AMD EPYC 9655 vs AMD EPYC 9734 Comparison

Head-to-Head Benchmarks

The benchmark data presents a decisive picture: the AMD EPYC 9655 wins 16 of the 17 recorded head-to-head tests, with the AMD EPYC 9734 taking a single narrow victory. The scale of the 9655's advantage is substantial across nearly every workload category.

The largest gap appears in PassMark physics, where the 9655 scores 25,947 against the 9734's 6,747, a delta of 284.6%. This is the most extreme difference in the entire comparison, indicating a fundamental advantage in simulation or physics-based workloads that goes far beyond what core count alone would suggest.

Cinebench results show a consistent pattern. Across R15, R20, and R23, both multicore and singlecore tests, the 9655 leads by 52.5% to 52.6%. For example, Cinebench R23 multicore shows the 9655 at 132,672 versus 86,943 for the 9734. The singlecore R23 result is similarly lopsided: 18,730 against 12,274. This consistency across rendering generations suggests the performance gap is architectural rather than workload-specific.

PassMark integer math shows a 38.4% advantage for the 9655, with scores of 1,139,161 versus 823,150. Floating point math follows at 20.7% (662,958 versus 549,045). The 9655 also leads in data encryption by 17.4% (210,555 versus 179,390) and data compression by 12.8% (3,271,896 versus 2,900,008).

The prime number search test reveals a 92.8% delta in favor of the 9655, with 1,598 versus 829. Single-thread performance shows a 66.5% gap, with scores of 3,847 versus 2,310. Random string sorting favors the 9655 by 22.9% (439,682 versus 357,638), and multithread performance shows a 52.6% delta (156,110 versus 102,286).

The only test where the 9734 wins is PassMark extended instructions, scoring 205,925 versus 203,285, a 1.3% margin. This is a narrow result and does little to offset the broader trend. The average benchmark score reflects the overall difference: the 9655 sits at 373,479, while the 9734 records 310,619, roughly 20% lower.

FAQ

Q: Which processor has the higher multicore performance?

A: The AMD EPYC 9655 leads in every multicore benchmark recorded. Cinebench R23 multicore shows 132,672 versus 86,943, a 52.6% advantage. PassMark multithread shows 156,110 versus 102,286, also a 52.6% delta.

Q: Does the EPYC 9734 win any benchmark at all?

A: Yes, one. The PassMark extended instructions test goes to the 9734 with 205,925 versus 203,285, a 1.3% margin. It is the only test out of 17 where the 9734 comes out ahead.

Q: How large is the single-thread performance gap?

A: The 9655 scores 3,847 in PassMark single-thread, compared to 2,310 for the 9734, a 66.5% advantage. Cinebench R23 singlecore shows a similar 52.6% gap: 18,730 versus 12,274.

Q: What explains the massive physics benchmark difference?

A: The PassMark physics test shows the 9655 at 25,947 versus 6,747 for the 9734, a 284.6% delta. This is far larger than the gaps in any other benchmark and suggests a significant architectural advantage in this specific workload.

Q: How do the two processors compare in memory-intensive tasks?

A: Data compression favors the 9655 at 3,271,896 versus 2,900,008, a 12.8% lead. Random string sorting also goes to the 9655, 439,682 versus 357,638, a 22.9% lead. Data encryption shows a 17.4% advantage for the 9655.

Q: Which processor has the higher average benchmark score?

A: The 9655 records an average benchmark score of 373,479, while the 9734 records 310,619. The 9655 ranks at the 100th percentile among all CPUs, while the 9734 ranks at the 99th percentile.

Architecture Differences

The two processors come from different generations of AMD's EPYC lineup. The 9655 belongs to the EPYC 9005 series with a Zen 5 architecture under the Turin codename, built on a 4 nm process at TSMC. The 9734 belongs to the EPYC 9004 series with a Zen 4c architecture under the Bergamo codename, built on a 5 nm process, also at TSMC.

The 9734 has more cores, 112 versus 96, and more threads, 224 versus 192. Despite this, it loses nearly every benchmark. The 9655 compensates with a higher base clock of 2.60 GHz against 2.20 GHz, and a substantially higher boost clock of 4.50 GHz against 3.00 GHz. The clock speed difference is particularly pronounced at boost, a 1.50 GHz gap.

Cache configurations differ as well. The 9655 has 80 KB of L1 per core, 1 MB of L2 per core, and 384 MB of shared L3 cache. The 9734 has 64 KB of L1 per core, 1 MB of L2 per core, and 256 MB of shared L3 cache. The 9655 therefore offers 128 MB more shared L3 cache.

Transistor counts and die sizes also diverge. The 9655 contains 99,780 million transistors across 12 dies of 70.6 mm² each. The 9734 contains 71,000 million transistors across 8 dies of 73 mm² each. The 9655 uses more transistors distributed across more, smaller dies.

Memory bandwidth favors the 9655 at 576.0 GB/s versus 460.8 GB/s for the 9734, despite both using DDR5 with a twelve-channel memory bus. Both support ECC memory. Both use PCIe Gen 5 with 128 lanes CPU-only.

The 9655 has a TDP of 400, while the 9734 has a TDP of 340. Both use AMD Socket SP5. Neither has an unlocked multiplier. The 9655 was released on 2024-10-09, while the 9734 was released on 2023-06-12.

The Verdict

The data points to a clear conclusion: the AMD EPYC 9655 is the stronger processor in nearly every measurable way. Its 16 benchmark wins out of 17 tests, combined with an average score of 373,479 against 310,619, establishes it as the higher-performing part. The 9655 sits at the 100th percentile among all CPUs, while the 9734 sits at the 99th.

The 9734's advantage in core count, 112 versus 96, does not translate into performance wins. The 9655's higher clocks, larger cache, newer architecture, and greater memory bandwidth overcome the 16-core deficit. The physics benchmark delta of 284.6% is particularly telling, as it shows workloads where the 9655's architecture provides an outsized benefit.

The 9734 does hold a single win in extended instructions, but the 1.3% margin is minimal. For buyers choosing between these two, the 9655 is the choice for virtually all workloads represented in the database. The 9734 may still have a place in scenarios where its lower TDP of 340 is a consideration, but the recorded performance data does not favor it.

Specification Differences

The two processors differ across multiple specification fields. The 9655 has 96 cores and 192 threads, while the 9734 has 112 cores and 224 threads. The 9655 has a base clock of 2.60 GHz and a boost clock of 4.50 GHz; the 9734 has a base clock of 2.20 GHz and a boost clock of 3.00 GHz.

TDP differs: 400 for the 9655, 340 for the 9734. The architecture is Zen 5 for the 9655 and Zen 4c for the 9734. The process node is 4 nm for the 9655 and 5 nm for the 9734. Transistor counts are 99,780 million for the 9655 and 71,000 million for the 9734. Die sizes are 12x 70.6 mm² for the 9655 and 8x 73 mm² for the 9734.

L1 cache is 80 KB per core for the 9655 and 64 KB per core for the 9734. L2 cache is 1 MB per core for both. L3 cache is 384 MB for the 9655 and 256 MB for the 9734. Memory bandwidth is 576.0 GB/s for the 9655 and 460.8 GB/s for the 9734.

The 9655 has a launch MSRP of $11852; the 9734 has a launch MSRP of $9600. The 9655 was released on 2024-10-09, the 9734 on 2023-06-12. The 9655 has a part number of 100-000000674, the 9734 has 100-000001235.

Where Each One Wins

The 9655 wins across the majority of tested workloads. In rendering, Cinebench R15, R20, and R23 all show a 52.6% multicore advantage, making the 9655 the clear choice for CPU-based rendering tasks. Single-thread performance also favors the 9655 by 52.6% to 66.5% depending on the test, which matters for lightly threaded applications and responsiveness.

Integer math, floating point math, data encryption, data compression, and random string sorting all go to the 9655 with deltas ranging from 12.8% to 38.4%. These are common enterprise workloads, and the 9655 handles them with a consistent edge. The prime number search test shows a 92.8% lead, and the physics test shows a 284.6% lead, indicating particular strength in computational and simulation-heavy tasks.

The 9734 wins exactly one test: PassMark extended instructions, with a 1.3% margin. This suggests that for workloads that rely heavily on specialized instruction extensions, the 9734 offers a marginal advantage. It is not a large enough delta to drive a purchasing decision on its own, but it does represent a niche where the older processor holds a slight edge.

The 9734 also has a lower TDP of 340 versus 400, which may be a consideration in power-constrained deployments. However, the recorded data does not include power efficiency benchmarks, so any such advantage is inferred from the specification difference alone. For overall performance, the 9655 is the superior part across the benchmark suite.