AMD EPYC 9535 vs AMD EPYC 9734 Comparison

Head-to-Head Benchmarks

The recorded data shows a clear split between these two server processors. The AMD EPYC 9734 takes seven of the eleven head-to-head wins, while the AMD EPYC 9535 takes four. But the margins tell a more interesting story than the raw win count.

The EPYC 9734 dominates in throughput-oriented workloads. In data compression, it scores 2,900,008 versus 2,308,822 for the EPYC 9535, a 20.4% advantage. Data encryption shows an even larger gap: 179,390 versus 127,372, which is 29% ahead. Random string sorting goes to the 9734 by 30.8% (357,638 versus 247,506). Floating point math favors the 9734 at 549,045 versus 457,047, a 16.8% lead. Integer math follows the same pattern: 823,150 versus 730,281, or 11.3% ahead. Extended instructions also go to the 9734 with 205,925 versus 175,784, a 14.6% margin.

The most dramatic difference appears in physics. The EPYC 9734 scores 6,747, which is 43.2% higher than the EPYC 9535's 3,834. That is the single largest percentage gap in either direction across all recorded tests.

The EPYC 9535 fights back in single-thread performance. Its score of 3,720 in the single-thread test is 61% higher than the 9734's 2,310. That is the largest winning margin for either chip. The 9535 also wins the multithread test with 114,528 versus 102,286, a 12% edge. Prime number finding goes to the 9535 by a narrower 5.4% margin (874 versus 829).

Looking at aggregate results, the EPYC 9535 posts an average benchmark score of 379,408, while the EPYC 9734 averages 310,619. That puts the 9535 roughly 22% higher on average, despite losing more individual tests. The explanation lies in the weighting of the tests and the fact that the 9535's wins include the heavily weighted single-thread and multithread metrics.

Percentile rankings also separate them. The EPYC 9535 sits at the 100th percentile against all CPUs in the database, while the EPYC 9734 sits at the 99th. Both are exceptional, but the 9535 edges out every other recorded processor in aggregate standing.

FAQ

Q: Which processor has the higher single-thread score?

A: The AMD EPYC 9535 scores 3,720 in the single-thread test, which is 61% higher than the AMD EPYC 9734's 2,310.

Q: Which processor wins in data encryption?

A: The AMD EPYC 9734 scores 179,390 in data encryption, beating the EPYC 9535's 127,372 by 29%.

Q: How do the two compare in the multithread test?

A: The AMD EPYC 9535 wins with 114,528 versus 102,286, a 12% advantage for the 9535.

Q: Which chip has more cores?

A: The AMD EPYC 9734 has 112 cores and 224 threads, while the AMD EPYC 9535 has 64 cores and 128 threads.

Q: What is the difference in average benchmark scores?

A: The EPYC 9535 averages 379,408, while the EPYC 9734 averages 310,619. The 9535 is about 22% higher in average score.

Q: Which processor has the higher boost clock?

A: The AMD EPYC 9535 boosts to 4.30 GHz, while the AMD EPYC 9734 boosts to 3.00 GHz.

Where Each One Wins

The AMD EPYC 9734 is the clear choice for workloads that scale with core count and memory bandwidth. Its wins in data compression, data encryption, random string sorting, floating point math, integer math, extended instructions, and physics all point to a processor optimized for bulk parallel throughput. If the workload involves processing large datasets, encrypting streams, or running physics simulations that can use every available core, the 9734's 112 cores and 224 threads give it a decisive edge. The 43.2% lead in physics is particularly notable for simulation-heavy environments.

The AMD EPYC 9535 wins where per-thread performance matters. The 61% single-thread advantage makes it the better fit for database queries, latency-sensitive services, or any application that cannot fully utilize 112 cores. Its 12% win in the multithread test is curious, given the 9734's core advantage, but it suggests the 9535's higher clock speeds (4.30 GHz boost versus 3.00 GHz) compensate for the core deficit in certain parallel workloads. The 5.4% edge in prime number finding reinforces this: the 9535's architecture handles certain algorithmic patterns more efficiently.

For mixed workloads, the average benchmark score of 379,408 for the 9535 versus 310,619 for the 9734 suggests the 9535 offers better all-around performance when tasks vary across the day. But for dedicated, high-density compute farms where every job is embarrassingly parallel, the 9734's individual test wins carry more weight.

Specification Differences

The core count difference is substantial: 64 cores and 128 threads for the EPYC 9535 versus 112 cores and 224 threads for the EPYC 9734. Clock speeds also differ sharply. The 9535 runs at a 2.40 GHz base clock and boosts to 4.30 GHz. The 9734 runs at 2.20 GHz base and boosts to only 3.00 GHz.

Thermal design power differs by 40 watts: the 9535 is rated at 300 TDP, while the 9734 is rated at 340 TDP. Both use the same AMD Socket SP5 and support twelve-channel DDR5 memory, but memory bandwidth is not equal. The 9535 delivers 576.0 GB/s, while the 9734 delivers 460.8 GB/s. That bandwidth gap explains some of the 9535's wins in memory-sensitive workloads.

Cache configurations differ at the L1 level. The 9535 has 80 KB of L1 per core, while the 9734 has 64 KB per core. Both have 1 MB of L2 per core and 256 MB of shared L3 cache. The 9734's higher core count means more total L2 aggregate, but per-core capacity is identical.

Both processors support PCIe Gen 5 with 128 lanes (CPU only), and both support ECC memory. Neither has integrated graphics. The 9535 was released on 2024-10-09, while the 9734 was released on 2023-06-12. The 9734 carries a launch MSRP of $9600, while the 9535 carries a launch MSRP of $8992.

Architecture Differences

The EPYC 9535 uses the Zen 5 architecture under the Turin codename, part of the EPYC 9005 series. The EPYC 9734 uses the Zen 4c architecture under the Bergamo codename, part of the EPYC 9004 series. Both are fabricated by TSMC, but on different nodes: the 9535 uses a 4 nm process, while the 9734 uses a 5 nm process.

Transistor counts are close, but not identical. The 9535 packs 66,520 million transistors across eight chiplets of 70.6 mm² each. The 9734 packs 71,000 million transistors across eight chiplets of 73 mm² each. The newer node on the 9535 delivers higher clock speeds and better efficiency per watt, despite the slightly lower transistor count.

The L1 cache difference of 80 KB versus 64 KB per core reflects the architectural change between Zen 5 and Zen 4c. The Zen 4c design in the 9734 prioritizes core density over per-core resources, which is why it fits 112 cores into the same socket. The Zen 5 design in the 9535 prioritizes per-core performance, which is why it achieves a 61% higher single-thread score.

Memory bandwidth also differs due to the architecture. The 9535's 576.0 GB/s versus the 9734's 460.8 GB/s represents a 25% bandwidth advantage for the newer chip. This is not a feature of the memory controller alone; it reflects the overall memory subsystem design in Zen 5 versus Zen 4c.

Both processors are active in production and target the server/workstation segment. Neither has an unlocked multiplier, which is expected for EPYC parts. The 9734's Zen 4c is a dense-core variant optimized for cloud and scale-out workloads, while the 9535's Zen 5 is optimized for raw per-core speed and modern instruction efficiency.

The Verdict

The data points to a straightforward recommendation for each use case.

Choose the AMD EPYC 9535 if your workloads are sensitive to single-thread performance, require the highest possible memory bandwidth, or involve a mix of tasks where the aggregate benchmark score matters. The 61% single-thread advantage is not a minor detail; it is the largest recorded gap between these two chips. The 576.0 GB/s memory bandwidth versus 460.8 GB/s gives it a real edge for in-memory databases and analytics. The 100th percentile ranking against all CPUs confirms its standing as the top aggregate performer in the database.

Choose the AMD EPYC 9734 if your workloads are heavily parallel and can use 112 cores effectively. The wins in data compression, encryption, floating point, integer math, and physics are all substantial, ranging from 11.3% to 43.2%. For dedicated batch processing, rendering farms, or scientific computing where every job is independent, the core count advantage will translate directly into shorter runtimes. The 99th percentile ranking is still elite, and the 7 individual test wins out of 11 show where its strengths lie.

For a single server that must handle everything, the 9535's average benchmark score of 379,408 versus 310,619 makes it the safer default. For a fleet of servers dedicated to one parallel task, the 9734's targeted advantages are more valuable than its lower average score. The 340 TDP versus 300 TDP also means the 9734 requires more cooling headroom, which is a practical consideration for dense deployments.