AMD EPYC 9535 vs Intel Xeon 696X Comparison

Head-to-Head Benchmarks

The benchmark comparison between the AMD EPYC 9535 and the Intel Xeon 696X is strikingly one-sided in aggregate. Across the eleven recorded PassMark tests, the AMD EPYC 9535 claims nine wins, while the Intel Xeon 696X takes only two. The magnitude of those victories, however, varies dramatically by workload.

The largest margin belongs to the AMD EPYC 9535 in random string sorting, where it posts a score of 247506 against the Intel part's 180392, a delta of 37.2%. This is the kind of workload that stresses memory subsystem latency and cache hierarchy efficiency, and the AMD part's advantage here is substantial. Integer math tells a similar story: the EPYC 9535 scores 730281 versus 572072 for the Xeon 696X, a 27.7% lead. This is the second-largest delta in the entire comparison and indicates a significant throughput advantage in general-purpose compute tasks that rely on integer operations.

Data encryption also favors the AMD processor decisively. The EPYC 9535 records 127372, while the Xeon 696X manages 112529, a 13.2% difference. This result suggests the AMD part's cryptographic instruction handling is more efficient in this PassMark workload. Physics simulation shows a comparable gap: 3834 for the EPYC 9535 versus 3382 for the Xeon 696X, a 13.4% edge for AMD.

The multithread benchmark, which often serves as a broad proxy for overall server throughput, gives the AMD EPYC 9535 a 9.1% advantage, with scores of 114528 and 104974 respectively. This is a meaningful margin for a workload that exercises all 64 cores and 128 threads on both processors.

The remaining AMD victories are narrower but consistent. In find prime numbers, the EPYC 9535 scores 874 against 853, a 2.5% lead. Data compression shows 2308822 versus 2264907, a 1.9% edge. Extended instructions produce 175784 versus 172975, a 1.6% difference. Floating point math rounds out the AMD wins at 457047 versus 450164, a 1.5% margin.

The Intel Xeon 696X takes the single-thread and singlethread benchmarks, both recording 3742 against the EPYC 9535's 3720. The delta is just 0.6%, which is a narrow lead but a consistent one across both entries. This indicates that in lightly threaded scenarios, the Intel part has a slight edge in per-core performance, likely tied to its higher boost clock.

Architecture Differences

The two processors come from fundamentally different design lineages. The AMD EPYC 9535 is built on the Zen 5 architecture, codenamed Turin, and belongs to the EPYC 9005 series. It is fabricated on a 4 nm process at TSMC, using a chiplet design that spans eight dies of 70.6 mm² each. The total transistor count is recorded at 66,520 million. The Intel Xeon 696X, by contrast, uses the Granite Rapids architecture, part of the Xeon 600 series. It is built on Intel's 5 nm process and uses two dies of 598 mm² each, for a combined die area of 1,196 mm². The Intel part's transistor count is not recorded in the database.

Cache organization differs substantially between the two. The AMD EPYC 9535 allocates 80 KB of L1 cache per core and 1 MB of L2 cache per core, with a shared L3 cache of 256 MB. The Intel Xeon 696X provides 112 KB of L1 per core and 2 MB of L2 per core, with a shared L3 cache of 336 MB. The Intel part therefore has a larger total cache footprint, but the AMD part's L3 is organized within its chiplet architecture, which may explain its strong showing in latency-sensitive workloads like random string sorting.

Memory channels also differ. The AMD EPYC 9535 supports twelve-channel DDR5 memory with a recorded bandwidth of 576.0 GB/s. The Intel Xeon 696X uses eight-channel DDR5 with a bandwidth of 409.6 GB/s. This 166.4 GB/s gap in theoretical memory bandwidth is a plausible contributor to the AMD part's wins in data compression and random string sorting, both of which are memory-intensive. Both processors support ECC memory and both provide PCIe Gen 5 with 128 lanes from the CPU.

The sockets are incompatible: the AMD EPYC 9535 uses AMD Socket SP5, while the Intel Xeon 696X uses Intel Socket 4710. The AMD part is not multiplier unlocked, while the Intel part is. The AMD processor was released on 2024-10-09, and the Intel processor has a release date of 2026-02-01, placing them roughly sixteen months apart in the database's timeline.

Where Each One Wins

The workload split is clear when the benchmark results are grouped by category. The AMD EPYC 9535 dominates in multithreaded throughput. Its 9.1% lead in the multithread benchmark, combined with a 27.7% lead in integer math and a 13.2% lead in encryption, makes it the stronger choice for database workloads, virtualized environments, and any server application that keeps all cores busy. The 37.2% advantage in random string sorting further reinforces this: sorting is a common operation in data processing pipelines, and a lead of that size could translate into visibly shorter batch job times.

The Intel Xeon 696X wins only in single-threaded performance, and even then by a slim 0.6% margin. This makes it marginally better suited for workloads that are latency-bound on a single core, such as certain legacy applications, some database query serialization points, or lightly threaded administrative tasks. But the margin is small enough that it may not be perceptible in real-world use. In floating point math, the AMD part leads by 1.5%, which is modest but still a win for the EPYC 9535. In physics simulation, the AMD part leads by 13.4%, which is a substantial margin for a workload that often blends integer and floating point operations.

For data compression, the AMD EPYC 9535's 1.9% lead is narrow, but in a workload that runs continuously, even small percentage differences compound over time. The extended instructions benchmark shows a 1.6% AMD lead, indicating that the Zen 5 instruction set implementation is at least on par with Granite Rapids for vectorized and specialized instruction streams.

Specification Differences

The two processors share several core specifications: both have 64 cores, 128 threads, a base clock of 2.40 GHz, support DDR5 memory with ECC, and offer PCIe Gen 5 with 128 lanes. Neither has integrated graphics. The differences are as follows.

The AMD EPYC 9535 boosts to 4.30 GHz, while the Intel Xeon 696X boosts to 4.80 GHz, a 0.50 GHz advantage for Intel. The AMD part has a TDP of 300 W, while the Intel part has a TDP of 350 W, a 50 W difference. The AMD part uses a 4 nm process from TSMC, while the Intel part uses a 5 nm process from Intel. The AMD part uses eight chiplets of 70.6 mm² each, while the Intel part uses two dies of 598 mm² each. The AMD part's L1 cache is 80 KB per core versus 112 KB per core for Intel; L2 is 1 MB per core versus 2 MB per core; L3 is 256 MB shared versus 336 MB shared. Memory bandwidth is 576.0 GB/s for AMD versus 409.6 GB/s for Intel. The AMD part uses twelve memory channels; the Intel part uses eight. The socket is SP5 for AMD and Socket 4710 for Intel. The AMD part is not multiplier unlocked; the Intel part is. The AMD part's part number is 100-000001147, while the Intel part's is SRWQ7. The AMD part's launch MSRP is $8992, and the Intel part's launch MSRP is $5599.

FAQ

Q: Which processor has the higher boost clock?

A: The Intel Xeon 696X boosts to 4.80 GHz, while the AMD EPYC 9535 boosts to 4.30 GHz.

Q: How much memory bandwidth does each processor support?

A: The AMD EPYC 9535 supports 576.0 GB/s over twelve DDR5 channels. The Intel Xeon 696X supports 409.6 GB/s over eight DDR5 channels.

Q: Which processor wins the PassMark multithread benchmark?

A: The AMD EPYC 9535 wins with a score of 114528, which is 9.1% ahead of the Intel Xeon 696X's 104974.

Q: Is there any benchmark where the Intel Xeon 696X beats the AMD EPYC 9535?

A: Yes. The Intel Xeon 696X wins the single-thread benchmark with 3742 against 3720 for the AMD part, a 0.6% margin. It wins the same result in the singlethread entry.

Q: What is the L3 cache size on each processor?

A: The AMD EPYC 9535 has 256 MB of shared L3 cache. The Intel Xeon 696X has 336 MB of shared L3 cache.

Q: What are the launch MSRP values?

A: The AMD EPYC 9535 has a launch MSRP of $8992. The Intel Xeon 696X has a launch MSRP of $5599.

The Verdict

The data points to the AMD EPYC 9535 as the stronger processor for the majority of server and workstation workloads. It wins nine of eleven benchmarks, with the largest margins in random string sorting (37.2%), integer math (27.7%), physics (13.4%), and encryption (13.2%). Its twelve-channel memory interface and 576.0 GB/s bandwidth give it a clear advantage in memory-bound tasks, and its 64 cores and 128 threads are on par with the Intel part in core count but deliver higher throughput in multithreaded tests. The 9.1% multithread lead is the kind of margin that shows up in real-world batch processing, virtualization density, and data analytics.

The Intel Xeon 696X is the better choice only for workloads that are strictly single-threaded, where its 0.6% lead in single-thread performance and its higher 4.80 GHz boost clock give it a slight edge. Its larger L3 cache (336 MB versus 256 MB) and larger per-core L1 and L2 caches are notable, but the benchmark data does not show these translating into wins outside of the single-thread tests. The Intel part consumes 50 W more TDP, which may factor into power-constrained deployments.

For buyers prioritizing raw throughput across a broad range of server tasks, the AMD EPYC 9535 is the recorded performance leader. For those running predominantly single-threaded workloads and willing to trade overall throughput for a marginal per-core advantage, the Intel Xeon 696X has the edge, but the data shows that edge is thin. The AMD part's nine wins against two, including the large deltas in integer math and random string sorting, make it the default recommendation based on the benchmark evidence.