AMD EPYC 9475F vs Intel Xeon 6980P Comparison

AMD EPYC 9475F vs Intel Xeon 6980P: a 48-core Zen 5 part against a 128-core Granite Rapids flagship. The benchmark data shows two very different philosophies for the same server socket class, and the results are split down the middle. The AMD chip dominates in single-thread and light-thread workloads, while the Intel chip takes the lead in raw math throughput and data compression. Here is the breakdown of where each processor stands based strictly on the recorded measurements.

Head-to-Head Benchmarks

The most striking result in the database is the PassMark physics test. The AMD EPYC 9475F scores 16,443, while the Intel Xeon 6980P scores 3,350. That is a 390.8% advantage for AMD, the largest delta in the entire comparison. This test heavily favors the AMD architecture and its higher per-core clock speeds, and the result is not close.

Single-thread performance follows a similar pattern. The EPYC 9475F records a PassMark single-thread score of 3,779, versus 1,681 for the Xeon 6980P. That is a 124.8% lead for AMD. The same scores appear under the singlethread test entry, confirming the result. This is a massive gap, and it explains why the AMD part feels snappier in any workload that does not scale perfectly across all cores.

The prime number test also goes to AMD, and it is another blowout. The EPYC 9475F scores 1,507, while the Xeon 6980P scores just 555. That is a 171.5% advantage for AMD. Prime number generation is a classic integer-heavy, branch-heavy workload, and the AMD core design with its higher base clock of 3.65 GHz versus 2.00 GHz appears to be the deciding factor.

The multithreaded PassMark score is interesting because it does not follow the core count narrative. The EPYC 9475F with 48 cores and 96 threads scores 122,476, while the Xeon 6980P with 128 cores and 256 threads scores 74,324. That is a 64.8% win for AMD. The database shows that the Xeon 6980P fails to convert its massive core advantage into a PassMark multithread win, likely due to the lower clock speeds and per-core efficiency.

Random string sorting goes narrowly to AMD. The EPYC 9475F scores 253,936 against 240,792 for Intel, a 5.5% difference. This is a memory-latency-sensitive workload, and the AMD chip's higher clock speeds help it edge out the Intel part despite the latter having a larger cache.

Now for the Intel wins. The floating-point math test is a clear victory for the Xeon 6980P. It scores 501,720 versus 406,524 for AMD, a 19% lead. The Intel part has way more cores to throw at FPU-heavy workloads, and it shows here.

Extended instructions go to Intel by a wider margin. The Xeon 6980P scores 214,794 versus 173,169 for AMD, a 19.4% lead. This suggests the Intel AVX-512 implementation is pulling its weight in vectorized code.

Data compression is also an Intel win. The Xeon 6980P scores 2,364,519 versus 2,156,305 for AMD, an 8.8% delta. The 504 MB of shared L3 cache on the Intel part likely helps here, as compression workloads benefit from large resident data sets.

Data encryption goes to Intel, but by a smaller margin. The Xeon 6980P scores 125,246 against 116,648 for AMD, a 6.9% lead. The Intel chip has more cores, but the AMD chip has higher clocks, and the result is a moderate win for Intel.

Integer math is the closest Intel victory. The Xeon 6980P scores 637,476 versus 605,696 for AMD, a 5% delta. This is a core-count-driven win, but the AMD chip keeps it respectable due to its per-core integer throughput.

In total, the head-to-head table records 6 wins for the AMD EPYC 9475F and 5 wins for the Intel Xeon 6980P. The magnitude of the AMD wins in physics and single-thread is enormous, while the Intel wins are mostly in the 5-19% range.

The Verdict

The data points to a clear split. The AMD EPYC 9475F is the pick for any workload that cares about single-thread responsiveness, low-latency operations, or lightly threaded tasks. Its 124.8% single-thread lead and 390.8% physics lead make it the obvious choice for databases, front-end web servers, or any application where a single core's speed is the bottleneck.

The Intel Xeon 6980P is the pick for massive parallel throughput, specifically in floating-point math and vectorized instruction streams. Its 19% lead in floating-point math and 19.4% lead in extended instructions show that it can chew through scientific computing or financial modeling workloads that scale across its 128 cores.

If the workload is pure PassMark multithread performance, the AMD EPYC 9475F wins by 64.8%. That is a surprising result given Intel's 128 cores versus AMD's 48 cores. The database suggests the AMD chip's higher clocks and more efficient core design deliver better aggregate throughput in this particular benchmark.

For data compression and encryption, the Intel Xeon 6980P is the better choice. It leads by 8.8% and 6.9% respectively. These are common server workloads, so the Intel part is not without merit in real-world deployments.

The overall average benchmark score favors AMD. The EPYC 9475F has an average score of 350,933, while the Xeon 6980P has an average score of 251,516. The AMD chip also sits at the 100th percentile against all CPUs, while the Intel part is at the 99th percentile.

Architecture Differences

The AMD EPYC 9475F is built on the Zen 5 architecture, codenamed Turin. It uses a 4 nm process node from TSMC. The chip is composed of 8 dies, each measuring 70.6 mm², for a total die size of 8x 70.6 mm². The transistor count is listed at 66,520 million. The L1 cache is 80 KB per core, L2 is 1 MB per core, and L3 is 256 MB shared.

The Intel Xeon 6980P uses the Granite Rapids architecture, part of the Xeon 6 generation. It is built on a 5 nm process node from Intel's own foundry. The die size is listed as 3x 598 mm², meaning three large dies. The transistor count is not recorded in the database. The L1 cache is 112 KB per core, L2 is 2 MB per core, and L3 is 504 MB shared.

The core counts differ by a factor of 2.67. The AMD chip has 48 cores and 96 threads, while the Intel chip has 128 cores and 256 threads. The AMD chip compensates with much higher clocks: a base clock of 3.65 GHz and a boost clock of 4.80 GHz. The Intel chip has a base clock of 2.00 GHz and a boost clock of 3.90 GHz.

Both chips use DDR5 memory with a twelve-channel memory bus. The AMD chip has a memory bandwidth of 576.0 GB/s, while the Intel chip has a higher bandwidth of 614.4 GB/s. Both support ECC memory.

PCIe support differs. The AMD EPYC 9475F provides Gen 5 with 128 lanes (CPU only), while the Intel Xeon 6980P provides Gen 5 with 96 lanes (CPU only). This gives AMD a 32-lane advantage for expansion cards or storage controllers.

Neither chip has integrated graphics. Both are marked as server/workstation parts with active production status.

Specification Differences

The two processors differ on several key specification fields. The core count is the most obvious: 48 for AMD versus 128 for Intel. Threads follow at 96 versus 256.

Clock speeds are significantly different. The AMD base clock is 3.65 GHz versus 2.00 GHz for Intel. Boost clocks are 4.80 GHz versus 3.90 GHz.

Thermal design power differs. The AMD EPYC 9475F is rated at 400 TDP, while the Intel Xeon 6980P is rated at 500 TDP. This is a 100-watt difference in favor of AMD.

The socket types are different. AMD uses Socket SP5, while Intel uses Socket 7529.

Cache hierarchy differs in size. L1 is 80 KB per core for AMD versus 112 KB per core for Intel. L2 is 1 MB per core for AMD versus 2 MB per core for Intel. L3 is 256 MB shared for AMD versus 504 MB shared for Intel.

Memory bandwidth differs slightly. AMD records 576.0 GB/s, Intel records 614.4 GB/s.

PCIe lanes differ. AMD has 128 lanes, Intel has 96 lanes.

The process node differs. AMD uses 4 nm from TSMC, Intel uses 5 nm from Intel.

Release dates are close. The AMD chip was released on 2024-10-09, while the Intel chip was released on 2024-09-23.

The launch MSRP for the AMD EPYC 9475F is $7592. The launch MSRP for the Intel Xeon 6980P is $12460.

FAQ

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

A: The AMD EPYC 9475F. It scores 3,779 in PassMark single-thread, which is 124.8% ahead of the Intel Xeon 6980P's score of 1,681.

Q: Which processor wins in floating-point math?

A: The Intel Xeon 6980P. It scores 501,720 versus 406,524 for the AMD EPYC 9475F, a 19% lead.

Q: Does the Intel Xeon 6980P win all the multithreaded benchmarks due to its 128 cores?

A: No. The AMD EPYC 9475F wins the PassMark multithread test with a score of 122,476 versus 74,324 for Intel, a 64.8% advantage. The Intel chip wins in integer math, floating-point math, and extended instructions.

Q: What is the memory bandwidth difference?

A: The Intel Xeon 6980P has a higher memory bandwidth at 614.4 GB/s, compared to 576.0 GB/s for the AMD EPYC 9475F. Both use twelve-channel DDR5.

Q: Which processor has more PCIe lanes?

A: The AMD EPYC 9475F has 128 Gen 5 lanes (CPU only), while the Intel Xeon 6980P has 96 Gen 5 lanes (CPU only).

Q: Which processor has the larger L3 cache?

A: The Intel Xeon 6980P has 504 MB of shared L3 cache, while the AMD EPYC 9475F has 256 MB of shared L3 cache.

Where Each One Wins

The AMD EPYC 9475F is the clear winner for single-threaded and lightly threaded workloads. Its 124.8% lead in single-thread performance and 390.8% lead in physics make it the go-to for latency-sensitive applications like transaction processing, real-time analytics, or any service where a single request needs to complete fast.

The AMD chip also wins in prime number generation (171.5% lead) and random string sorting (5.5% lead). These results point to a processor that excels at integer-heavy, branch-predictor-sensitive code with high clock speeds.

The AMD chip wins the PassMark multithread test by a 64.8% margin, which is a strong indicator for workloads that are not purely parallel but have some serial component. The higher per-core performance helps in mixed workloads like database queries with many small transactions.

The Intel Xeon 6980P is the winner for massively parallel numeric workloads. Its 19% lead in floating-point math and 19.4% lead in extended instructions make it suitable for scientific simulations, financial risk modeling, or any AVX-512-heavy application.

Intel also wins in data compression (8.8% lead) and data encryption (6.9% lead). These are common server tasks, and the Intel part's larger L3 cache and core count help it here.

For integer math, Intel wins by a narrow 5% margin. This is a workload where the core count matters, but the AMD chip's clock speed keeps it close.

The overall average benchmark score favors AMD, and the AMD chip holds the 100th percentile rank. The Intel chip sits at the 99th percentile. If a deployment is mixed and needs a balance of single-thread speed and multithread throughput, the database suggests the AMD EPYC 9475F is the safer default. If the workload is exclusively large-scale parallel math, the Intel Xeon 6980P has the edge.