AMD EPYC 9335 vs AMD EPYC 9684X Comparison

Head-to-Head Benchmarks

The benchmark data delivers a decisive, one-sided verdict: the AMD EPYC 9684X wins all 11 recorded head-to-head comparisons against the AMD EPYC 9335. The margin of victory varies dramatically by workload, from a narrow 5.8% single-thread edge to a staggering 1195.9% blowout in the physics test.

Starting with the largest gap, the PassMark physics test shows the 9684X scoring 24,686 versus the 9335's 1,905. That 1195.9% delta is far and away the biggest difference in the entire comparison. It reflects the 9684X's massive thread count and cache resources being leveraged in a test that scales aggressively with parallel execution. The 9335 is not competitive in this specific metric; the data suggests the 9684X is in a completely different performance class for physics simulation workloads.

The prime number finding test follows a similar pattern, though slightly less extreme. The 9684X posts 2,020 versus the 9335's 340, a 494.1% advantage. This workload is highly sensitive to integer throughput and memory latency, and the 9684X's 96 cores simply overwhelm the 9335's 32 cores. The 9335's score of 340 is among its weakest showings relative to its rival.

Moving to data encryption, the 9684X scores 178,453 against the 9335's 63,159, a 182.5% lead. Encryption workloads often benefit from both core count and specialized instruction support, and the 9684X dominates on both fronts. Similarly, random string sorting shows the 9684X at 349,126 versus 116,608, a 199.4% advantage. Sorting is memory-bandwidth sensitive, and while the 9335 actually has a higher rated memory bandwidth (576.0 GB/s versus 460.8 GB/s), the 9684X's larger cache and higher core count compensate entirely.

The integer math test gives the 9684X an 844,145 score against 346,291, a 143.8% win. Floating point math shows 472,174 versus 228,123, a 107% lead. Data compression results put the 9684X at 2,698,807 versus 1,203,096, a 124.3% margin. Extended instructions see the 9684X at 177,956 versus 105,706, a 68.3% advantage. The multithread benchmark shows 121,595 versus 65,811, an 84.8% lead. These are all substantial wins, ranging from roughly two-thirds to nearly triple the rival's score.

The closest contest is single-thread performance. The 9684X scores 2,891 and the 9335 scores 2,732, a 5.8% difference. While the 9335 has a higher boost clock (4.40 GHz versus 3.70 GHz), the 9684X still edges ahead. This narrow margin indicates that for purely single-threaded tasks, the two processors are nearly equivalent, with the 9684X holding a slight, albeit consistent, advantage.

The overall average benchmark score in the database places the 9684X at 266,914 and the 9335 at 194,228. The 9684X sits in the 99th percentile of all CPUs, as does the 9335, but the raw average gap is substantial. The nearest rivals for the 9684X include the AMD Ryzen Threadripper 9970X at 279,778 (4.6% higher), the Intel Xeon 6780E at 280,438 (4.8% higher), and the Intel Xeon 6980P at 251,516 (6.1% lower). The 9335's nearest rivals are the Intel Xeon 6741P at 194,901 (0.3% lower), the Intel Xeon 678X at 193,477 (0.4% higher), and the Intel Xeon 6740E at 187,718 (3.5% higher). This places both chips among the fastest server processors recorded, but the 9684X occupies a higher absolute tier.

FAQ

Q: Which processor wins the most head-to-head benchmarks?

A: The AMD EPYC 9684X wins all 11 recorded head-to-head comparisons against the AMD EPYC 9335. The 9335 records zero wins in the database.

Q: What is the largest performance gap between the two?

A: The largest gap is in the PassMark physics test, where the 9684X scores 24,686 against the 9335's 1,905, a 1195.9% delta.

Q: Is the 9335 competitive in any workload?

A: The closest result is single-thread performance, where the 9684X scores 2,891 and the 9335 scores 2,732, a 5.8% difference. The 9335 is within striking distance there but still loses.

Q: How do the two processors compare to their nearest rivals?

A: The 9684X's nearest rival, the AMD Ryzen Threadripper 9970X, scores 279,778, which is 4.6% higher. The Intel Xeon 6980P trails the 9684X by 6.1%. The 9335's nearest rival, the Intel Xeon 6741P, is 0.3% lower, while the Intel Xeon 678X is 0.4% higher.

Q: What is the average benchmark score for each?

A: The 9684X has an average benchmark score of 266,914, while the 9335 has an average of 194,228. Both are in the 99th percentile of all CPUs.

Q: Does the 9335's higher memory bandwidth help it win any test?

A: No. The 9335 has a rated memory bandwidth of 576.0 GB/s versus the 9684X's 460.8 GB/s, but the 9684X still wins every recorded benchmark, including memory-sensitive ones like random string sorting.

The Verdict

The data is unambiguous: the AMD EPYC 9684X is the superior processor across every measured workload. It wins all 11 head-to-head tests, with deltas ranging from 5.8% to 1195.9%. The 9684X's average benchmark score of 266,914 is 37.4% higher than the 9335's 194,228. For any workload that benefits from high core counts, large caches, or heavy parallel execution, the 9684X is the clear choice.

The 9335 does have advantages in the specification sheet: a higher base clock (3.00 GHz versus 2.55 GHz), a higher boost clock (4.40 GHz versus 3.70 GHz), a newer architecture (Zen 5 versus Zen 4), and a smaller process node (4 nm versus 5 nm). Yet these advantages do not translate into a single benchmark win. The 9335's 32 cores and 128 MB of L3 cache are simply outclassed by the 9684X's 96 cores and 1152 MB of L3 cache.

Buyers should choose the 9684X when maximum throughput is the priority, particularly for multi-threaded server workloads such as virtualization, database processing, or scientific computing. The 9335 is a lower-power alternative (210 W TDP versus 400 W), but the recorded data shows its performance is dramatically lower in every category. The 9335's only near-competitive result is single-thread performance, where it trails by just 5.8%. For workloads that are purely single-threaded, the two are nearly interchangeable, but the 9684X still wins.

From a competitive positioning standpoint, the 9684X's nearest rival is the AMD Ryzen Threadripper 9970X, which is 4.6% faster on average, while the Intel Xeon 6980P is 6.1% slower. The 9335's nearest rival is the Intel Xeon 6741P, which is 0.3% slower. This places both chips in the top tier of server processors, but the 9684X is in a higher performance bracket entirely.

Specification Differences

The two processors differ across nearly every core specification. The 9684X has 96 cores and 192 threads, while the 9335 has 32 cores and 64 threads. The 9335 has a higher base clock at 3.00 GHz versus 2.55 GHz, and a higher boost clock at 4.40 GHz versus 3.70 GHz.

Thermal design power differs substantially: the 9684X is rated at 400 W, while the 9335 is rated at 210 W. Both use the AMD Socket SP5, but the 9684X is from the EPYC 9004 series, while the 9335 is from the EPYC 9005 series. The launch MSRP for the 9684X is $14756, and the launch MSRP for the 9335 is $3178.

Memory bandwidth favors the 9335 at 576.0 GB/s versus the 9684X's 460.8 GB/s, though both use twelve-channel DDR5 with ECC support. The 9335 is built on a 4 nm process with 33,260 million transistors across 4x 70.6 mm² dies. The 9684X uses a 5 nm process with 135,240 million transistors across 12x 72 mm² dies. Both support PCIe Gen 5 with 128 lanes (CPU only).

Architecture Differences

The 9684X uses the Zen 4 architecture under the codename "Genoa-X" and belongs to the EPYC 9004 series generation. The 9335 uses the Zen 5 architecture under the codename "Turin" and belongs to the EPYC 9005 series generation. Both are manufactured by TSMC, but on different process nodes: 5 nm for the 9684X and 4 nm for the 9335.

Cache configurations differ markedly. The 9684X has 64 KB of L1 cache per core, 1 MB of L2 cache per core, and a massive 1152 MB of shared L3 cache. The 9335 has 80 KB of L1 per core, 1 MB of L2 per core, and 128 MB of shared L3 cache. The 9684X's L3 cache is nine times larger, a decisive factor in its benchmark dominance.

The transistor counts reflect the scale difference. The 9684X packs 135,240 million transistors across 12 dies, while the 9335 uses 33,260 million across 4 dies. The 9684X's die size is 12x 72 mm², while the 9335's is 4x 70.6 mm². The 9335's integrated graphics are listed as "N/A", and the 9684X has no integrated graphics listed either. Both have locked multipliers, and both are active production server/workstation parts. The 9684X was released on 2023-06-12, while the 9335 was released on 2024-10-09.

Where Each One Wins

The 9684X wins everywhere in the recorded data. It dominates the physics test with a 1195.9% lead, prime number finding with 494.1%, and random string sorting with 199.4%. It also leads in data encryption (182.5%), integer math (143.8%), data compression (124.3%), floating point math (107%), multithread (84.8%), extended instructions (68.3%), and single-thread (5.8%). There is no workload category in the database where the 9335 comes out ahead.

For use-case selection, the 9684X is the choice for workloads that scale with core count and cache capacity. This includes heavy virtualization environments, large-scale database workloads, scientific simulation, and any batch processing where parallel throughput is paramount. The 9684X's 96 cores and 1152 MB of L3 cache make it particularly suited for in-memory computing and data-intensive tasks where the working set can reside in the cache hierarchy.

The 9335, despite losing every benchmark, still has a role. Its lower TDP of 210 W versus 400 W makes it a more power-conscious option for density-constrained deployments. Its higher clocks (3.00 GHz base, 4.40 GHz boost) and newer Zen 5 architecture suggest it could be a better fit for lightly threaded workloads where per-core performance matters more, though the data shows the 9684X still wins even single-thread tests by 5.8%. The 9335's smaller die footprint (4x 70.6 mm² versus 12x 72 mm²) and lower transistor count (33,260 million versus 135,240 million) also indicate a simpler, potentially more available part.

For buyers, the decision is straightforward based on the data: if raw performance is the goal, the 9684X is the definitive pick. If power efficiency or architectural recency is a priority, the 9335 offers those traits, but at a substantial performance cost. The record shows no benchmark where the 9335's advantages translate into a win.