AMD EPYC 4484PX vs AMD EPYC 9115 Comparison
AMD EPYC 4484PX
EPYC 9115
PERFORMANCE BENCHMARKS
Analysis: AMD EPYC 4484PX vs AMD EPYC 9115
The AMD EPYC 9115 and AMD EPYC 4484PX represent two distinct approaches to server processing within AMD’s current lineup. The 9115, a 16-core Zen 5 part from the EPYC 9005 series, squares off against the 12-core Zen 4 EPYC 4484PX from the 4004 series. Benchmark data shows a clear split: the 4484PX wins 14 of 17 head-to-head comparisons, while the 9115 takes 3 decisive victories in specific computational workloads. The overall average benchmark scores are close, with the 9115 at 69,288 and the 4484PX at 67,822, placing both at the 94th percentile among all CPUs. However, the nature of their wins reveals fundamentally different performance profiles driven by architecture and configuration.
Head-to-Head Benchmarks
The most striking pattern in the data is the 4484PX’s dominance in single-threaded and lightly-threaded workloads. In PassMark single-thread testing, the 4484PX scores 4,119 against the 9115’s 3,360, a substantial 18.4% advantage. This gap is consistent across Cinebench single-core tests: the 4484PX leads by 2.3% in R15 (611 vs 597), 2.2% in R20 (2,547 vs 2,490), and 2.2% in R23 (6,065 vs 5,929). The 4484PX also wins PassMark physics by 15.2% (4,938 vs 4,188), a test that often responds to high clock speeds and strong per-core efficiency. These results align with the clock specifications: the 4484PX has a base clock of 4.40 GHz and boost of 5.60 GHz, compared to the 9115’s 2.60 GHz base and 4.10 GHz boost.
In multi-threaded Cinebench tests, the 4484PX maintains a consistent edge, though the margins are smaller than in single-thread. The 4484PX scores 4,330 in R15 multi-core versus 4,233 for the 9115, a 2.2% lead. In R20, the gap is 18,044 vs 17,641 (2.2%), and in R23 it is 42,964 vs 42,003 (2.2%). PassMark multithread shows a 3.2% win for the 4484PX (50,547 vs 48,936). This is notable because the 9115 has 16 cores and 32 threads versus the 4484PX’s 12 cores and 24 threads. The 4484PX’s higher clocks apparently offset the core deficit, at least in these tests. The 4484PX also wins PassMark data compression by 0.5% (603,371 vs 600,099) and random string sorting by 2.1% (71,642 vs 70,151).
The 9115’s three wins are large and specific. The biggest is PassMark floating point math, where it scores 113,853 against 96,446, an 18% advantage. Integer math shows an 11.5% lead (181,807 vs 162,993). Extended instructions go to the 9115 by 5% (45,477 vs 43,315). These are computationally intensive, throughput-oriented workloads where the 9115’s extra cores and Zen 5 architecture likely provide an advantage. The 4484PX wins data encryption by 8.2% (36,499 vs 33,489), but the 9115 counters with a 32% deficit in find prime numbers, where the 4484PX scores 425 versus 289. That 32% margin is the largest single delta in either direction across all tests.
The Verdict
The data supports a clear split recommendation. For users prioritizing single-thread performance, high clock speeds, and memory bandwidth at the platform level, the EPYC 4484PX is the stronger choice. It wins all Cinebench single-core tests, PassMark single-thread, physics, and multithread by margins ranging from 2.2% to 18.4%. Its 5.60 GHz boost clock and 4.40 GHz base clock are the highest in this comparison, and it delivers wins in 14 of 17 benchmarks despite having 4 fewer cores. The 4484PX’s average benchmark score of 67,822 ranks it just 0.1% behind the AMD Ryzen Threadripper PRO 5955WX and 0.2% ahead of the Intel Xeon w5-3525, per its nearestRivals data.
The EPYC 9115 is the pick for specific compute-heavy workloads. Its wins in floating point math (18% ahead), integer math (11.5% ahead), and extended instructions (5% ahead) indicate a strong capability for scientific, financial, or engineering simulations that leverage heavy math throughput. The 9115’s average score of 69,288 is 0.1% behind the Intel Core i7-14700K and 0.3% behind the AMD Ryzen 9 7950X, but its raw math scores are decisive. It also offers 64 MB of shared L3 cache versus 128 MB on the 4484PX, yet still loses on most memory-sensitive benchmarks, suggesting the 4484PX’s higher clocks matter more in those tests.
Neither chip is universally superior. The 4484PX is the default recommendation for general server workloads, database queries, and any application that benefits from high clock speeds. The 9115 is the specialist for math-heavy tasks where its extra 4 cores and Zen 5 efficiency translate into large percentage wins.
Where Each One Wins
The EPYC 4484PX wins everywhere except three math-specific PassMark tests. Its 14 wins cover the entire Cinebench suite (R15, R20, R23, both single and multi-core), PassMark multithread, single-thread, physics, data compression, data encryption, random string sorting, and find prime numbers. The find prime numbers result is particularly lopsided at 425 vs 289, a 32% advantage that suggests a major per-core throughput advantage in integer-heavy, low-level operations. The 4484PX also wins data encryption by 8.2%, a relevant metric for secure server workloads. For single-thread responsiveness, the 18.4% lead in PassMark single-thread is the strongest argument for this chip.
The EPYC 9115 wins exactly three tests, but they are high-value ones. Floating point math at 113,853 vs 96,446 (18% lead) is critical for scientific computing, 3D rendering, and financial modeling. Integer math at 181,807 vs 162,993 (11.5% lead) matters for general-purpose computing, database operations, and compression algorithms. Extended instructions at 45,477 vs 43,315 (5% lead) covers SIMD and vectorized workloads. These three wins are the 9115’s raison d’être: if your workload is dominated by math throughput, the 9115 is the better choice despite losing most other benchmarks.
FAQ
Q: Which CPU has higher single-thread performance?
A: The AMD EPYC 4484PX. It scores 4,119 in PassMark single-thread versus 3,360 for the 9115, an 18.4% advantage. It also leads in all Cinebench single-core tests: R15 (611 vs 597), R20 (2,547 vs 2,490), and R23 (6,065 vs 5,929), each by 2.2% to 2.3%.
Q: Does the EPYC 9115 win any benchmarks?
A: Yes, it wins 3 of 17 head-to-head tests. These are PassMark floating point math (113,853 vs 96,446, an 18% lead), integer math (181,807 vs 162,993, an 11.5% lead), and extended instructions (45,477 vs 43,315, a 5% lead).
Q: How do the core counts compare?
A: The EPYC 9115 has 16 cores and 32 threads, while the EPYC 4484PX has 12 cores and 24 threads. Despite having 4 fewer cores, the 4484PX wins the PassMark multithread test by 3.2% (50,547 vs 48,936) and all Cinebench multi-core tests by 2.2%.
Q: What is the largest performance gap between the two?
A: The largest margin is in PassMark find prime numbers, where the 4484PX scores 425 versus 289, a 32% advantage. The next largest is PassMark single-thread at 18.4% in favor of the 4484PX.
Q: Which CPU has a higher average benchmark score?
A: The EPYC 9115 has a higher average benchmark score of 69,288 compared to 67,822 for the 4484PX. However, both sit at the 94th percentile among all CPUs, and the 9115’s nearest rival, the Intel Core i7-14700K, is only 0.1% behind it.
Q: Are both CPUs in the same performance percentile?
A: Yes, both the EPYC 9115 and the EPYC 4484PX rank at the 94th percentile among all CPUs. Their average benchmark scores differ by only 1,466 points (69,288 vs 67,822).
Architecture Differences
The two processors are built on different architectures and process nodes. The EPYC 9115 uses Zen 5 architecture under the codename Turin, fabricated on a 4 nm process by TSMC. It has 16,630 million transistors across a die size of 2x 70.6 mm². The EPYC 4484PX uses Zen 4 architecture under the codename Raphael, fabricated on a 5 nm process by TSMC. It has 17,840 million transistors across a die size of 2x 71 mm². Despite the 9115 being on a smaller node, the 4484PX has more transistors, indicating a denser design or additional components.
Cache hierarchies differ significantly. The 9115 has 80 KB of L1 cache per core, 1 MB of L2 cache per core, and 64 MB of shared L3 cache. The 4484PX has 64 KB of L1 per core, 1 MB of L2 per core, and 128 MB of shared L3 cache. The 4484PX also includes a 1x 64 MB V-Cache slice, which likely contributes to its higher L3 total. Memory support is also different: the 9115 uses twelve-channel DDR5 memory with a bandwidth of 576.0 GB/s, while the 4484PX uses dual-channel DDR5 with a bandwidth of 83.2 GB/s. Both support ECC memory.
The 9115 offers 128 PCIe Gen 5 lanes (CPU only), while the 4484PX offers 28 PCIe Gen 5 lanes (CPU only). The 4484PX includes integrated Radeon Graphics, while the 9115 has N/A for integrated graphics. Both use different sockets: the 9115 uses AMD Socket SP5, and the 4484PX uses AMD Socket AM5. The 9115 is part of the EPYC 9005 series, while the 4484PX is part of the EPYC 4004 series. Neither has an unlocked multiplier.
Specification Differences
The most direct specification differences are core and thread counts: the EPYC 9115 has 16 cores and 32 threads, while the EPYC 4484PX has 12 cores and 24 threads. Clock speeds differ substantially: the 9115 has a base clock of 2.60 GHz and boost clock of 4.10 GHz, while the 4484PX has a base clock of 4.40 GHz and boost clock of 5.60 GHz. Thermal design power is close, with the 9115 at 125 TDP and the 4484PX at 120 TDP.
Memory configuration is a major differentiator. The 9115 supports twelve-channel memory with 576.0 GB/s bandwidth, while the 4484PX supports dual-channel memory with 83.2 GB/s bandwidth. This gives the 9115 nearly 7x the memory bandwidth on paper, though benchmark results do not always reflect this advantage. PCIe lanes also differ: the 9115 has 128 Gen 5 lanes (CPU only), while the 4484PX has 28 Gen 5 lanes (CPU only). The 4484PX includes integrated Radeon Graphics; the 9115 has no integrated graphics.
Process node and transistor count differ: the 9115 is on 4 nm with 16,630 million transistors, while the 4484PX is on 5 nm with 17,840 million transistors. Die size is similar at 2x 70.6 mm² for the 9115 and 2x 71 mm² for the 4484PX. Cache differs in L1 (80 KB per core vs 64 KB per core) and L3 (64 MB shared vs 128 MB shared), with the 4484PX adding a 1x 64 MB V-Cache slice. Release dates are also different: the 9115 launched on 2024-10-09, while the 4484PX launched on 2024-05-20. The 9115’s launch MSRP is $726, and the 4484PX’s launch MSRP is $599.