AMD EPYC 9655 vs AMD Ryzen Threadripper PRO 9995WX Comparison

The AMD Ryzen Threadripper PRO 9995WX and AMD EPYC 9655 are both 96-core, 192-thread Zen 5 processors built on TSMC’s 4 nm process, yet they target different segments of the high-performance computing market. The Threadripper PRO 9995WX ships with a higher boost clock and a workstation-oriented feature set, while the EPYC 9655 counters with a higher base clock, a twelve-channel memory bus, and a larger TDP envelope. Benchmark data shows the Threadripper PRO 9995WX winning 14 of 17 head-to-head comparisons, but the EPYC 9655 takes decisive victories in specific physics and sorting workloads. This analysis quantifies those differences, interprets the scores against nearest rivals, and outlines the architectural and specification gaps that explain the performance deltas.

Head-to-Head Benchmarks

The Cinebench suite delivers a consistent and emphatic win for the Threadripper PRO 9995WX across every test. In Cinebench R15 multi-core, the Threadripper PRO 9995WX scores 14978 against the EPYC 9655’s 13373, a 12% advantage. That same 12% delta appears in R15 single-core (2114 vs 1887), R20 multi-core (62412 vs 55722), R20 single-core (8811 vs 7866), R23 multi-core (148601 vs 132672), and R23 single-core (20979 vs 18730). The uniformity of this 12% gap across both multi-threaded and single-threaded rendering workloads indicates a fundamental clock-speed advantage rather than a scaling edge — the Threadripper PRO 9995WX’s 5.40 GHz boost clock versus the EPYC 9655’s 4.50 GHz is the likely driver.

PassMark results paint a more nuanced picture. The Threadripper PRO 9995WX leads in data compression with 3627619 versus 3271019, a 10.9% margin, and in extended instructions with 247436 versus 203302, a substantial 21.7% lead. Floating-point math favors the Threadripper PRO 9995WX at 707600 vs 662949 (6.7%), and integer math follows at 1220090 vs 1139221 (7.1%). The multi-thread score shows a 12% advantage (174825 vs 156085), matching the Cinebench pattern. Single-thread performance is the largest PassMark win for the Threadripper PRO 9995WX: 4565 vs 3845, an 18.7% delta.

The EPYC 9655, however, wins three workloads. PassMark physics shows a dramatic 22.9% lead for the EPYC 9655 (25958 vs 20011). Random string sorting goes to the EPYC 9655 by a narrower 3.1% margin (440608 vs 426857). Find prime numbers is a near-tie, with the EPYC 9655 edging ahead at 1599 vs 1573, a 1.6% advantage. Data encryption is effectively a draw — the Threadripper PRO 9995WX leads by just 2.4% (215680 vs 210541).

Context from nearest rivals strengthens the Threadripper PRO 9995WX’s position. Its average benchmark score of 406395 places it 8.8% ahead of the EPYC 9655 (373484), 2.5% ahead of the EPYC 9655P (396673), and 11.5% ahead of the EPYC 9754 (364371). Only the EPYC 9745 (425973) sits above it, by 4.6%. For the EPYC 9655, its average score is 5.8% behind the EPYC 9655P and 8.4% ahead of the Ryzen Threadripper 9980X (344569). Both processors hit the 100th percentile among all CPUs, indicating top-tier placement overall, but the Threadripper PRO 9995WX’s aggregate score is measurably higher.

FAQ

Q: Which processor has a higher boost clock and how does that impact single-threaded benchmarks?

A: The AMD Ryzen Threadripper PRO 9995WX boosts to 5.40 GHz, while the AMD EPYC 9655 reaches 4.50 GHz. This 0.90 GHz difference aligns with single-thread results: the Threadripper PRO 9995WX leads by 12% in Cinebench R23 single-core (20979 vs 18730) and by 18.7% in PassMark single-thread (4565 vs 3845).

Q: How do the two processors compare in memory bandwidth and channel count?

A: The EPYC 9655 uses a twelve-channel memory bus with 576.0 GB/s of bandwidth, whereas the Threadripper PRO 9995WX uses eight channels with 409.6 GB/s. This gives the EPYC 9655 a 166.4 GB/s advantage, which may contribute to its wins in memory-sensitive workloads like random string sorting.

Q: Are both processors on the same manufacturing node and with the same transistor count?

A: Yes, both are fabricated on TSMC’s 4 nm process with 99,780 million transistors. Their die sizes are also identical at 12x 70.6 mm². The performance differences therefore stem from clock speeds, memory architecture, and cache configuration, not from silicon density.

Q: Which processor wins in PassMark physics, and what is the score difference?

A: The EPYC 9655 wins PassMark physics by a wide margin, scoring 25958 against the Threadripper PRO 9995WX’s 20011. That is a 22.9% advantage, the largest single-workload gap in either direction across the entire head-to-head set.

Q: What is the average benchmark score for each processor, and how does that rank against their nearest rivals?

A: The Threadripper PRO 9995WX has an average score of 406395, which is 8.8% higher than the EPYC 9655’s 373484. The Threadripper PRO 9995WX also leads the EPYC 9655P by 2.5% and the EPYC 9754 by 11.5%, while trailing the EPYC 9745 by 4.6%. The EPYC 9655 trails the 9655P by 5.8% and leads the Threadripper 9980X by 8.4%.

Q: Do both processors support the same PCIe configuration?

A: Yes, both support PCIe Gen 5 with 128 lanes (CPU only). This is a point of parity, meaning expansion capability does not differentiate the two in benchmark terms.

Architecture Differences

Both CPUs share the Zen 5 architecture, but they belong to different product families with distinct codenames. The Threadripper PRO 9995WX is codenamed Shimada Peak, while the EPYC 9655 is codenamed Turin. This is not a generational gap — both are Zen 5, 4 nm TSMC parts with identical transistor counts (99,780 million) and die sizes (12x 70.6 mm²). The architectural divergence lies in cache and memory hierarchy.

The L1 cache differs: the Threadripper PRO 9995WX has 64 KB per core, while the EPYC 9655 has 80 KB per core. That 16 KB per-core difference could influence the EPYC 9655’s performance in physics simulations and sorting tasks, which often rely on tightly coupled data. Both share 1 MB L2 per core and 384 MB shared L3, so the L1 delta is the only cache distinction.

Memory architecture is a major split. The EPYC 9655 uses a twelve-channel memory bus, delivering 576.0 GB/s, while the Threadripper PRO 9995WX uses eight channels at 409.6 GB/s. This 40.6% bandwidth advantage for the EPYC 9655 likely explains its edge in random string sorting (3.1%) and contributes to the physics win. The Threadripper PRO 9995WX compensates with a higher boost clock (5.40 vs 4.50 GHz), which drives most of its single-thread and many multi-thread victories.

The sockets are different: the Threadripper PRO 9995WX uses AMD Socket sTR5, while the EPYC 9655 uses AMD Socket SP5. This means they are not interchangeable in existing platforms, and each requires a motherboard designed for its respective socket. Both support DDR5 memory and ECC, and both have no integrated graphics. The Threadripper PRO 9995WX has an unlocked multiplier, whereas the EPYC 9655 is locked — a feature that matters for overclocking but does not appear in the benchmark scores.

Specification Differences

The two processors differ in several key specification fields. Base clocks are close but not identical: the EPYC 9655 runs at 2.60 GHz, while the Threadripper PRO 9995WX runs at 2.50 GHz — a 0.10 GHz advantage for the EPYC. Boost clocks reverse the trend decisively: the Threadripper PRO 9995WX reaches 5.40 GHz, versus 4.50 GHz for the EPYC 9655, a 0.90 GHz gap. TDP also differs, with the EPYC 9655 rated at 400 W and the Threadripper PRO 9995WX at 350 W — a 50 W difference.

Memory channels and bandwidth are distinct as noted: twelve channels and 576.0 GB/s for the EPYC 9655, versus eight channels and 409.6 GB/s for the Threadripper PRO 9995WX. L1 cache per core is 80 KB for the EPYC 9655 and 64 KB for the Threadripper PRO 9995WX. The socket differs (SP5 vs sTR5), as does the multiplier lock state (unlocked vs locked). Release dates are not listed in the fact pack, but the production status is Active for both. Launch MSRP is $11852 for the EPYC 9655 and $11700 for the Threadripper PRO 9995WX — a $152 difference.

All other specification fields match: 96 cores, 192 threads, 1 MB L2 per core, 384 MB shared L3, DDR5 memory support, ECC support, PCIe Gen 5 with 128 lanes, no integrated graphics, and the same market segment (Server/Workstation).

The Verdict

The data points to the Threadripper PRO 9995WX as the stronger overall performer in this head-to-head. It wins 14 of 17 benchmarks, including all six Cinebench tests, and holds an 8.8% average score advantage over the EPYC 9655. Its 12% lead across the entire Cinebench suite, plus an 18.7% single-thread PassMark margin, makes it the clear choice for workloads that prioritize rendering, compression, extended instructions, and general multi-threaded throughput. The fact that it achieves this with a lower TDP (350 W vs 400 W) and a lower launch MSRP ($11700 vs $11852) reinforces its position as the more efficient option in this pairing.

The EPYC 9655 is not without purpose, however. Its 22.9% victory in PassMark physics and 3.1% win in random string sorting point to a memory-bandwidth advantage (576.0 GB/s vs 409.6 GB/s) that matters in specific data-intensive tasks. The 1.6% edge in find prime numbers is marginal but real. For users running physics simulations, string-heavy data processing, or workloads that scale with memory channels, the EPYC 9655’s twelve-channel bus and higher base clock (2.60 vs 2.50 GHz) provide a tangible benefit.

The verdict depends on workload mix. For a workstation that needs peak single-thread speed, broad multi-core dominance, and an unlocked multiplier for tuning, the Threadripper PRO 9995WX is the data-backed selection. For a server environment where memory bandwidth drives performance and physics or sorting workloads dominate, the EPYC 9655’s wins justify its selection. Both sit at the 100th percentile of all CPUs, so neither is a weak choice — but the Threadripper PRO 9995WX’s 14-to-3 win count and higher average score make it the default recommendation unless the EPYC 9655’s specific strengths align with the intended application.