AMD EPYC 9135 vs Intel Xeon 654 Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive overall win for the Intel Xeon 654, taking 16 of the 17 head-to-head benchmark comparisons against the AMD EPYC 9135. The only test where the AMD part comes out ahead is PassMark random string sorting, where the EPYC 9135 scores 90064 versus 82828 for the Intel Xeon 654, an 8% difference in favor of AMD.

The Intel Xeon 654's largest victories come in compute-heavy workloads. In PassMark find prime numbers, the Intel part scores 390 against 292 for the EPYC 9135, a 33.6% advantage. Floating point math shows a similar gap: 163093 for the Xeon 654 versus 126679 for the EPYC 9135, a 28.7% lead. Extended instructions favor Intel by 13.8%, with scores of 63539 and 55822 respectively. Data compression is also a clear Intel win at 10.8% (818902 versus 739277).

Across the Cinebench suite, the Intel Xeon 654 maintains a consistent edge. In Cinebench R15 multicore, the Xeon 654 scores 5256 against 4952 for the EPYC 9135, a 6.1% difference. Single-core R15 shows 742 versus 699, a 6.2% lead. Cinebench R20 multicore delivers 21903 versus 20637, again 6.1%. The single-core R20 result is 3092 against 2913, a 6.1% margin. Cinebench R23 multicore shows 52150 versus 49136, and single-core R23 shows 7362 versus 6936, both at 6.1%. This consistency across the Cinebench generations suggests a stable architectural advantage rather than a workload-specific quirk.

Smaller but still positive Intel leads appear in PassMark multithread (61353 versus 57170, 7.3%), PassMark single thread (3778 versus 3672, 2.9%), integer math (207745 versus 202962, 2.4%), and physics (5596 versus 5477, 2.2%). Data encryption is nearly a tie, with the Xeon 654 at 40675 and the EPYC 9135 at 40295, a 0.9% edge for Intel.

The average benchmark score for the Intel Xeon 654 is 90717, placing it in the 96th percentile among all CPUs. The EPYC 9135 averages 82980, also in the 96th percentile. The Intel part's nearest rivals include the AMD Ryzen AI Max+ 392 with an average score of 90541 (0.2% behind) and the AMD Ryzen 9 9955HX at 91199 (0.5% ahead). The EPYC 9135's closest competitor is the AMD EPYC 4584PX at 83090 (0.1% behind) and the Intel Core Ultra 9 285K at 83807 (1% ahead). The 90717 average for the Xeon 654 is roughly 9.3% higher than the 82980 average for the EPYC 9135, which aligns with the per-test deltas observed in the head-to-head data.

Where Each One Wins

Based on the benchmark results, the Intel Xeon 654 is the stronger choice for floating-point-heavy computation, prime number searches, and extended instruction workloads. Its 28.7% lead in floating point math and 33.6% lead in find prime numbers indicate a substantial advantage in scientific computing and numerical simulation tasks. The 13.8% edge in extended instructions suggests better support for advanced vector and SIMD workloads. Data compression also falls firmly in Intel's camp with a 10.8% advantage, which matters for database compression, file archiving, and data transfer pipelines.

The Intel Xeon 654 also wins all Cinebench tests by a consistent 6.1% to 6.2% margin, making it the preferable part for 3D rendering and content creation workloads where Cinebench scores correlate with real-world performance. The multithread score advantage of 7.3% reinforces this pattern for heavily threaded applications.

The AMD EPYC 9135 wins only random string sorting, with an 8% advantage (90064 versus 82828). This single win indicates a specific strength in sorting algorithms and possibly workloads that benefit from the EPYC 9135's higher base clock of 3.65 GHz versus 3.10 GHz for the Intel part. However, the EPYC 9135 does not translate this into broader victories elsewhere.

The EPYC 9135's boost clock is lower at 4.30 GHz versus 4.80 GHz for the Xeon 654, which may explain why the Intel part wins single-thread tests by 2.9% to 6.2% despite the AMD part's higher base frequency. For workloads that are primarily single-threaded and latency-sensitive, the Intel Xeon 654's higher boost clock appears to provide the advantage.

The Intel Xeon 654 has 18 cores and 36 threads versus 16 cores and 32 threads for the EPYC 9135, giving Intel a 12.5% core count advantage. This translates into the observed multithread performance gap of 7.3%, though the per-core efficiency of the AMD Zen 5 architecture narrows the difference in some tests. In integer math, the 2.4% Intel lead is much smaller than the core count gap would suggest, indicating that the EPYC 9135's Zen 5 cores are more efficient per core for integer operations.

FAQ

Q: Which CPU wins the majority of benchmark comparisons?

A: The Intel Xeon 654 wins 16 of the 17 head-to-head benchmarks, with the AMD EPYC 9135 winning only PassMark random string sorting.

Q: What is the largest performance gap between the two in any single test?

A: The largest gap is in PassMark find prime numbers, where the Intel Xeon 654 scores 390 versus 292 for the EPYC 9135, a 33.6% advantage. Floating point math is close behind at 28.7% in favor of Intel.

Q: How do the two compare in Cinebench R23 multicore performance?

A: The Intel Xeon 654 scores 52150 in Cinebench R23 multicore, while the AMD EPYC 9135 scores 49136, giving Intel a 6.1% lead.

Q: Does the AMD EPYC 9135 win any benchmark at all?

A: Yes, the EPYC 9135 wins PassMark random string sorting with a score of 90064, which is 8% higher than the Intel Xeon 654's 82828.

Q: What are the average benchmark scores for each CPU?

A: The Intel Xeon 654 has an average benchmark score of 90717, while the AMD EPYC 9135 averages 82980. Both rank in the 96th percentile among all CPUs.

Q: Which CPU has the higher single-thread performance?

A: The Intel Xeon 654 leads in single-thread tests, scoring 3778 in PassMark single thread versus 3672 for the EPYC 9135, a 2.9% advantage. In Cinebench R23 single-core, Intel scores 7362 versus 6936, a 6.1% lead.

Specification Differences

The two processors differ in several core specifications. The Intel Xeon 654 has 18 cores and 36 threads, while the AMD EPYC 9135 has 16 cores and 32 threads. Base clock speeds are 3.10 GHz for Intel and 3.65 GHz for AMD. Boost clocks are 4.80 GHz for Intel and 4.30 GHz for AMD. Both have a TDP of 200.

The Intel Xeon 654 uses Intel Socket 4710, while the EPYC 9135 uses AMD Socket SP5. The Intel part is built on a 5 nm process by Intel, while the EPYC 9135 uses a 4 nm process from TSMC. The Intel die size is listed as 2x 598 mm², whereas the AMD die is 2x 70.6 mm² and has a transistor count of 16,630 million.

Cache configurations differ notably. The Intel Xeon 654 has 112 KB of L1 cache per core, 2 MB of L2 per core, and 72 MB of shared L3. The EPYC 9135 has 80 KB of L1 per core, 1 MB of L2 per core, and 64 MB of shared L3. Both support DDR5 memory with ECC, but the memory channels differ: Intel uses eight-channel while AMD uses twelve-channel. Memory bandwidth is 409.6 GB/s for Intel and 576.0 GB/s for AMD.

Both have PCIe Gen 5 with 128 lanes (CPU only), and neither has integrated graphics. The market segment for both is Server/Workstation, and both are in active production. The Intel Xeon 654 has an unlocked multiplier, while the EPYC 9135 is locked. The Intel launch MSRP is $1199, and the AMD launch MSRP is $1214. The Intel part numbers include SA2DP, and the AMD part number is 100-000001150. The Intel Xeon 654 was released on 2026-02-01, while the EPYC 9135 was released on 2024-10-09.

Architecture Differences

The Intel Xeon 654 is built on the Granite Rapids architecture, specifically the Xeon 600 (Granite Rapids-WS) generation, using a 5 nm process from Intel. The AMD EPYC 9135 uses the Zen 5 architecture, codenamed Turin, as part of the EPYC 9005 series, built on a 4 nm process by TSMC. This process difference may contribute to the EPYC 9135's higher base clock of 3.65 GHz despite having fewer cores, as the smaller process node typically allows for better power efficiency per transistor.

The Intel part uses a dual-die design with a combined die size of 2x 598 mm², which is substantially larger than the AMD's 2x 70.6 mm² dies. The AMD EPYC 9135's transistor count is listed at 16,630 million, while no transistor count is provided for the Intel Xeon 654. The larger Intel die size may accommodate the bigger L3 cache of 72 MB versus 64 MB for AMD, as well as the higher per-core L2 of 2 MB versus 1 MB.

The memory architecture favors AMD with twelve-channel DDR5 support and 576.0 GB/s bandwidth, compared to Intel's eight-channel and 409.6 GB/s. This bandwidth advantage could benefit the EPYC 9135 in memory-intensive workloads, though the recorded benchmarks do not show AMD winning any memory-bound tests except random string sorting. The Intel Xeon 654 compensates with a higher boost clock of 4.80 GHz versus 4.30 GHz, which likely explains its single-thread benchmark wins.

Both processors provide PCIe Gen 5 with 128 lanes from the CPU, making them comparable for expansion and I/O-heavy server deployments. Neither includes integrated graphics, and both support ECC memory, aligning them for reliability-focused server environments. The Intel part's unlocked multiplier offers overclocking flexibility, whereas the EPYC 9135 is locked. The EPYC 9135's earlier release date and established production status contrast with the Intel Xeon 654's later release, but both are currently active in production.