08/15/2025. 2P 32C AMD EPYC 9375F powered reference system, 775 estimated SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power; 2P 32C Intel Xeon 6737P powered production system, 540 SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on;
9xx5-232: SPECrate®2017_fp_base GCC 13 compiler comparison based on internal testing as of 08/15/2025. 2P 32C AMD EPYC 9375F powered reference system, 898 estimated SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT OFF , NPS1, Determinism=Power; 2P 32C Intel Xeon 6737P powered production system, 545 SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on; 775/540=1.44x; 898/545=1.65x
2. "9xx5-234: Server side multi-instance max operations (m-jOPS) and critical operations (c-jOPS) comparisons based on AMD internal testing as of 08/15/2025
2P 32C AMD EPYC 9375F powered reference system, 460753 m-jOPS, 340290 c-jOPS, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 316781 m-jOPS, 266176c-jOPS, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
3. "9XX5-237: MySQL TPROC-H workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-H workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-H™ results, as the results do not comply with the TPC-H Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, Scale Factor 30, aggregate New Queries per Hour (QphH)
2P 32C AMD EPYC 9375F powered reference system, 49822 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 41756 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
"9XX5-238: MySQL TPROC-C workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-C workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-C™ results, as the results do not comply with the TPC-C Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, duration 5min, 32 v users, warehouses 128, aggregate New Orders Per Minute (NOPM)
2P 32C AMD EPYC 9375F powered reference system, 10388986 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 6540736 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. TPC, TPC Benchmark, and TPC-C are trademarks of the Transaction Processing Performance Council.
4. 9xx5-240: Redis™ GET and SET workloads based on AMD internal measurements as of 08/15/2025.
Workload configs: Redis 7.4.0, 4 core nodes (Multi-SUT), data size 1000, tests GET SET, pipelinerequests=10
2P 32C AMD EPYC 9375F powered reference system, GET 36931448 rps, SET 35028533 rps, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, GET 20922983 rps, SET 21071463 RPS, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 36931448/20922983=1.77x; 35028533/21071463=1.66x
5. 9xx5-242: MongoDB workload based on AMD internal measurements as of 08/15/2025.
2P 32C AMD EPYC 9375F powered reference system, 936185 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 565212 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 936185 /565212 =1.66x
6. 9xx5-244: Memcached workload based on AMD internal measurements as of 08/15/2025.
2P 32C AMD EPYC 9375F powered reference system, 43945796 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 31703721 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 43945796/31703721=1.39x
7. 9xx5-246: NGINX™ WRK workload based on AMD internal measurements as of 08/15/2025.
2P 32C AMD EPYC 9375F powered reference system, 8754271 req/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 7709525 req/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 8754271/7709525=1.14x
8. 9xx5-248: FFMpeg workload based on AMD internal measurements as of 08/15/2025.
FFMPEG 6.1.1, 1 instance/thread, video source: ducks take-off, 302 frames (Raw to VP9, Raw to h264, vp9 to h264, h264 to vp9)
2P 32C AMD EPYC 9375F powered reference system, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Workload 6737P 9375F Relative
raw_vp9 2165650.96 3110672.00 1.436
raw_h264 2891767.40 4155961.81 1.437
vp9_h264 2810406.80 4141432.03 1.474
h264_vp9 2551334.76 3687912.21 1.445
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings.
9. 9xx5-229: SPECrate®2017_int_base GCC 13 compiler comparison based on internal testing as of 08/15/2025. 2P 64C AMD EPYC 9575F powered reference system, 1190 estimated SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power; 2P 64C Intel Xeon 6767P powered production system, 914 SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on; 1190/914=1.30x;
"9xx5-230: SPECrate®2017_fp_base GCC 13 compiler comparison based on internal testing as of 08/15/2025;
2P 64C AMD EPYC 9575F powered reference system, 1020 estimated SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT off, NPS1, Determinism=Power; 2P 64C Intel Xeon 6767P powered production system, 825 SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=off, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on; 1020/825=1.24x
10. 9xx5-257: SPECjbb® 2015-MultiJVM max-jOPS and critical-jOPS comparisons based on published scores from www.spec.org as of 10/20/2025 2P 64C AMD EPYC 9575F, 621179 SPECjbb®2015-MultiJVM max-jOPS, 568932 SPECjbb®2015-MultiJVM critical-jOPS, https://spec.org/jbb2015/results/res2025q4/jbb2015-20250916-01648.html 2P 64C AMD EPYC 9575F, 753796 SPECjbb®2015-MultiJVM max-jOPS, 365455 SPECjbb®2015-MultiJVM critical-jOPS, https://spec.org/jbb2015/results/res2025q4/jbb2015-20250916-01647.html 2P 64C Intel Xeon 6767P 592688 SPECjbb®2015-MultiJVM max-jOPS, 386396 SPECjbb®2015-MultiJVM critical-jOPS, http://www.spec.org/jbb2015/results/res2025q2/jbb2015-20250402-01571.html 2P 64C Intel Xeon 6767P 500063 SPECjbb®2015-MultiJVM max-jOPS, 457394 SPECjbb®2015-MultiJVM critical-jOPS, http://www.spec.org/jbb2015/results/res2025q2/jbb2015-20250402-01570.html For 1.27x the highest Max-jOPS performance For 1.24x the highest Critical-jOPS performance SPEC®, SPEC CPU®, and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information.ion Corporation. See www.spec.org for more information. 753796/592688=1.26x; 621179/500063=1.24x
11. 9XX5-235: MySQL TPROC-H workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-H workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-H™ results, as the results do not comply with the TPC-H Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, Scale Factor 30, aggregate New Queries per Hour (QphH)
2P 64C AMD EPYC 9575F powered reference system, 108384 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 76955 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
"9XX5-236: MySQL TPROC-C workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-C workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-C™ results, as the results do not comply with the TPC-C Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, duration 5min, 32 v users, warehouses 128, aggregate New Orders Per Minute (NOPM)
2P 64C AMD EPYC 9575F powered reference system, 19317869 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 12299768 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. TPC, TPC Benchmark, and TPC-H are trademarks of the Transaction Processing Performance Council.
108384/76955= 1.41x 19317869/12299768= 1.57x
12. 9xx5-239: Redis™ GET and SET workloads based on AMD internal measurements as of 08/15/2025.
Workload configs: Redis 7.4.0, 4 core nodes (Multi-SUT), data size 1000, tests GET SET, pipelinerequests=10
2P 64C AMD EPYC 9575F powered reference system, GET 78802838 SET 72507450 RPS, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, GET 37476165 rps, SET 39696937 rps, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 78802838/37476165=2.1x; 72507450/39696937=1.83x
13. 9xx5-241: MongoDB workload based on AMD internal measurements as of 08/15/2025.
2P 64C AMD EPYC 9575F powered reference system, 2240882 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 1159512 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 2240882 / 1159512 = 1.93x
14. 9xx5-243: Memcached workload based on AMD internal measurements as of 08/15/2025.
2P 64C AMD EPYC 9575F powered reference system, 58692041 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 32672882 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 58692041/32672882=1.8x
15. 9xx5-245: NGINX™ WRK workload estimates based on AMD internal measurements as of 08/15/2025.
2P 64C AMD EPYC 9575F powered reference system, 23882083 req/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 13190279 req/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 23882083/13190279=1.81x
16. 9xx5-247: FFMpeg workload based on AMD internal measurements as of 08/15/2025.
FFMPEG 6.1.1, 1 instance/thread, video source: ducks take-off, 302 frames (Raw to VP9, Raw to h264, vp9 to h264, h264 to vp9)
2P 64C AMD EPYC 9575F powered reference system, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Workload 6767P 9575F Relative
raw_vp9 3882916.51 5482051.86 1.412
raw_h264 4894392.76 7502108.76 1.533
vp9_h264 4864390.48 7448709.95 1.531
h264_vp9 4672757.08 6433209.13 1.377
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings.
Footnotes
References
1. 9xx5-231: SPECrate®2017_int_base GCC 13 compiler comparison based on internal testing as of 08/15/2025. 2P 32C AMD EPYC 9375F powered reference system, 775 estimated SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power; 2P 32C Intel Xeon 6737P powered production system, 540 SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on;
9xx5-232: SPECrate®2017_fp_base GCC 13 compiler comparison based on internal testing as of 08/15/2025. 2P 32C AMD EPYC 9375F powered reference system, 898 estimated SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT OFF , NPS1, Determinism=Power; 2P 32C Intel Xeon 6737P powered production system, 545 SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on; 775/540=1.44x; 898/545=1.65x
2. "9xx5-234: Server side multi-instance max operations (m-jOPS) and critical operations (c-jOPS) comparisons based on AMD internal testing as of 08/15/2025
2P 32C AMD EPYC 9375F powered reference system, 460753 m-jOPS, 340290 c-jOPS, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 316781 m-jOPS, 266176c-jOPS, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
3. "9XX5-237: MySQL TPROC-H workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-H workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-H™ results, as the results do not comply with the TPC-H Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, Scale Factor 30, aggregate New Queries per Hour (QphH)
2P 32C AMD EPYC 9375F powered reference system, 49822 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 41756 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
"9XX5-238: MySQL TPROC-C workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-C workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-C™ results, as the results do not comply with the TPC-C Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, duration 5min, 32 v users, warehouses 128, aggregate New Orders Per Minute (NOPM)
2P 32C AMD EPYC 9375F powered reference system, 10388986 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 6540736 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. TPC, TPC Benchmark, and TPC-C are trademarks of the Transaction Processing Performance Council.
4. 9xx5-240: Redis™ GET and SET workloads based on AMD internal measurements as of 08/15/2025.
Workload configs: Redis 7.4.0, 4 core nodes (Multi-SUT), data size 1000, tests GET SET, pipelinerequests=10
2P 32C AMD EPYC 9375F powered reference system, GET 36931448 rps, SET 35028533 rps, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, GET 20922983 rps, SET 21071463 RPS, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 36931448/20922983=1.77x; 35028533/21071463=1.66x
5. 9xx5-242: MongoDB workload based on AMD internal measurements as of 08/15/2025.
2P 32C AMD EPYC 9375F powered reference system, 936185 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 565212 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 936185 /565212 =1.66x
6. 9xx5-244: Memcached workload based on AMD internal measurements as of 08/15/2025.
2P 32C AMD EPYC 9375F powered reference system, 43945796 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 31703721 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 43945796/31703721=1.39x
7. 9xx5-246: NGINX™ WRK workload based on AMD internal measurements as of 08/15/2025.
2P 32C AMD EPYC 9375F powered reference system, 8754271 req/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 7709525 req/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 8754271/7709525=1.14x
8. 9xx5-248: FFMpeg workload based on AMD internal measurements as of 08/15/2025.
FFMPEG 6.1.1, 1 instance/thread, video source: ducks take-off, 302 frames (Raw to VP9, Raw to h264, vp9 to h264, h264 to vp9)
2P 32C AMD EPYC 9375F powered reference system, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 32C Intel Xeon 6737P powered production system, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Workload 6737P 9375F Relative
raw_vp9 2165650.96 3110672.00 1.436
raw_h264 2891767.40 4155961.81 1.437
vp9_h264 2810406.80 4141432.03 1.474
h264_vp9 2551334.76 3687912.21 1.445
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings.
9. 9xx5-229: SPECrate®2017_int_base GCC 13 compiler comparison based on internal testing as of 08/15/2025. 2P 64C AMD EPYC 9575F powered reference system, 1190 estimated SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power; 2P 64C Intel Xeon 6767P powered production system, 914 SPECrate2017®_int_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on; 1190/914=1.30x;
"9xx5-230: SPECrate®2017_fp_base GCC 13 compiler comparison based on internal testing as of 08/15/2025;
2P 64C AMD EPYC 9575F powered reference system, 1020 estimated SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT off, NPS1, Determinism=Power; 2P 64C Intel Xeon 6767P powered production system, 825 SPECrate2017®_fp_base, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=off, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on; 1020/825=1.24x
10. 9xx5-257: SPECjbb® 2015-MultiJVM max-jOPS and critical-jOPS comparisons based on published scores from www.spec.org as of 10/20/2025 2P 64C AMD EPYC 9575F, 621179 SPECjbb®2015-MultiJVM max-jOPS, 568932 SPECjbb®2015-MultiJVM critical-jOPS, https://spec.org/jbb2015/results/res2025q4/jbb2015-20250916-01648.html 2P 64C AMD EPYC 9575F, 753796 SPECjbb®2015-MultiJVM max-jOPS, 365455 SPECjbb®2015-MultiJVM critical-jOPS, https://spec.org/jbb2015/results/res2025q4/jbb2015-20250916-01647.html 2P 64C Intel Xeon 6767P 592688 SPECjbb®2015-MultiJVM max-jOPS, 386396 SPECjbb®2015-MultiJVM critical-jOPS, http://www.spec.org/jbb2015/results/res2025q2/jbb2015-20250402-01571.html 2P 64C Intel Xeon 6767P 500063 SPECjbb®2015-MultiJVM max-jOPS, 457394 SPECjbb®2015-MultiJVM critical-jOPS, http://www.spec.org/jbb2015/results/res2025q2/jbb2015-20250402-01570.html For 1.27x the highest Max-jOPS performance For 1.24x the highest Critical-jOPS performance SPEC®, SPEC CPU®, and SPECjbb® are registered trademarks of the Standard Performance Evaluation Corporation. See www.spec.org for more information.ion Corporation. See www.spec.org for more information. 753796/592688=1.26x; 621179/500063=1.24x
11. 9XX5-235: MySQL TPROC-H workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-H workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-H™ results, as the results do not comply with the TPC-H Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, Scale Factor 30, aggregate New Queries per Hour (QphH)
2P 64C AMD EPYC 9575F powered reference system, 108384 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 76955 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM, Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
"9XX5-236: MySQL TPROC-C workload (SQL Server OLTP Brokerage) based on internal AMD measurements as of 08/15/2025. The HammerDB TPROC-C workload is an open-source workload derived from TPC-Benchmark™ Standard, and as such is not comparable to published TPC-C™ results, as the results do not comply with the TPC-C Benchmark Standard.
Workload configs: MySQL 8.0.41, 8 core nodes (Multi-SUT), HammerDB-4.4, duration 5min, 32 v users, warehouses 128, aggregate New Orders Per Minute (NOPM)
2P 64C AMD EPYC 9575F powered reference system, 19317869 QphH, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 12299768 QphH, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. TPC, TPC Benchmark, and TPC-H are trademarks of the Transaction Processing Performance Council.
108384/76955= 1.41x 19317869/12299768= 1.57x
12. 9xx5-239: Redis™ GET and SET workloads based on AMD internal measurements as of 08/15/2025.
Workload configs: Redis 7.4.0, 4 core nodes (Multi-SUT), data size 1000, tests GET SET, pipelinerequests=10
2P 64C AMD EPYC 9575F powered reference system, GET 78802838 SET 72507450 RPS, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, GET 37476165 rps, SET 39696937 rps, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 78802838/37476165=2.1x; 72507450/39696937=1.83x
13. 9xx5-241: MongoDB workload based on AMD internal measurements as of 08/15/2025.
2P 64C AMD EPYC 9575F powered reference system, 2240882 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 1159512 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 2240882 / 1159512 = 1.93x
14. 9xx5-243: Memcached workload based on AMD internal measurements as of 08/15/2025.
2P 64C AMD EPYC 9575F powered reference system, 58692041 Operations/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 32672882 Operations/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 58692041/32672882=1.8x
15. 9xx5-245: NGINX™ WRK workload estimates based on AMD internal measurements as of 08/15/2025.
2P 64C AMD EPYC 9575F powered reference system, 23882083 req/s, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 13190279 req/s, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings. 23882083/13190279=1.81x
16. 9xx5-247: FFMpeg workload based on AMD internal measurements as of 08/15/2025.
FFMPEG 6.1.1, 1 instance/thread, video source: ducks take-off, 302 frames (Raw to VP9, Raw to h264, vp9 to h264, h264 to vp9)
2P 64C AMD EPYC 9575F powered reference system, 1.5TB 24x64GB DDR5-6400 (running at 6000 MT/s), 2 x 40 GbE Mellanox CX-7, 3.84TB NVMe, BIOS RVOT1006C, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Mitigations enabled, SMT ON, NPS1, Determinism=Power
2P 64C Intel Xeon 6767P powered production system, 1.5TB 24x64GB DDR5-6400, 4 x 1GbE Broadcom NetXtreme BCM5719 Gigabit Ethernet PCIe, 3.84TB NVMe, BIOS 1.22, VMware ESXI 8.0.3 (build 24022510) with 16 threads and 48GB memory per VM Ubuntu 22.04.2 LTS | Linux 6.8.0-63-generic, Hyper-Threading=Enabled, Turbo boost=enabled, SNC3, Performance Bias, Mitigations=on
Workload 6767P 9575F Relative
raw_vp9 3882916.51 5482051.86 1.412
raw_h264 4894392.76 7502108.76 1.533
vp9_h264 4864390.48 7448709.95 1.531
h264_vp9 4672757.08 6433209.13 1.377
Results may vary based on factors including but not limited to system configurations, software versions, and BIOS settings.
Related Blogs
-
-
Open Source AI Week Recap
Learn more about AMD AI DevDay and how AMD participated at the PyTorch Conference during Open Source AI Week.
October 24, 2025
-
Building Trust Through Open Innovation
Read how AMD is making significant strides in the realm of open-source AI models and safety by embracing transparency, collaboration, and responsible AI development.
October 23, 2025
-
STX-B0T: Real-time AI Robot Assistant Powered by RyzenAI and ROCm — ROCm Blogs
STX-B0T explores the potential of RyzenAI PCs to power robotics applications on NPU+GPU. This blog demonstrates how our hardware and software interoperate to unlock real-time perception.
October 22, 2025
-
Empowering Developers to Build a Robust PyTorch Ecosystem on AMD ROCm™ with Better Insights and Monitoring — ROCm Blogs
Production ROCm support for N-1 to N+1 PyTorch releases is in progress. The AI Software Head-Up Dashboard shows status of PyTorch on ROCm.
October 20, 2025
-
Making Telcos Run Leaner and AI-Ready with AMD
Learn how telecommunication companies are running leaner and becoming AI-ready with AMD processors
October 20, 2025
-
-
ROCm 7.9 Technology Preview: ROCm Core SDK and TheRock Build System — ROCm Blogs
Introduce ROCm Core SDK, and learn to install and build ROCm components easily using TheRock.
October 19, 2025
