AMD EPYC 7F72 vs Intel Xeon 6736P Comparison

The Intel Xeon 6736P and AMD EPYC 7F72 are both 96th-percentile server processors, yet they achieve that standing through fundamentally different designs. The Xeon 6736P leans on a modern 5 nm process and 36 cores, while the EPYC 7F72 counters with higher clock speeds and a larger shared cache. Benchmark results show the AMD EPYC 7F72 wins 12 of 17 direct comparisons, but the Intel Xeon 6736P takes decisive victories in floating-point math, integer math, and extended instructions, making the choice highly workload-dependent.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon 6736P has 36 cores and 72 threads, while the AMD EPYC 7F72 has 24 cores and 48 threads. The Xeon holds a 50% core-count advantage.

Q: How do their boost clocks compare?

A: The AMD EPYC 7F72 boosts to 3.70 GHz, which is lower than the Intel Xeon 6736P's 4.10 GHz boost clock. However, the EPYC has a higher base clock at 3.20 GHz versus 2.00 GHz for the Xeon.

Q: Which processor wins in single-threaded workloads?

A: The AMD EPYC 7F72 wins all single-core Cinebench tests by 5% to 5.1% and leads PassMark single-thread by 15.1%. Its scores of 637 in Cinebench R15 and 2384 in PassMark single-thread outpace the Xeon's 605 and 2024.

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

A: The Intel Xeon 6736P leads by 38.1% in PassMark floating-point math (149770 vs 108437). The AMD EPYC 7F72 counters with a 21.3% lead in PassMark find prime numbers (498 vs 392).

Q: Do both processors support ECC memory?

A: Yes, both the Intel Xeon 6736P and AMD EPYC 7F72 support ECC memory. They also both use an eight-channel memory bus, though the Xeon supports DDR5 while the EPYC uses DDR4.

Q: What are the memory bandwidth figures?

A: The Intel Xeon 6736P delivers 409.6 GB/s of memory bandwidth, exactly double the AMD EPYC 7F72's 204.8 GB/s. This is a direct consequence of the Xeon's newer DDR5 support.

The Verdict

The data points to the AMD EPYC 7F72 as the better all-rounder for mixed server workloads, winning 12 of 17 head-to-head benchmarks. Its leads in every Cinebench test (5% to 5.1%) and PassMark multithread (5.1%) make it the safer pick for general rendering and threaded applications. The EPYC also dominates data encryption by 17.8% (56261 vs 46236) and prime-number finding by 21.3% (498 vs 392), which are common database and security tasks.

However, the Intel Xeon 6736P is the clear choice for math-intensive scientific computing. Its 38.1% lead in floating-point math and 14.2% lead in integer math (206833 vs 181103) show a massive advantage in number-crunching. It also wins extended instructions by 18.4% (55563 vs 46936), making it suitable for SIMD-heavy code. The Xeon's 96th-percentile ranking matches the EPYC's, but its 5 wins are concentrated where they matter most for HPC.

For buyers prioritizing memory bandwidth, the Xeon's 409.6 GB/s is unmatched by the EPYC's 204.8 GB/s, suggesting better performance in bandwidth-bound workloads like large-scale data processing. The EPYC's higher base clock (3.20 GHz vs 2.00 GHz) and larger L3 cache (192 MB vs 144 MB) explain its consistency across most benchmarks. No single processor wins every category, so the verdict hinges on whether the workload favors the EPYC's balanced throughput or the Xeon's raw math capability.

Architecture Differences

The Intel Xeon 6736P is built on Granite Rapids architecture using a 5 nm process from Intel, with a die size of 598 mm². The AMD EPYC 7F72 uses the older Zen 2 architecture (Rome) on a 7 nm process from TSMC, with a much smaller 74 mm² die but a transistor count of 3,800 million. This process gap is significant: the Xeon's smaller node allows higher boost clocks (4.10 GHz vs 3.70 GHz) despite having more cores.

Cache organization differs substantially. The Xeon allocates 112 KB of L1 and 2 MB of L2 per core, while the EPYC uses 96 KB L1 and 512 KB L2 per core. In L3, the EPYC has the advantage with 192 MB shared, compared to 144 MB shared on the Xeon. This larger L3 cache helps the EPYC compensate for its lower core count in many workloads.

Memory architecture is a major differentiator. The Xeon supports DDR5 with eight channels and 409.6 GB/s bandwidth, while the EPYC uses DDR4 with eight channels and 204.8 GB/s bandwidth. The Xeon also offers PCIe Gen 5 with 88 lanes (CPU only), whereas the EPYC provides PCIe Gen 4 without a lane count specified. Both are server/workstation parts with ECC support and active production status, but the Xeon sockets into Intel Socket 4710 while the EPYC uses AMD Socket SP3.

Head-to-Head Benchmarks

The AMD EPYC 7F72 sweeps all six Cinebench tests, with consistent 5% to 5.1% margins. In Cinebench R23 multicore, the EPYC scores 44829 against the Xeon's 42561, a 5.1% lead. Single-core R23 shows the same pattern: 6328 vs 6008, also 5.1% ahead. The EPYC's PassMark multithread score of 52740 beats the Xeon's 50072 by 5.1%, confirming its strength in heavily threaded rendering workloads.

The Intel Xeon 6736P dominates in raw math throughput. Its PassMark floating-point score of 149770 crushes the EPYC's 108437 by a massive 38.1%. Integer math follows with 206833 vs 181103, a 14.2% lead. Extended instructions show an 18.4% advantage for the Xeon (55563 vs 46936), indicating superior SIMD and cryptographic instruction execution.

The EPYC wins the remaining categories by varying margins. Data encryption shows a 17.8% lead (56261 vs 46236), and find prime numbers is 21.3% faster (498 vs 392). Data compression is close at 1.5% (808795 vs 796658), while physics and random string sorting are nearly tied, with the Xeon winning by 1.1% and 1.3% respectively. Single-thread PassMark goes to the EPYC by 15.1% (2384 vs 2024).

Specification Differences

The core and thread counts differ dramatically: 36 cores/72 threads for the Intel Xeon 6736P versus 24 cores/48 threads for the AMD EPYC 7F72. Base clocks are 2.00 GHz (Xeon) and 3.20 GHz (EPYC), while boost clocks are 4.10 GHz (Xeon) and 3.70 GHz (EPYC). TDP values also differ, with the Xeon rated at 205 W and the EPYC at 240 W.

Cache per core is distinct: the Xeon has 112 KB L1 and 2 MB L2, while the EPYC has 96 KB L1 and 512 KB L2. Shared L3 cache favors the EPYC at 192 MB versus 144 MB for the Xeon. Memory support is DDR5 for the Xeon and DDR4 for the EPYC, with bandwidth figures of 409.6 GB/s and 204.8 GB/s respectively. The Xeon uses PCIe Gen 5 with 88 lanes, while the EPYC uses PCIe Gen 4. Process nodes are 5 nm (Intel) and 7 nm (TSMC), with die sizes of 598 mm² and 74 mm². The Xeon's launch MSRP is $3351; the EPYC has no listed launch MSRP.

Where Each One Wins

The AMD EPYC 7F72 wins in general-purpose server tasks, taking all Cinebench benchmarks, PassMark multithread, data compression, data encryption, find prime numbers, and single-thread performance. It is the better choice for virtualization, database workloads, and any application that benefits from high base clocks (3.20 GHz) and a large 192 MB L3 cache. Its 5.1% lead in Cinebench R23 multicore and 17.8% lead in encryption make it ideal for mixed enterprise environments.

The Intel Xeon 6736P wins in mathematical and computational workloads, specifically floating-point math (38.1% lead), integer math (14.2% lead), and extended instructions (18.4% lead). Its 409.6 GB/s memory bandwidth and 36 cores also make it suitable for memory-bound HPC simulations. The Xeon's 4.10 GHz boost clock and 5 nm process give it an edge in bursty, low-threaded math tasks, even though it loses single-thread benchmarks overall. For physics calculations (1.1% lead) and random string sorting (1.3% lead), the Xeon also edges ahead, but these margins are negligible in practice.