AMD EPYC 9174F vs Intel Xeon Gold 6348 Comparison

The Intel Xeon Gold 6348 and AMD EPYC 9174F are both active server/workstation processors, but they represent fundamentally different design philosophies. The data shows a clear, consistent edge for the AMD part across every single benchmark recorded, yet the Intel chip counters with a vastly higher core count and a different architectural approach. This analysis breaks down where each processor excels, what the underlying silicon differences mean, and who should ultimately choose which.

Where Each One Wins

The benchmark results are unambiguous: the AMD EPYC 9174F wins every head-to-head comparison in the FACT PACK. Across six Cinebench tests spanning R15, R20, and R23, the AMD chip takes the top spot each time, with a deltaPct of -5.2% or -5.3% relative to the Intel Xeon Gold 6348. This means the EPYC is consistently over 5% faster in both single-core and multi-core workloads, regardless of the rendering engine version.

The Intel Xeon Gold 6348, however, wins on paper in one crucial aspect: core count. It offers 28 cores and 56 threads, compared to the EPYC 9174F’s 16 cores and 32 threads. The data implies that the Intel part is designed for heavily threaded, parallel workloads where raw core quantity matters more than per-core speed. The EPYC, with its higher clock speeds and newer architecture, appears to be the choice for workloads that are latency-sensitive or moderately threaded, where its superior single-core performance and per-core efficiency shine.

In terms of raw throughput, the EPYC’s wins are not massive, but they are universal. The data shows the AMD part is the better performer in every measured scenario, making it the default pick for pure performance. The Intel part’s advantage is purely structural—more cores for parallel scaling—but the benchmark results show that even in multi-core tests, the EPYC 9174F’s higher frequency and efficiency overcome the core deficit.

Architecture Differences

The two processors come from different generations and foundries, which explains their performance gap. The Intel Xeon Gold 6348 is built on Intel’s 10 nm process node, using the Ice Lake-SP architecture. It features 28 cores based on the Ice Lake design, with 64 KB of L1 cache per core and 1 MB of L2 cache per core. The shared L3 cache is 42 MB. This is a mature, high-core-count design aimed at density and throughput in traditional data centers.

The AMD EPYC 9174F is a Zen 4 part, built on TSMC’s 5 nm process node, and is part of the EPYC 9004 series (Genoa). It has 16 cores, but each core is significantly faster. The L1 and L2 cache per core are identical to Intel (64 KB and 1 MB respectively), but the L3 cache is a massive 256 MB shared pool—over six times larger than Intel’s. The EPYC also has a transistor count of 52,560 million and a die size of 8x 72 mm², indicating a chiplet-based design, whereas the Intel part is a monolithic die (though die size is not listed for it).

Memory and I/O also differ drastically. The Intel Xeon supports DDR4 with an eight-channel memory bus and a bandwidth of 204.8 GB/s. The EPYC supports DDR5 with a twelve-channel bus and a bandwidth of 460.8 GB/s—more than double. PCIe lanes also favor AMD: Gen 5 with 128 lanes versus Intel’s Gen 4 with 64 lanes. These architectural choices mean the EPYC 9174F is not just faster per core; it has far more headroom for memory-intensive and I/O-heavy applications, which could explain its multi-core wins despite fewer cores.

Head-to-Head Benchmarks

The head-to-head data is a series of narrow but consistent defeats for Intel. In Cinebench R15 multi-core, the Intel Xeon Gold 6348 scores 4441 against the EPYC’s 4686, a -5.2% delta. The single-core R15 test shows 626 versus 661, a -5.3% delta. This pattern repeats exactly in R20: multi-core 18507 versus 19525 (-5.2%), and single-core 2612 versus 2756 (-5.2%). R23 multi-core has Intel at 44066 and AMD at 46489 (-5.2%), while single-core is 6221 versus 6563 (-5.2%).

The consistency of the -5.2% delta across multi-core tests is telling. It suggests the EPYC’s advantage is not just about clock speed or cores, but about overall efficiency. The EPYC’s base clock of 4.10 GHz and boost of 4.40 GHz are significantly higher than Intel’s 2.60 GHz base and 3.50 GHz boost. This frequency advantage, combined with a newer architecture, allows the 16-core EPYC to beat the 28-core Intel part even in multi-threaded workloads. The data implies that the Intel chip’s extra 12 cores cannot compensate for the per-core performance deficit.

In single-core tests, the delta is similar (-5.2% to -5.3%), showing that the EPYC’s higher boost clock and Zen 4 IPC are decisive. The Intel part’s single-core scores are lower, which would hurt in lightly threaded applications like legacy database queries or certain simulation tasks. The EPYC also has a Geekbench score of 2244 single-core and 17363 multi-core, which are not matched by Intel in the FACT PACK, further reinforcing its dominance in mixed workloads.

FAQ

Q: Which processor has more cores?

A: The Intel Xeon Gold 6348 has 28 cores and 56 threads, while the AMD EPYC 9174F has 16 cores and 32 threads.

Q: What is the performance difference in multi-core workloads?

A: The AMD EPYC 9174F wins all multi-core Cinebench tests by about 5.2%. For example, in Cinebench R23 multi-core, the EPYC scores 46489 versus Intel’s 44066.

Q: How do their memory systems compare?

A: The Intel Xeon uses DDR4 with an eight-channel bus and 204.8 GB/s bandwidth. The AMD EPYC uses DDR5 with a twelve-channel bus and 460.8 GB/s bandwidth, which is more than double the Intel part’s bandwidth.

Q: Is the AMD EPYC 9174F faster in single-core tests?

A: Yes, it wins all single-core Cinebench tests by about 5.2-5.3%. For instance, in R23 single-core, it scores 6563 versus Intel’s 6221.

Q: What is the process node for each chip?

A: The Intel Xeon Gold 6348 is on Intel’s 10 nm process, while the AMD EPYC 9174F is on TSMC’s 5 nm process.

Q: Which chip has a larger L3 cache?

A: The AMD EPYC 9174F has a 256 MB shared L3 cache, whereas the Intel Xeon Gold 6348 has only 42 MB of shared L3 cache.

Specification Differences

The core specifications diverge sharply, with the AMD EPYC 9174F being the newer, more advanced part. The Intel Xeon Gold 6348 has a base clock of 2.60 GHz and a boost of 3.50 GHz, with a TDP of 235 W. The AMD EPYC 9174F has a base clock of 4.10 GHz and a boost of 4.40 GHz, with a higher TDP of 320 W. The Intel part uses Intel Socket 4189, while the AMD part uses AMD Socket SP5.

Cache sizes differ mostly in L3: Intel has 42 MB shared, while AMD has 256 MB shared. Memory support is DDR4 for Intel and DDR5 for AMD, with bus widths of eight channels versus twelve channels, and bandwidths of 204.8 GB/s versus 460.8 GB/s. PCIe support is Gen 4 with 64 lanes for Intel, and Gen 5 with 128 lanes for AMD. The process node is 10 nm for Intel and 5 nm for AMD, with the AMD part using TSMC as the foundry. The release dates also differ: Intel’s is 2021-04-05, and AMD’s is 2022-11-09. The AMD part has a launch MSRP of $3850, which is the only price data available. The Intel part has no launch MSRP listed.

The Verdict

The data is clear for anyone prioritizing raw performance: the AMD EPYC 9174F is the superior processor in every benchmark category recorded. It wins all six Cinebench tests by roughly 5.2%, despite having 12 fewer cores. This confirms its higher clock speeds, newer Zen 4 architecture, and significantly better memory bandwidth (460.8 GB/s versus 204.8 GB/s). For workloads that are sensitive to single-thread performance or memory throughput—such as real-time analytics, financial modeling, or high-frequency trading—the EPYC 9174F is the definitive choice.

For the Intel Xeon Gold 6348, the appeal lies in its core count and potentially lower power draw per core (235 W TDP versus 320 W). The data suggests it is suited for massively parallel, throughput-oriented tasks where the software can scale across 28 cores, such as large-scale virtualization, batch processing, or rendering farms that can utilize every thread. However, the benchmark results show that even in multi-core Cinebench tests, the EPYC’s fewer but faster cores win out, meaning the Intel part would only be preferable if the specific workload does not benefit from the EPYC’s memory or clock advantages, or if the software is heavily optimized for Intel’s instruction set. The data does not show any workload where the Intel part wins, so the EPYC 9174F is the safer bet for general-purpose performance, while the Xeon Gold 6348 is a niche pick for extreme core-count scaling.