AMD EPYC 9335 vs Intel Xeon 676X Comparison
AMD EPYC 9335
Xeon 676X
PERFORMANCE BENCHMARKS
Analysis: AMD EPYC 9335 vs Intel Xeon 676X
The AMD EPYC 9335 and the Intel Xeon 676X are two 32-core server processors aimed at the same workload class but built around very different design philosophies. The EPYC 9335, part of the EPYC 9005 series with the Turin codename and Zen 5 architecture, reaches the 99th percentile among all CPUs in the database with an average benchmark score of 194,228. The Xeon 676X, from the Granite Rapids generation, sits at the 98th percentile with an average score of 158,540. Despite that lower average, the Xeon wins 10 of the 11 recorded head-to-head benchmark comparisons, making this a matchup between a high overall average on one side and raw dominance in specific tests on the other.
FAQ
Q: Which processor is faster in single-threaded performance?
A: The Intel Xeon 676X scores 4015 in the PassMark single-thread test, while the AMD EPYC 9335 scores 2732. The Xeon leads by 32% in this test.
Q: Which processor has more L3 cache?
A: The Xeon 676X has 144 MB of shared L3 cache, whereas the EPYC 9335 has 128 MB. The Xeon also offers 2 MB of L2 per core against 1 MB, and 112 KB of L1 per core versus 80 KB.
Q: Which processor has the higher memory bandwidth?
A: The AMD EPYC 9335 uses a twelve-channel memory bus with 576.0 GB/s of bandwidth. The Intel Xeon 676X uses an eight-channel bus with 409.6 GB/s, making the EPYC the bandwidth leader by a substantial margin.
Q: Which processor draws less power?**
A: The EPYC 9335 is rated at 210 W TDP, while the Xeon 676X is rated at 275 W. The AMD part consumes less power under its rated thermal envelope.
Q: When did each processor launch?**
A: The AMD EPYC 9335 launched on 2024-10-09. The Intel Xeon 676X has a later release date of 2026-02-01.
Q: Which processor wins the PassMark multithread test?**
A: The Xeon 676X scores 91115 in PassMark multithread, versus 65811 for the EPYC 9335. That is a 27.8% advantage for Intel.
Architecture Differences
The two chips come from different foundries and design families. The AMD EPYC 9335 uses the Zen 5 architecture, codenamed Turin, fabricated at TSMC on a 4 nm process. It packs 33,260 million transistors across a 4x 70.6 mm² die. The Intel Xeon 676X uses the Granite Rapids architecture, fabricated by Intel on a 5 nm process, with a die size of 2x 598 mm² and no transistor count listed in the database.
Cache layout differs significantly. The EPYC 9335 allocates 80 KB of L1 and 1 MB of L2 per core, with 128 MB of L3 shared across the chip. The Xeon 676X gives each core 112 KB of L1 and 2 MB of L2, with 144 MB of L3 shared. The Xeon therefore has more cache at every level, which contributes to its high single-thread and physics results.
Memory architecture diverges as well. The EPYC 9335 runs DDR5 over a twelve-channel bus, delivering 576.0 GB/s of bandwidth. The Xeon 676X runs DDR5 over an eight-channel bus with 409.6 GB/s. Both support ECC memory and both use PCIe Gen 5 with 128 lanes available from the CPU. Neither has integrated graphics. The Xeon 676X has an unlocked multiplier, while the EPYC 9335 is locked.
The EPYC 9335 uses the AMD Socket SP5, while the Xeon 676X uses Intel Socket 4710. Production status for both is listed as Active. The EPYC part number is 100-000001149; the Xeon part number is SA2CY.
Head-to-Head Benchmarks
The head-to-head PassMark comparison covers 11 tests, and the Intel Xeon 676X wins 10 of them. The largest margin comes in the physics test, where the Xeon scores 8281 against the EPYC's 1905, a difference of 77%. That is the most lopsided result in the entire comparison, and it shows the Xeon's ability to handle simulation-style floating-point workloads.
The second-largest gap is in prime number computation. The Xeon scores 738 in the PassMark find-prime-numbers test, while the EPYC scores 340, a difference of 53.9%. This is a strong indicator of integer-heavy algorithmic work. The single-thread test also goes to the Xeon, 4015 versus 2732, a 32% gap that matters for lightly threaded workloads.
The multithread score follows the same pattern: 91115 for the Xeon versus 65811 for the EPYC, a 27.8% gap. Floating-point math also favors the Xeon, 283570 versus 228123, a 19.6% gap. Random string sorting goes to the Xeon at 137976 versus 116608, a 15.5% gap. Data compression shows the Xeon at 1355807 versus 1203096, an 11.3% gap. Data encryption favors the Xeon at 67638 versus 63159, a 6.6% gap. Integer math is closer, with the Xeon at 354777 versus 346291, a 2.4% gap.
The AMD EPYC 9335 takes only one test: extended instructions, scoring 105706 versus 105231, a narrow 0.5% win. That is the sole head-to-head victory for the AMD part.
These results indicate that the Xeon 676X is the stronger performer across nearly all measured PassMark workloads, with the biggest advantages in physics, prime-number search, and single-threaded execution. The EPYC's only lead is in the extended instruction set test, where it barely edges the Xeon.
Specification Differences
The following fields differ between the two processors:
| Field | AMD EPYC 9335 | Intel Xeon 676X |
|-------|----------------|-----------------|
| Base clock | 3.00 GHz | 2.80 GHz |
| Boost clock | 4.40 GHz | 4.90 GHz |
| TDP | 210 W | 275 W |
| Socket | AMD Socket SP5 | Intel Socket 4710 |
| Process node | 4 nm (TSMC) | 5 nm (Intel) |
| Transistors | 33,260 million | not listed |
| Die size | 4x 70.6 mm² | 2x 598 mm² |
| L1 per core | 80 KB | 112 KB |
| L2 per core | 1 MB | 2 MB |
| L3 shared | 128 MB | 144 MB |
| Memory bus | Twelve-channel | Eight-channel |
| Memory bandwidth | 576.0 GB/s | 409.6 GB/s |
| Release date | 2024-10-09 | 2026-02-01 |
| Launch MSRP | $3178 | $2499 |
| Multiplier | Locked | Unlocked |
The EPYC 9335 has a higher base clock, a lower TDP, a smaller process node, and more memory bandwidth. The Xeon 676X has a higher boost clock, larger caches at every level, a higher TDP, and an unlocked multiplier.
Where Each One Wins
The Intel Xeon 676X wins in every category that involves raw compute throughput. The physics test result of 8281 versus 1905 shows a large lead in simulation-style floating-point work. The prime number test at 738 versus 340 points to strong integer algorithmic throughput. The single-thread score of 4015 versus 2732 makes the Xeon the clear choice for latency-sensitive, low-thread-count applications. The multithread score of 91115 versus 65811 extends that advantage to fully threaded workloads.
The AMD EPYC 9335 wins in the extended instructions test with a 0.5% margin. It also offers a higher memory bandwidth of 576.0 GB/s versus 409.6 GB/s, which can matter for memory-bound code that does not saturate the compute cores. Its lower TDP of 210 W versus 275 W gives it a power advantage in dense server installations. The EPYC also has a higher average benchmark score overall, 194,228 versus 158,540, though that average reflects a broader set of tests than the 11 head-to-head comparisons.
The Verdict
For workloads that stress compute throughput, the Intel Xeon 676X is the clear choice. It wins 10 of the 11 head-to-head tests, with especially large margins in physics (77% ahead) and prime number search (53.9% ahead). The Xeon also leads in single-thread performance by 32%, which makes it suitable for latency-sensitive tasks that cannot spread across many cores.
For workloads that are memory-bound or power-constrained, the AMD EPYC 9335 has its own strengths. It offers 576.0 GB/s of memory bandwidth versus 409.6 GB/s on the Xeon, and it draws 210 W versus 275 W. It also has a higher overall average benchmark score, 194,228 versus 158,540. The EPYC's single head-to-head win in extended instructions is narrow but indicates that certain instruction-set-heavy workloads can run at parity or slightly ahead.
The decision comes down to the nature of the workload. The Intel Xeon 676X dominates the compute-oriented PassMark suite and is the safer pick for general compute, physics, prime, and single-threaded tasks. The AMD EPYC 9335 is the better option for memory-hungry environments and power-sensitive deployments, provided the workload does not rely on the specific tests where the Xeon holds a large lead.