AMD EPYC 9535 vs AMD EPYC 9655 Comparison

FAQ

Q: Which processor has the higher average benchmark score?

A: The AMD EPYC 9535 has a higher average benchmark score of 379,408, while the AMD EPYC 9655 scores 373,479. This puts the 9535 1.6% ahead of the 9655, according to the nearestRivals data.

Q: How does the AMD EPYC 9655 compare to the Intel Xeon 6960P?

A: The EPYC 9655 has an average score of 373,479, which is 2.3% higher than the Intel Xeon 6960P's 365,194. The EPYC 9535, meanwhile, is 3.9% ahead of that same Intel part.

Q: In which benchmark does the EPYC 9655 show its most dominant win?

A: The largest margin is in the PassMark physics test, where the EPYC 9655 scores 25,947 versus the EPYC 9535's 3,834 — a massive 85.2% difference. This is the single biggest delta in the head-to-head data.

Q: What is the single-thread performance difference between the two?

A: The EPYC 9655 leads in PassMark single-thread with a score of 3,847, which is only 3.3% higher than the EPYC 9535's 3,720. Both chips post identical scores across the two single-thread entries in the data.

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

A: Yes. Both the AMD EPYC 9535 and AMD EPYC 9655 are fabricated by TSMC on a 4 nm process node, and both use the Zen 5 architecture with the Turin codename.

Q: Which chip has more cores and threads?

A: The AMD EPYC 9655 has 96 cores and 192 threads, while the AMD EPYC 9535 has 64 cores and 128 threads. The 9655 also carries a higher transistor count of 99,780 million versus 66,520 million.

The Verdict

The benchmark data paints a clear picture: the AMD EPYC 9655 is the superior multi-threaded performer, winning all 11 head-to-head benchmark comparisons. Its advantages range from a narrow 3.3% in single-thread to a colossal 85.2% in physics. For workloads that scale with core count — data compression, encryption, integer math, floating-point math — the 9655 is the definitive choice.

However, the average benchmark score tells a more nuanced story. The EPYC 9535's average of 379,408 edges out the 9655's 373,479 by 1.6%. This suggests that in certain mixed or lightly threaded workloads, the 9535 may hold its own, despite losing every individual test in the head-to-head set. The 9535 also consumes less power, with a 300 TDP versus the 9655's 400 TDP, making it a more energy-conscious option for dense server deployments.

The verdict splits by use case. The EPYC 9655 is for users who need maximum throughput across every available core — database compression, encryption pipelines, and physics simulations will all benefit disproportionately. The EPYC 9535, with its lower TDP and comparable average score, suits environments where power density matters more than raw core count, or where the 64-core configuration is sufficient for the workload mix. Both are 100th percentile CPUs, but the 9655 is the brute-force king while the 9535 is the efficiency pick.

Head-to-Head Benchmarks

The EPYC 9655 dominates every single benchmark in the head-to-head set, but the margins vary wildly by workload type. The most lopsided result is in PassMark physics, where the 9655 scores 25,947 against the 9535's 3,834 — an 85.2% advantage. This is a physics simulation test that scales heavily with core count, and the 96-core 9655 crushes the 64-core 9535.

Data compression shows a similar but less extreme pattern. The 9655 hits 3,271,896 versus 2,308,822, a 29.4% lead. Encryption follows with a 39.5% gap (210,555 vs 127,372), and random string sorting shows a 43.7% difference (439,682 vs 247,506). These are all heavily parallel workloads where the extra 32 cores and 64 threads of the 9655 translate directly into higher throughput.

Integer math is another strong showing for the 9655, scoring 1,139,161 against 730,281 — a 35.9% advantage. Floating-point math follows at 31.1% (662,958 vs 457,047). The extended instructions test is closer, with the 9655 leading by 13.5% (203,285 vs 175,784), suggesting that SIMD-heavy code benefits less from raw core count.

The find prime numbers test shows a 45.3% gap (1,598 vs 874), and the multithread score is 26.6% higher on the 9655 (156,110 vs 114,528). The closest contest is single-thread performance, where the 9655 leads by just 3.3% (3,847 vs 3,720). This indicates that per-core IPC is nearly identical between the two Zen 5 parts, and the 9655's small edge comes from its higher boost clock of 4.50 GHz versus 4.30 GHz.

The 9535 wins zero head-to-head tests. Its average score advantage comes from the broader benchmark suite, which includes Cinebench results only listed for the 9655, not the 9535. The 9655's Cinebench scores (13,373 in R15 multi, 132,672 in R23 multi) are not directly comparable since the 9535 lacks those entries in the data.

Specification Differences

The core count difference is the headline spec: the EPYC 9655 packs 96 cores and 192 threads, while the EPYC 9535 offers 64 cores and 128 threads. This 50% core increase is the primary driver of the 9655's benchmark dominance in parallel workloads.

Clock speeds also differ. The 9655 has a base clock of 2.60 GHz and a boost clock of 4.50 GHz, while the 9535 runs at 2.40 GHz base and 4.30 GHz boost. The 9655's higher clocks contribute to its 3.3% single-thread advantage.

TDP is a significant differentiator. The 9655 is rated at 400 W, while the 9535 draws 300 W. This 100 W gap has real implications for server power budgets and cooling requirements — the 9655 needs substantially more thermal headroom.

Cache hierarchies scale with core count. The 9655 has 384 MB of shared L3 cache across its 12 CCDs, while the 9535 has 256 MB across 8 CCDs. Both share the same per-core L1 (80 KB) and L2 (1 MB) allocations.

Transistor counts reflect the silicon differences: the 9655 has 99,780 million transistors, versus 66,520 million for the 9535. The die size is 12x 70.6 mm² for the 9655 and 8x 70.6 mm² for the 9535, showing that the 9655 uses 50% more compute dies.

Both parts share identical platform specs: DDR5 memory support, twelve-channel memory bus, 576.0 GB/s memory bandwidth, ECC memory, PCIe Gen 5 with 128 lanes (CPU only), and AMD Socket SP5. Both have no integrated graphics, are not multiplier-unlocked, and are active production parts with a launch date of 2024-10-09. The launch MSRP is $8,992 for the 9535 and $11,852 for the 9655.

Architecture Differences

Both processors belong to the AMD EPYC 9005 series and share the Zen 5 architecture with the Turin codename. They are fabricated on the same 4 nm TSMC process, meaning the fundamental microarchitecture is identical — same IPC, same instruction set, same memory controller design.

The architectural difference is purely one of scale. The 9655 uses 12 compute dies, each 70.6 mm², while the 9535 uses 8 of the same dies. This is why the 9655 has 96 cores versus 64 — it simply has more of the same silicon. The L3 cache scales accordingly: 384 MB shared across 12 dies versus 256 MB across 8, both maintaining the same 32 MB per die allocation.

The transistor count difference (99,780 million vs 66,520 million) is a direct consequence of the extra dies. The 9655 does not have any architectural enhancements over the 9535 — no additional cache per core, no different memory support, no extra PCIe lanes. It is a wider implementation of the same Zen 5 design.

The 9655's higher base and boost clocks indicate better binning or more aggressive power delivery, enabled by its 400 W TDP. The 9535, at 300 W, is binned for lower power consumption. Both support the same DDR5 twelve-channel memory configuration with identical bandwidth, so memory-bound workloads will see similar per-core performance, but the 9655's additional cores can drive more concurrent memory requests.

The only other architectural distinction is the part number: the 9535 is 100-000001147, and the 9655 is 100-000000674. Both are server/workstation market segments with active production status. No 3D V-Cache is present on either part, and both lack integrated graphics entirely.