AMD EPYC 9115 vs AMD EPYC 9184X Comparison

The AMD EPYC 9115 and AMD EPYC 9184X are both 16-core server processors built for the SP5 platform, yet they represent distinct architectural approaches from consecutive generations. The 9115, a Zen 5 "Turin" part, and the 9184X, a Zen 4 "Genoa-X" part, deliver nearly identical average benchmark scores—69,288 for the 9115 versus 68,202 for the 9184X—but achieve this parity through very different strengths. The 9115 wins 12 of 17 head-to-head tests, while the 9184X secures 5 decisive victories, primarily in cache-sensitive and specialized workloads. Both sit at the 94th percentile among all CPUs, indicating top-tier performance overall, but the data reveals clear specialization patterns that dictate which processor suits a given workload.

Head-to-Head Benchmarks

The most striking pattern is the EPYC 9115’s dominance across all Cinebench render tests. In Cinebench R23 multi-core, the 9115 scores 42,003 against the 9184X’s 40,515, a 3.7% advantage. This same 3.7% delta repeats in R20 multi-core (17,641 vs. 17,016) and R15 multi-core (4,233 vs. 4,083). The single-core results tell an identical story: the 9115 leads by 3.6% in R15 (597 vs. 576) and by 3.7% in both R20 (2,490 vs. 2,401) and R23 (5,929 vs. 5,719). This consistent margin across render workloads suggests the 9115’s newer Zen 5 architecture provides a uniform IPC uplift over the Zen 4-based 9184X.

The 9115’s lead extends into several PassMark integer and floating-point tests. Floating point math shows a massive 19.2% advantage for the 9115 (113,853 vs. 95,476), while integer math trails at a 15.4% margin (181,807 vs. 157,483). Single-thread performance in PassMark reinforces this, with the 9115 scoring 3,360 against 2,822—a 19.1% gap. Extended instructions also favor the 9115, but by a smaller 4.4% (45,477 vs. 43,562). The multi-thread test rounds out the 9115’s wins with a 2.7% edge (48,936 vs. 47,665).

The 9184X fights back with commanding wins in specific tasks. Its most dramatic victory is in find prime numbers, where it scores 465 versus the 9115’s 289—a 37.8% advantage. Physics tests show a nearly identical 37.2% lead for the 9184X (6,674 vs. 4,188). Data encryption favors the 9184X by 10.4% (37,376 vs. 33,489), and random string sorting goes to the 9184X by 12.2% (79,913 vs. 70,151). Data compression is the closest contest, with the 9184X winning by just 2.4% (614,873 vs. 600,099). These results indicate the 9184X’s massive 768 MB L3 cache provides a decisive advantage in workloads that thrash memory or rely on large working sets.

FAQ

Q: Which processor is faster in multi-core rendering?

A: The EPYC 9115 wins all multi-core Cinebench tests by 3.7%. In Cinebench R23, it scores 42,003 versus the 9184X’s 40,515.

Q: Does the 9184X ever beat the 9115?

A: Yes. The 9184X wins 5 of 17 head-to-head tests, with its biggest wins in prime number finding (37.8% ahead), physics (37.2% ahead), and data encryption (10.4% ahead).

Q: How do the single-thread scores compare?

A: The 9115 leads in every single-thread test. PassMark single-thread shows a 19.1% advantage (3,360 vs. 2,822), while Cinebench R23 single-core shows a 3.7% edge (5,929 vs. 5,719).

Q: What explains the 9184X’s strong performance in compression and sorting?

A: The 9184X’s 768 MB shared L3 cache allows it to hold larger datasets on-die. It wins data compression by 2.4% and random string sorting by 12.2%, both tasks that benefit from a large cache.

Q: Are these processors comparable in overall performance?

A: The average benchmark scores are close: 69,288 for the 9115 and 68,202 for the 9184X. Both rank at the 94th percentile among all CPUs, though the 9115 has a slight edge in total wins (12 vs. 5).

Q: Which processor has a higher base clock speed?

A: The 9184X operates at a 3.55 GHz base clock, while the 9115 runs at 2.60 GHz. The 9184X also has a slightly higher boost clock at 4.20 GHz versus 4.10 GHz.

Architecture Differences

The architectural chasm between these processors defines their benchmark behavior. The EPYC 9115 uses the Zen 5 architecture, codenamed Turin, built on TSMC’s 4 nm process. It features 16 cores and 32 threads. Its L1 cache is 80 KB per core, L2 is 1 MB per core, and L3 is 64 MB shared. The transistor count is 16,630 million, spread across a die size of 2x 70.6 mm². This is a monolithic-style design with fewer, larger compute dies.

The EPYC 9184X relies on the older Zen 4 architecture, codenamed Genoa-X, fabricated on TSMC’s 5 nm process. It also has 16 cores and 32 threads, but its cache hierarchy is radically different. L1 drops to 64 KB per core, L2 remains at 1 MB per core, and L3 expands to a massive 768 MB shared. This enormous L3 comes from 3D V-Cache technology, which stacks additional cache on the compute dies. The 9184X contains 90,160 million transistors across 8x 72 mm² dies.

These architectural choices lead to different power and thermal profiles. The 9115 has a TDP of 125 watts, while the 9184X consumes 320 watts. The 9184X’s higher power draw is a direct result of its larger cache and higher clocks. The memory bandwidth also shifts: the 9115 delivers 576.0 GB/s, while the 9184X offers 460.8 GB/s, reflecting the 9115’s newer memory controller. Both use DDR5 with twelve-channel memory buses and support PCIe Gen 5 with 128 lanes from the CPU.

Specification Differences

The specification sheet reveals a clear generational split. The 9115 launches with a base clock of 2.60 GHz and a boost clock of 4.10 GHz, while the 9184X starts at 3.55 GHz and boosts to 4.20 GHz. The 9184X’s higher clocks benefit latency-sensitive tasks, but the 9115 compensates with architectural efficiency. TDP differs substantially: 125 watts for the 9115 versus 320 watts for the 9184X, meaning the 9115 offers a significantly lower power envelope.

Memory bandwidth favors the 9115 at 576.0 GB/s versus 460.8 GB/s for the 9184X. The 9115 uses a 4 nm process node, while the 9184X uses 5 nm. Transistor counts differ by a factor of over five: 16,630 million for the 9115 versus 90,160 million for the 9184X, a disparity driven by the 9184X’s stacked cache dies. The 9115 has an L3 cache of 64 MB, while the 9184X has 768 MB. Release dates also differ: the 9115 launched on 2024-10-09, while the 9184X launched on 2023-06-12. The 9115 has a launch MSRP of $726, while the 9184X has a launch MSRP of $4928. The 9115 carries part number 100-000001552, while the 9184X has no listed part number.

Where Each One Wins

The EPYC 9115 is the clear winner for general-purpose compute, rendering, and single-threaded workloads. Its 3.7% lead across all Cinebench tests makes it the better choice for 3D rendering and content creation. The 19.2% advantage in floating-point math and 15.4% in integer math positions it well for scientific simulations and financial modeling. The 19.1% single-thread lead in PassMark indicates superior responsiveness for latency-sensitive applications. Its lower 125-watt TDP also makes it more suitable for dense deployments where power density matters. The 9115’s higher memory bandwidth of 576.0 GB/s gives it an edge in memory-intensive operations that fit within its 64 MB cache.

The EPYC 9184X excels in cache-bound and specialized workloads. Its 37.8% victory in prime number finding and 37.2% win in physics suggest a strong fit for computational mathematics and physics simulations that benefit from the 768 MB L3 cache. The 10.4% advantage in data encryption makes it suitable for security appliances and cryptographic workloads. The 12.2% lead in random string sorting points to advantages in database operations and sorting algorithms. The 2.4% win in data compression positions it well for archival and storage systems. However, these wins come at the cost of a 320-watt TDP, making the 9184X less efficient for general workloads. The 9184X’s higher base and boost clocks (3.55 GHz and 4.20 GHz) provide an advantage in lightly threaded tasks that are not cache-limited, but the 9115’s architectural improvements ultimately deliver better performance per watt in most benchmarks.