AMD EPYC 9745 vs Intel Xeon 6780E Comparison

The AMD EPYC 9745 and Intel Xeon 6780E are both high-core-count server processors, but the benchmark data shows them in entirely different performance tiers. The EPYC 9745, built on Zen 5c, consistently outperforms the Xeon 6780E across every recorded workload. The following analysis breaks down the recorded measurements from the database, focusing on the magnitude of the performance differences and what they mean for server deployment.

Head-to-Head Benchmarks

The database records 16 head-to-head benchmark comparisons between the AMD EPYC 9745 and the Intel Xeon 6780E. The AMD part wins all 16. The margins are not small: the smallest advantage is an 18.9% lead in PassMark data encryption, while the largest is a 146% lead in PassMark extended instructions.

In Cinebench, the EPYC 9745 is uniformly dominant. In Cinebench R15 multicore, it scores 11198 against 7431 for the Xeon, a 50.7% advantage. The single-core R15 result is similarly lopsided: 1580 against 1049, again a 50.6% lead. Cinebench R20 multicore shows 46659 versus 30963, a 50.7% delta, and the single-core R20 test shows 6586 versus 4371, also 50.7%. The pattern holds in Cinebench R23 multicore, where the AMD chip scores 111093 compared to 73723 for Intel, exactly 50.7% ahead.

The PassMark suite confirms the trend with even larger deltas in some tests. The most striking result is PassMark extended instructions: 280477 for the EPYC 9745 versus 114008 for the Xeon 6780E. That is a 146% difference, meaning the AMD part more than doubles the Intel result in this specific workload. PassMark integer math shows a 90.8% lead for AMD: 1224315 versus 641817. Floating point math follows with a 75.5% margin: 761219 against 433862. PassMark physics shows 17122 versus 10951, a 56.4% edge.

Memory-sensitive workloads also favor the EPYC 9745. PassMark data compression scores 3929890 against 2557582, a 53.7% lead. Random string sorting shows 468975 versus 326954, a 43.4% advantage. PassMark find prime numbers records 979 versus 708, a 38.3% delta. PassMark multithread scores 130698 against 86734, a 50.7% margin.

The only workload where the two are relatively close is data encryption: 229447 versus 193004, with the EPYC 9745 leading by 18.9%. Even there, the AMD processor wins decisively. There is no recorded benchmark in the database where the Xeon 6780E outperforms the EPYC 9745.

Where Each One Wins

Given that the EPYC 9745 wins all 16 recorded benchmarks, the use-case split is one-sided. The AMD part is the clear choice for compute-intensive server workloads that rely on multi-threaded performance, single-thread performance, and memory bandwidth.

The EPYC 9745 wins in rendering workloads, as shown by Cinebench R15, R20, and R23 multicore scores. It wins in general-purpose integer and floating point math, broad instruction set execution, and physics simulations. It also wins in data compression and random string sorting, tasks that depend heavily on memory subsystem throughput. Even in encryption, where the margin is smallest, the EPYC 9745 still comes out ahead.

The Xeon 6780E has no recorded wins. Its only potential advantage is its higher core count of 144 versus 128, but the benchmark data shows that the EPYC 9745's 128 cores deliver more work per clock and per thread. The Xeon 6780E does have a lower TDP of 330 watts compared to 400 watts, which could matter in power-constrained environments, but the database has no power efficiency benchmarks to quantify that trade-off.

For workloads that are purely single-threaded, the EPYC 9745 is also superior. Its PassMark single-thread score of 2806 versus 1923 for the Xeon is a 45.9% margin. This means that even lightly threaded applications will run faster on the AMD part.

Architecture Differences

The two processors are built on fundamentally different designs. The AMD EPYC 9745 uses the Zen 5 architecture with the codename Turin, fabricated on a 3 nm process at TSMC. The Intel Xeon 6780E uses the Sierra Forest architecture, fabricated on Intel's 5 nm process. The process node difference likely contributes to the performance gap, as the 3 nm node allows for higher transistor density and better power efficiency per clock.

The EPYC 9745 has 128 cores and 256 threads, while the Xeon 6780E has 144 cores and 144 threads. The Xeon has no simultaneous multithreading, which explains why its thread count equals its core count. The AMD part has two threads per core. This difference is critical: the EPYC 9745 has 256 threads versus 144, giving it a 77.8% thread advantage despite having 16 fewer cores.

Cache layouts also differ significantly. The EPYC 9745 has 80 KB of L1 cache per core, 1 MB of L2 per core, and 256 MB of shared L3 cache. The Xeon 6780E has 96 KB of L1 per core, 4 MB of L2 per module, and 108 MB of shared L3. The AMD part has 256 MB of L3 versus 108 MB, a 137% advantage in total L3 capacity. This larger cache pool helps with data-heavy workloads.

Memory architecture is another differentiator. The EPYC 9745 supports twelve-channel DDR5 memory with a bandwidth of 576.0 GB/s. The Xeon 6780E supports eight-channel DDR5 with 409.6 GB/s. The AMD part has 40.6% more memory bandwidth, which directly explains its dominance in data compression and random string sorting.

PCIe lane counts also differ. The EPYC 9745 provides 128 PCIe Gen 5 lanes, while the Xeon 6780E provides 88. Both support ECC memory, and neither has integrated graphics. The EPYC 9745 uses AMD Socket SP5, while the Xeon uses Intel Socket 4710.

The Xeon 6780E has a die size of 578 mm², while the EPYC 9745 does not have a recorded die size in the database. The EPYC 9745 has a base clock of 2.40 GHz and a boost clock of 3.70 GHz, compared to the Xeon's 2.20 GHz base and 3.00 GHz boost. The AMD part has a 23.3% higher boost clock.

FAQ

Q: Which processor has more cores?

A: The Intel Xeon 6780E has 144 cores, while the AMD EPYC 9745 has 128 cores. However, the EPYC 9745 has 256 threads versus 144 for the Xeon, because the AMD part supports two threads per core.

Q: How much faster is the AMD EPYC 9745 in multi-threaded workloads?

A: In Cinebench R23 multicore, the EPYC 9745 scores 111093 versus 73723 for the Xeon 6780E, a 50.7% advantage. PassMark multithread shows a similar 50.7% margin, with 130698 versus 86734.

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

A: The largest delta is in PassMark extended instructions, where the EPYC 9745 scores 280477 against 114008 for the Xeon 6780E, a 146% lead. This means the AMD part more than doubles the Intel result in that workload.

Q: Does the Intel Xeon 6780E win any benchmark in the database?

A: No. The database records 16 head-to-head benchmarks, and the AMD EPYC 9745 wins all 16. The Intel Xeon 6780E has zero wins.

Q: How do their memory bandwidths compare?

A: The EPYC 9745 supports twelve-channel DDR5 with 576.0 GB/s bandwidth. The Xeon 6780E supports eight-channel DDR5 with 409.6 GB/s. The AMD part has 40.6% more memory bandwidth.

Q: What are the launch prices of these processors?

A: The AMD EPYC 9745 has a launch MSRP of $12141. The Intel Xeon 6780E has a launch MSRP of $11350.

Specification Differences

The two processors differ in several key specification fields. The AMD EPYC 9745 has 128 cores and 256 threads, while the Intel Xeon 6780E has 144 cores and 144 threads. The EPYC 9745 has a base clock of 2.40 GHz and a boost clock of 3.70 GHz, compared to 2.20 GHz and 3.00 GHz for the Xeon. The TDP is 400 watts for AMD and 330 watts for Intel.

The EPYC 9745 uses AMD Socket SP5 and the Zen 5 architecture with the codename Turin, fabricated on a 3 nm process at TSMC. The Xeon 6780E uses Intel Socket 4710 and the Sierra Forest architecture, fabricated on Intel's 5 nm process. The Xeon has a die size of 578 mm², while the EPYC does not have a recorded die size.

Cache configurations differ: the EPYC 9745 has 80 KB L1 per core, 1 MB L2 per core, and 256 MB shared L3. The Xeon 6780E has 96 KB L1 per core, 4 MB L2 per module, and 108 MB shared L3. Memory support is twelve-channel DDR5 with 576.0 GB/s for AMD versus eight-channel DDR5 with 409.6 GB/s for Intel. PCIe lanes are 128 for AMD versus 88 for Intel.

The release dates differ: the EPYC 9745 was released on 2024-10-09, while the Xeon 6780E was released on 2024-06-02. The part numbers are 100-000001460 for AMD and SRPG3 for Intel. Both have a production status of Active, neither has an unlocked multiplier, and both lack integrated graphics.

The Verdict

The data is unambiguous. The AMD EPYC 9745 outperforms the Intel Xeon 6780E in every single recorded benchmark. The smallest margin, data encryption, is still an 18.9% lead for AMD. Most margins hover around 50%, and the extended instructions workload shows a 146% advantage. The EPYC 9745 wins on single-thread performance, multi-thread performance, memory bandwidth, cache capacity, and PCIe lane count.

The Xeon 6780E does have two advantages in the specification sheet: 144 cores versus 128, and a lower TDP of 330 watts versus 400 watts. But the core count advantage does not translate into performance wins, because the Xeon lacks simultaneous multithreading and has much lower clock speeds. The EPYC 9745's 256 threads and 3.70 GHz boost clock simply outwork the Xeon's 144 threads and 3.00 GHz boost.

For server deployments that prioritize raw throughput, the EPYC 9745 is the clear choice. It delivers 50.7% higher Cinebench multicore scores and 90.8% higher PassMark integer math scores. The Xeon 6780E might be considered in environments where the 70-watt TDP difference is critical, but the database shows no performance scenario where the Intel part is preferable. The EPYC 9745 also has a higher average benchmark score of 425973 compared to 280438 for the Xeon, and it sits at the 100th percentile versus the 99th percentile for Intel.

The verdict from the recorded measurements is straightforward: any workload that benefits from more threads, more memory bandwidth, or higher clock speeds will run faster on the AMD EPYC 9745. The Intel Xeon 6780E offers no recorded performance advantage, making the EPYC 9745 the superior processor in every benchmark category present in the database.