AMD EPYC 9555P vs AMD Ryzen Threadripper PRO 9995WX Comparison

The recorded data tells a lopsided story: the AMD Ryzen Threadripper PRO 9995WX wins all 17 head-to-head benchmarks against the AMD EPYC 9555P, often by wide margins. Both chips are Zen 5 designs on TSMC's 4 nm node, aimed at the server and workstation segment, but they serve different roles. The Threadripper PRO 9995WX leads the database at the 100th percentile versus all CPUs in the database, while the EPYC 9555P sits at the 99th percentile, still elite territory but a clear step behind in raw throughput.

FAQ

Q: Which CPU is faster overall?

A: The Ryzen Threadripper PRO 9995WX. It has an average benchmark score of 412068 versus 287066 for the EPYC 9555P, and it won every single one of the 17 recorded head-to-head tests.

Q: How big is the multi-core gap?

A: In Cinebench R23 multi-core, the Threadripper PRO 9995WX scored 148601 against 115186, a 29% lead. In PassMark multithread the gap widens to 38.5% (171200 vs 123576).

Q: Is the single-core performance different too?

A: Yes. The 9995WX scored 4542 in PassMark single thread versus 3410 for the EPYC 9555P, a 33.2% advantage, driven by its 5.40 GHz boost clock compared to 4.40 GHz.

Q: Which platform has more memory bandwidth?

A: The EPYC 9555P. Its twelve-channel DDR5 bus delivers 576.0 GB/s, well above the eight-channel, 409.6 GB/s setup of the 9995WX. Both support ECC memory.

Q: Can either CPU be overclocked?

A: Only the Threadripper PRO 9995WX has an unlocked multiplier. The EPYC 9555P does not.

Q: What was the launch MSRP of each?

A: The Ryzen Threadripper PRO 9995WX launched at $11700, and the EPYC 9555P launched at $7983.

Architecture Differences

Both processors are built on AMD's Zen 5 architecture and manufactured on TSMC's 4 nm process, which makes this a same-generation comparison rather than a cross-generation one. The differences come down to how each product carves up that silicon.

The Threadripper PRO 9995WX uses the "Shimada Peak" implementation of Zen 5 with 96 cores and 192 threads. It packs 99,780 million transistors across 12 dies of 70.6 mm² each. The EPYC 9555P uses the "Turin" server implementation with 64 cores and 128 threads, containing 66,520 million transistors across 8 dies of the same 70.6 mm² per-die size. The core-count advantage of 32 cores and 64 extra threads is the single biggest reason the 9995WX dominates the multi-core charts.

Cache configuration differs as well. The 9995WX carries 64 KB of L1 per core, 1 MB of L2 per core, and 384 MB of L3. The 9555P has a larger 80 KB of L1 per core, the same 1 MB of L2 per core, and 256 MB of shared L3. The 9995WX holds the L3 advantage by 128 MB, which matters for large working sets in rendering and simulation.

Clock behavior splits the two. The 9995WX runs a 2.50 GHz base clock but boosts to 5.40 GHz, while the 9555P runs a higher 3.20 GHz base but tops out at 4.40 GHz. That one full gigahertz of extra boost headroom on the Threadripper explains the 29% single-core leads in every Cinebench test. TDP is close: 350 W for the 9995WX and 360 W for the 9555P.

Platform features are where the EPYC pulls back ground. It uses AMD Socket SP5 with a twelve-channel DDR5 memory bus, while the Threadripper PRO uses Socket sTR5 with eight-channel DDR5. Both offer 128 lanes of PCIe Gen 5 from the CPU and both support ECC, but neither has integrated graphics.

The 9995WX was released on 2025-07-22; the 9555P arrived earlier, on 2024-10-09. Both remain in active production.

The Verdict

The data makes the split straightforward. If the workload is compute-bound, the Threadripper PRO 9995WX is the clear pick: 17 wins from 17 tests, leads ranging from 9% in physics to 52.9% in integer math, plus an unlocked multiplier for tuning. It also sits at the 100th percentile against all CPUs in the database and outperforms its own nearest rivals, including the EPYC 9655P (3.6% behind) and EPYC 9535 (8.6% behind).

The EPYC 9555P earns its place when memory bandwidth and density matter more than raw core count. Its 576.0 GB/s of bandwidth is a 40% increase over the 9995WX's 409.6 GB/s, and it wins none of the benchmark tests but competes closely with its own peer group: the Intel Xeon 696X is within 0.3%, the AMD EPYC 9565 within 0.6%, the Intel Xeon 6780E within 2.4%, and the AMD Ryzen Threadripper 9970X within 2.6%. In that server neighborhood, the 9555P is a finely balanced part; against the 9995WX it is simply outgunned in throughput.

Note the launch MSRP figures before deciding: $11700 for the 9995WX and $7983 for the 9555P.

Specification Differences

• Series: Ryzen Threadripper 9000 series (9995WX) vs EPYC 9005 series (9555P)
• Codename: Shimada Peak vs Turin
• Cores / Threads: 96 / 192 vs 64 / 128
• Base clock: 2.50 GHz vs 3.20 GHz
• Boost clock: 5.40 GHz vs 4.40 GHz
• TDP: 350 W vs 360 W
• Socket: AMD Socket sTR5 vs AMD Socket SP5
• Transistors: 99,780 million vs 66,520 million
• Die configuration: 12x 70.6 mm² vs 8x 70.6 mm²
• L1 cache: 64 KB per core vs 80 KB per core
• L3 cache: 384 MB vs 256 MB (shared)
• Memory bus: Eight-channel vs Twelve-channel
• Memory bandwidth: 409.6 GB/s vs 576.0 GB/s
• Release date: 2025-07-22 vs 2024-10-09
• Launch MSRP: $11700 vs $7983
• Multiplier unlocked: Yes vs No
• Part number: 100-000001361 vs 100-000001523

Identical between the two: Zen 5 architecture, 4 nm TSMC process, 1 MB L2 per core, DDR5 support, ECC support, 128 PCIe Gen 5 lanes from the CPU, no integrated graphics, and active production status.

Head-to-Head Benchmarks

The 9995WX swept the board, 17 tests to zero.

Cinebench shows a remarkably consistent story. R15 multi-core: 14978 vs 11610, a 29% win. R20 multi-core: 62412 vs 48378, again 29%. R23 multi-core: 148601 vs 115186, 29% once more. Single-core results are just as uniform: 2114 vs 1638 in R15, 8811 vs 6829 in R20, and 20979 vs 16261 in R23, all 29% or 29.1% in the Threadripper's favor. Consistency at that level indicates the lead comes from structural advantages (more cores, higher boost) rather than test-specific quirks.

PassMark tilts even harder toward the 9995WX. Integer math is the biggest blowout: 1203634 vs 787106, a 52.9% win. Random string sorting follows at 49.4% (418973 vs 280398), floating point math at 49.1% (725066 vs 486407), extended instructions at 41.6% (270647 vs 191082), data compression at 41.1% (3723652 vs 2639400), encryption at 38.8% (206662 vs 148896), find prime numbers at 38.3% (1476 vs 1067), and the overall multithread score at 38.5% (171200 vs 123576).

The narrowest margin is PassMark physics, where the 9995WX scored 16860 against 15474, a 9% edge. Single thread lands in between at 33.2% (4542 vs 3410).

Where Each One Wins

Ryzen Threadripper PRO 9995WX wins on: rendering (all six Cinebench results, 29% each), heavy compute (integer math 52.9%, floating point 49.1%), data workloads (compression 41.1%, encryption 38.8%, string sorting 49.4%), single-threaded responsiveness (33.2%), physics simulation (9%), and enthusiast tuning via the unlocked multiplier. Its 384 MB of L3 also favors large in-memory datasets. Anyone doing 3D rendering, code compilation, or simulation will see the full 29 to 53% advantage in practice.

EPYC 9555P wins on: memory bandwidth, with twelve channels delivering 576.0 GB/s versus 409.6 GB/s, and per-core L1 capacity at 80 KB versus 64 KB. It records zero benchmark wins against the 9995WX, so its case rests entirely on platform characteristics: bandwidth-hungry server workloads that scale with memory channels rather than core count, and a peer group (Xeon 696X, EPYC 9565) where it is highly competitive. Its higher 3.20 GHz base clock also suggests steadier all-core sustained behavior, though the recorded data shows the 9995WX ahead regardless.

The bottom line from the data: for outright performance the 9995WX wins everywhere; the 9555P is a bandwidth-first server platform, not a benchmark rival.