AMD EPYC 9335 vs AMD EPYC 9355P Comparison

The AMD EPYC 9335 and AMD EPYC 9355P are two 32-core server processors from the same Zen 5 (Turin) generation, both built on TSMC's 4 nm process and designed for the AMD Socket SP5 platform. Despite their shared core count and architecture, the benchmark data reveals a significant performance gulf between the two, driven by differences in clock speeds, cache size, and physical configuration. The following analysis breaks down the recorded measurements to understand where each processor excels and what the data implies for server deployment.

Head-to-Head Benchmarks

The head-to-head comparison is decisively one-sided. The AMD EPYC 9355P wins all 11 recorded benchmark comparisons, with the AMD EPYC 9335 failing to secure a single victory. The most dramatic disparity appears in the PassMark physics test, where the 9355P scores 13,515 against the 9335's 1,905, a delta of -85.9% for the 9335. This suggests a massive difference in raw computational throughput in workloads that stress core efficiency and clock speed.

The single-thread performance gap is equally telling. The 9355P records a PassMark single-thread score of 3,747, while the 9335 manages 2,732, a 27.1% deficit for the latter. This is a substantial difference for processors sharing the same Zen 5 architecture and 4.40 GHz boost clock. The data implies the 9335 is unable to sustain its maximum boost under the test conditions, likely due to power or thermal constraints, whereas the 9355P, with its higher 280 TDP, can maintain higher sustained clocks.

In multithreaded workloads, the gap persists. The 9355P's PassMark multithread score of 96,603 is 31.9% higher than the 9335's 65,811. This is a notable divergence considering both chips have 32 cores and 64 threads. The difference is likely attributable to the 9355P's larger 256 MB L3 cache versus the 9335's 128 MB, allowing better data locality and reduced memory latency in cache-sensitive parallel tasks.

Integer and floating-point math show similar patterns. The 9355P leads in integer math with 412,067 versus 346,291 (a 16% advantage) and in floating-point math with 256,635 versus 228,123 (an 11.1% advantage). The prime number finding test demonstrates the most extreme relative difference: the 9355P scores 1,044 compared to the 9335's 340, a 67.4% deficit for the 9335. This particular workload appears highly sensitive to the combination of clock speed and cache capacity.

Data compression and encryption tasks also favor the 9355P. In data compression, the 9355P scores 1,429,976 against 1,203,096 (15.9% higher), while encryption shows 80,961 versus 63,159 (22% higher). Even extended instructions, typically a more architecture-bound test, show a small but consistent 1.8% advantage for the 9355P. Random string sorting, another memory-latency-sensitive test, shows a 34% advantage for the 9355P, reinforcing the cache size impact.

The Verdict

The data presents a clear hierarchy: the AMD EPYC 9355P is the superior performer across every measured category. Its average benchmark score of 160,358 places it in the 98th percentile of all CPUs, while the 9335's average of 194,228 places it in the 99th percentile. This apparent contradiction, where the lower-scoring chip has a higher percentile, is explained by the rival sets. The 9335's nearest rivals are high-scoring Intel Xeon parts (the Xeon 6741P at 194,901, only 0.3% above the 9335), whereas the 9355P's nearest rivals are a mix of older EPYC and Xeon parts with lower average scores.

For single-threaded and lightly threaded workloads, the 9355P is the clear choice. Its 3.55 GHz base clock versus the 9335's 3.00 GHz gives it a substantial head start, and the benchmark data confirms this translates to real-world performance. The 27.1% single-thread advantage is too large to ignore for applications that depend on per-core speed, such as database query processing, certain high-frequency trading algorithms, or legacy software that cannot scale across many cores.

For heavily multithreaded, cache-intensive workloads, the 9355P again dominates. The combination of double the L3 cache and higher sustained clocks provides a 31.9% multithread advantage. Workloads like virtualization hosts, in-memory databases, and scientific simulations that fit within the 256 MB L3 cache would benefit significantly from the 9355P.

The 9335 is not without merit, but its strengths are relative. Its lower 210 TDP makes it a more thermally conservative option for dense server installations where power delivery or cooling is constrained. The data does not include power efficiency metrics directly, but the 9335's lower power envelope suggests it could be preferable in scenarios where peak performance is secondary to operational stability. However, from a pure performance standpoint as measured by the database, the 9355P is the superior processor in every recorded test.

Architecture Differences

Both processors share the same fundamental architecture: Zen 5, codename Turin, built on a 4 nm TSMC process, using the AMD Socket SP5. They are both classified as part of the EPYC 9005 series and were released on the same date. The core complexity is identical: each core has 80 KB of L1 and 1 MB of L2 cache.

The first major divergence is transistor count and die configuration. The 9335 uses 33,260 million transistors spread across 4x 70.6 mm² dies, while the 9355P uses 66,520 million transistors across 8x 70.6 mm² dies. This doubling of compute dies and transistor count in the 9355P is the primary architectural differentiator, allowing it to accommodate twice the L3 cache: 256 MB shared versus 128 MB shared. The 9335's smaller die count suggests a more power-efficient design, which aligns with its lower 210 TDP, while the 9355P's larger configuration requires the higher 280 TDP.

The base clocks differ significantly: 3.00 GHz for the 9335 and 3.55 GHz for the 9355P. Both share the same 4.40 GHz boost clock, but the 9355P's higher base clock means it starts from a higher performance floor. The practical implication is that under sustained all-core loads, the 9355P can maintain higher minimum performance due to its higher base frequency.

Memory support is identical: DDR5 with a twelve-channel bus and 576.0 GB/s of bandwidth, along with ECC support. PCIe connectivity is also the same: Gen 5 with 128 lanes from the CPU. Neither processor includes integrated graphics, and both have locked multipliers, meaning no overclocking via multiplier adjustment.

FAQ

Q: Which processor has the higher base clock speed?

A: The AMD EPYC 9355P has a base clock of 3.55 GHz, while the AMD EPYC 9335 has a base clock of 3.00 GHz. Both share the same 4.40 GHz boost clock.

Q: How much larger is the L3 cache on the EPYC 9355P?

A: The EPYC 9355P has 256 MB of shared L3 cache, exactly double the 128 MB found on the EPYC 9335. This is reflected in the 9355P's higher transistor count of 66,520 million versus 33,260 million.

Q: What is the single-thread performance difference?

A: The EPYC 9355P scores 3,747 in the PassMark single-thread test, while the EPYC 9335 scores 2,732. This represents a 27.1% advantage for the 9355P.

Q: Are both processors built on the same manufacturing process?

A: Yes, both are built on TSMC's 4 nm process and use the Zen 5 architecture. They are both part of the EPYC 9005 series and use the AMD Socket SP5.

Q: Which processor has a higher average benchmark score?

A: The EPYC 9335 has a higher average benchmark score of 194,228, compared to the EPYC 9355P's 160,358. However, this is influenced by the different sets of rival processors used for comparison, and the head-to-head tests show the 9355P winning all 11 comparisons.

Q: What is the difference in TDP?

A: The EPYC 9335 has a TDP of 210, while the EPYC 9355P has a TDP of 280. This represents a 70-watt difference, likely contributing to the 9355P's higher sustained performance.

Where Each One Wins

The AMD EPYC 9355P wins in every benchmark category recorded in the database. Its strengths are most pronounced in physics (85.9% faster), prime number finding (67.4% faster), and multithreaded workloads (31.9% faster). These results point to workloads that benefit from high core clocks and large cache capacity, such as real-time physics simulation, cryptographic key generation, and heavily parallel server-side processing.

The 9355P also excels in single-threaded performance, making it suitable for applications where per-core speed is the bottleneck. The 27.1% single-thread advantage over the 9335 is substantial for a same-architecture comparison. This makes the 9355P a strong candidate for database servers running transaction-heavy workloads, web servers handling high request rates, or any application with serial code sections that cannot be fully parallelized.

The AMD EPYC 9335's wins are not in benchmark scores but in operational characteristics. Its 210 TDP versus the 9355P's 280 implies lower power draw and heat output, which can be advantageous in densely packed server racks or environments with power delivery limitations. The 9335 also uses half the transistor count (33,260 million versus 66,520 million) and half the dies (4x 70.6 mm² versus 8x 70.6 mm²), suggesting a more streamlined production footprint. For organizations prioritizing energy efficiency over raw performance, the 9335 may be the more appropriate choice, though the database does not provide direct power efficiency measurements.

Specification Differences

The following specifications differ between the two processors:

• Base Clock: 3.00 GHz (9335) versus 3.55 GHz (9355P)
• TDP: 210 (9335) versus 280 (9355P)
• Transistors: 33,260 million (9335) versus 66,520 million (9355P)
• Die Size: 4x 70.6 mm² (9335) versus 8x 70.6 mm² (9355P)
• L3 Cache: 128 MB shared (9335) versus 256 MB shared (9355P)
• Part Number: 100-000001149 (9335) versus 100-000001521 (9355P)
• Launch MSRP: $3178 (9335) versus $2998 (9355P)

All other specifications are identical: 32 cores, 64 threads, 4.40 GHz boost clock, AMD Socket SP5, Zen 5 architecture, Turin codename, 4 nm process, TSMC foundry, 80 KB L1 per core, 1 MB L2 per core, DDR5 memory support, twelve-channel memory bus, 576.0 GB/s memory bandwidth, ECC memory, Gen 5 PCIe with 128 lanes, no integrated graphics, server/workstation market segment, active production status, same release date, and locked multipliers.