AMD EPYC 9355P vs AMD EPYC 9455P Comparison

Head-to-Head Benchmarks

The data shows a decisive overall victory for the AMD EPYC 9455P, which wins 15 of the 17 recorded head-to-head benchmark comparisons. The margins are substantial in most heavily threaded workloads, reflecting the 48-core versus 32-core configuration.

The largest single advantage appears in PassMark integer math, where the 9455P scores 606,239 against the 9355P's 412,067, a 47.1% lead. Data encryption follows closely, with the 9455P at 115,403 versus 80,961, a 42.5% advantage. Floating-point math shows a 39.4% gap (357,783 versus 256,635), and random string sorting lands at 37.4% (242,701 versus 176,697). These are not marginal differences; they represent the kind of scaling you expect when core count rises by 50%.

Data compression favors the 9455P by 34.9%, with scores of 1,928,897 against 1,429,976. Extended instructions show a 27.7% lead (137,485 versus 107,622), and physics simulation is 28.1% ahead (17,315 versus 13,515). PassMark multithread shows a 21% gap (116,927 versus 96,603), which is closer than the raw core-count math might suggest, but still a clear win.

The Cinebench suite tells a remarkably consistent story. Across R15, R20, and R23, both multicore and singlecore, the 9455P leads by exactly 20% in every single test. Multicore R23 scores are 99,206 versus 82,666, singlecore R23 is 14,005 versus 11,670, R20 multicore is 41,666 versus 34,719, and R20 singlecore is 5,882 versus 4,901. R15 shows the same pattern: 9,999 versus 8,332 multicore and 1,411 versus 1,176 singlecore. The uniformity of that 20% delta across all six Cinebench runs suggests consistent per-core performance parity, with the difference driven almost entirely by the additional cores.

The only benchmark where the 9355P wins is PassMark single-thread, where it scores 3,747 against 3,745, a 0.1% margin. That is effectively a tie, but it does appear in both single-thread and singlethread entries in the database. Find prime numbers is close as well: 1,107 versus 1,044, a 6% lead for the 9455P, which is the smallest multicore gap recorded.

The 9455P's average benchmark score is 217,854, placing it in the 99th percentile of all CPUs in the database. The 9355P averages 160,358, in the 98th percentile. For context, the 9455P sits 3.8% above the Intel Xeon w9-3595X and 11.8% above the Intel Xeon 6741P, while trailing the Intel Xeon 6747P by 8.6%. The 9355P's nearest rivals include the AMD EPYC 7663 at 1% behind, the Intel Xeon 676X at 1.1% behind, and the AMD EPYC 9375F at 1.3% ahead, a tight cluster.

FAQ

Q: Which processor is faster in multi-threaded workloads?

A: The AMD EPYC 9455P wins every multicore benchmark in the database. PassMark multithread shows 116,927 versus 96,603 (21% ahead), and Cinebench R23 multicore shows 99,206 versus 82,666 (20% ahead).

Q: Is there any workload where the 9355P wins?

A: Yes, PassMark single-thread, where it scores 3,747 versus 3,745 for the 9455P, a 0.1% margin. This appears in both the single-thread and singlethread entries.

Q: How do the two compare in encryption performance?

A: The 9455P is substantially ahead. PassMark data encryption scores 115,403 versus 80,961, a 42.5% advantage.

Q: What is the core and thread difference?

A: The 9455P has 48 cores and 96 threads, while the 9355P has 32 cores and 64 threads.

Q: Do they share the same socket and memory support?

A: Yes. Both use AMD Socket SP5, support DDR5 memory, and have a twelve-channel memory bus with 576.0 GB/s bandwidth.

Q: What is the difference in average benchmark score?

A: The 9455P averages 217,854, which is 36% above the 9355P's 160,358. The 9455P sits in the 99th percentile of all CPUs, the 9355P in the 98th.

The Verdict

The benchmark data points to a clear split. For workloads that scale with core count, the AMD EPYC 9455P is the stronger choice. It wins every multicore benchmark in the database, often by large margins: 47.1% in integer math, 42.5% in encryption, 39.4% in floating-point math, and a consistent 20% across all Cinebench runs. The 48-core configuration delivers real, measurable performance advantages across rendering, compression, and simulation workloads.

The AMD EPYC 9355P is the pick for situations where single-thread performance is the priority and the workload does not use the extra cores. Its PassMark single-thread score of 3,747 edges out the 9455P's 3,745, though the margin is negligible. With 32 cores and 64 threads, it still delivers strong multi-threaded performance, as its 98th percentile ranking shows, but it trails the 9455P by 36% in average benchmark score.

Buyers who run heavily threaded server workloads, database compression, encryption pipelines, or scientific computing should choose the 9455P. The performance gap is consistent and often dramatic. Buyers who need a capable server processor with high per-core throughput and do not require the full 48-core complement should consider the 9355P. The single-thread scores are effectively identical, so the decision rests entirely on whether the workload can use 16 additional cores.

Neither processor offers an unlocked multiplier, and both are active production parts from the EPYC 9005 series. The 9455P carries a launch MSRP of $4819, while the 9355P has a launch MSRP of $2998. Both are server and workstation parts with ECC memory support and no integrated graphics.

Specification Differences

The core count is the primary differentiator. The 9455P offers 48 cores and 96 threads, while the 9355P offers 32 cores and 64 threads. Base clocks differ as well: the 9455P runs at 3.15 GHz, the 9355P at 3.55 GHz. Boost clocks are identical at 4.40 GHz for both.

Thermal design power differs by 20 watts, with the 9455P rated at 300 and the 9355P at 280. Both processors share the same socket, AMD Socket SP5, and both are built on the same 4 nm process at TSMC. Transistor counts are identical at 66,520 million, and die size is the same at 8x 70.6 mm².

Cache configurations match exactly: 80 KB of L1 per core, 1 MB of L2 per core, and 256 MB of shared L3. Memory support is identical as well, with DDR5, a twelve-channel memory bus, and 576.0 GB/s of bandwidth. Both support ECC memory. PCIe is Gen 5 with 128 lanes (CPU only) on both parts.

The part numbers differ: 100-000001563 for the 9455P and 100-000001521 for the 9355P. Both were released on the same date, October 9, 2024. The launch MSRP differs, with the 9455P at $4819 and the 9355P at $2998.

Architecture Differences

Both processors are built on the same Zen 5 architecture with the Turin codename, part of the EPYC 9005 series. The process node is 4 nm at TSMC for both, and the transistor count and die size are identical. The architecture is fundamentally the same; the differences come from core count and clock configuration.

The 9455P has 50% more cores than the 9355P (48 versus 32) and 50% more threads (96 versus 64). The 9355P compensates with a higher base clock of 3.55 GHz versus 3.15 GHz, but both boost to 4.40 GHz. The identical per-core L1 and L2 cache amounts mean each core has the same private cache allocation, and the 256 MB shared L3 is the same on both.

The higher base clock on the 9355P explains its single-thread edge, however small. The 9455P's advantage in multi-threaded workloads comes entirely from the additional cores, since the per-core architecture, cache hierarchy, memory subsystem, and boost clock are all identical. The 20 watt TDP difference reflects the additional cores in the 9455P.

Both have the same memory bus width (twelve-channel), same bandwidth (576.0 GB/s), and same PCIe lane count (128 Gen 5 lanes). Neither has integrated graphics, and both support ECC memory. The architecture story is simple: the same Zen 5 design, scaled to different core counts with different base clocks.

Where Each One Wins

The AMD EPYC 9455P is the clear winner for multi-threaded compute. The PassMark integer math result of 606,239 versus 412,067 (47.1% ahead) makes it the choice for integer-heavy workloads such as compilation, data processing, and general server tasks. Data encryption at 115,403 versus 80,961 (42.5% ahead) points to a strong fit for security, VPN, and cryptographic workloads. Floating-point math at 357,783 versus 256,635 (39.4% ahead) suits scientific computing, simulation, and financial modeling.

Data compression, scoring 1,928,897 against 1,429,976 (34.9% ahead), makes the 9455P the better option for storage servers, backup systems, and database workloads that compress data on the fly. Random string sorting at 242,701 versus 176,697 (37.4% ahead) reinforces the multi-threaded sorting and indexing advantage. Physics simulation at 17,315 versus 13,515 (28.1% ahead) and extended instructions at 137,485 versus 107,622 (27.7% ahead) round out a comprehensive multi-threaded win.

The Cinebench results, all at exactly 20% ahead for the 9455P, make it the pick for rendering and content creation workloads that use all available cores. The 99th percentile ranking versus the 98th percentile for the 9355P confirms the overall standing.

The AMD EPYC 9355P wins only in PassMark single-thread performance, 3,747 versus 3,745, a 0.1% margin. That makes it the choice for workloads that are strictly single-threaded and where every bit of per-core performance matters. The higher base clock of 3.55 GHz contributes here, even though the 4.40 GHz boost clock is the same on both. For legacy applications, license-limited software that cannot use many cores, or lightly threaded network services, the 9355P offers essentially identical per-core performance at a lower core count.

The 9355P also makes sense in power-constrained environments. Its 280 watt TDP is lower than the 9455P's 300 watt rating, and with 32 cores rather than 48, it produces less heat under load while still delivering 98th percentile performance. For rack density and cooling considerations, that 20 watt difference per socket can matter across a large deployment.

In short, the 9455P wins everywhere that core count matters, and the 9355P wins only where it does not. The find prime numbers benchmark, at 1,107 versus 1,044 (6% ahead for the 9455P), shows that even in less parallel workloads, the 9455P still leads. The single-thread result is the only exception, and it is a tie in practical terms.