AMD EPYC 9015 vs Intel Xeon Platinum 8260M Comparison

# Intel Xeon Platinum 8260M vs AMD EPYC 9015

The Intel Xeon Platinum 8260M and AMD EPYC 9015 represent two very different approaches to server processing, separated by roughly six years of architectural evolution. The Xeon Platinum 8260M is a 24-core Cascade Lake-SP part built on Intel's 14 nm process, while the EPYC 9015 is an 8-core Zen 5 (Turin) processor on TSMC's 4 nm node. Despite the core count disparity, the benchmark data shows a surprisingly competitive matchup: the Xeon wins 6 of 11 head-to-head tests, while the EPYC takes 5. The average benchmark scores are remarkably close—58,323 for Intel versus 57,555 for AMD—placing both processors at the 92nd percentile of all CPUs. The real story lies in how each chip dominates different workload categories, with the EPYC's single-thread speed contrasting sharply against the Xeon's raw multi-core throughput.

Where Each One Wins

The Intel Xeon Platinum 8260M is the clear winner in throughput-oriented workloads that scale with core count and thread parallelism. Its 24 cores and 48 threads give it a decisive edge in integer math, where it scores 129,213 versus the EPYC's 92,524—a 39.7% advantage. The gap widens even further in random string sorting, where the Xeon posts 69,236 against 39,772, a 74.1% lead that reflects the benefit of having three times as many physical cores. Data compression follows the same pattern: the Xeon's 573,896 score dwarfs the EPYC's 353,014 by 62.6%. Floating-point math also favors Intel, with 73,514 versus 61,615 (19.3% ahead), and the multithread benchmark shows a 12.7% edge at 34,592 versus 30,689. Extended instruction workloads round out Intel's wins with 35,529 versus 27,970, a 27% margin. The Xeon's advantage is rooted in parallel execution capability, not clock speed or architectural efficiency.

The AMD EPYC 9015 wins where single-thread performance and per-core efficiency matter most. Its single-thread score of 3,265 crushes the Xeon's 2,320—a 28.9% lead that reflects both the higher 4.10 GHz boost clock against 3.90 GHz and the fundamental IPC advantages of Zen 5 over Cascade Lake. The physics benchmark shows a similar 28.1% edge (2,893 versus 2,080), indicating stronger per-core computational power in latency-sensitive tasks. Prime number finding is the EPYC's most dramatic victory: 304 versus 141, a 53.6% margin that highlights the Zen 5 core's arithmetic efficiency. Data encryption is essentially a tie, with the EPYC narrowly ahead at 17,790 versus 17,625 (0.9% difference), suggesting that cryptographic workloads do not benefit significantly from the Xeon's extra cores in this comparison. The EPYC achieves these wins with only 8 cores and 16 threads, demonstrating that architectural superiority can overcome a 3:1 core deficit in certain workloads.

Architecture Differences

The two processors could hardly be more different in construction. The Intel Xeon Platinum 8260M uses Cascade Lake-SP architecture on a 14 nm process with 8,000 million transistors. It packs 24 cores and 48 threads, with a base clock of 2.40 GHz and boost clock of 3.90 GHz. The cache hierarchy includes 64 KB of L1 per core, 1 MB of L2 per core, and 35.75 MB of shared L3 cache. It supports DDR4 memory with ECC, sits in Intel Socket 3647, and was released in December 2018. The Xeon is not multiplier-unlocked and targets the server/workstation market segment.

The AMD EPYC 9015 belongs to the EPYC 9005 series, built on Zen 5 (Turin) architecture. It uses a 4 nm process from TSMC with 16,630 million transistors spread across two chiplets, each measuring 70.6 mm². The core configuration is much leaner: 8 cores and 16 threads, but with significantly higher clocks—3.60 GHz base and 4.10 GHz boost. The cache design differs substantially: 80 KB of L1 per core, 1 MB of L2 per core, and a much larger 64 MB of shared L3. Memory support jumps to DDR5 with a twelve-channel memory bus and 576.0 GB/s of bandwidth, versus the Xeon's unspecified DDR4 configuration. The EPYC also brings PCIe Gen 5 with 128 lanes (CPU only), whereas the Xeon's PCIe capability is not listed. The EPYC was released in October 2024, has a launch MSRP of $527, and is marked as Active in production status. The Xeon's launch MSRP is not provided.

The transistor count difference is striking: 16,630 million versus 8,000 million, despite the EPYC having one-third the cores. This reflects the denser 4 nm process and the more complex Zen 5 core design. The EPYC's dual-chiplet layout with 70.6 mm² dies also contrasts with the Xeon's monolithic approach, though the Xeon's die size is not specified. The memory bandwidth advantage of the EPYC—576.0 GB/s through twelve DDR5 channels—likely contributes to its strong performance in memory-sensitive workloads, while the Xeon's larger core count compensates in compute-bound tasks.

Head-to-Head Benchmarks

The benchmark suite reveals a clear workload split. Starting with the Xeon's strongest showing, random string sorting delivers a 74.1% margin (69,236 versus 39,772). This benchmark heavily rewards parallel thread execution, and the Xeon's 48 threads simply overwhelm the EPYC's 16. Data compression follows at 62.6% (573,896 versus 353,014), again leveraging the core advantage. Integer math shows a 39.7% lead (129,213 versus 92,524), while extended instructions trail at 27% (35,529 versus 27,970). Floating-point math is closer at 19.3% (73,514 versus 61,615), and the multithread benchmark shows a 12.7% edge (34,592 versus 30,689). These six wins for Intel are all in parallelizable, throughput-oriented workloads.

The EPYC's five wins are concentrated in single-thread and per-core performance. The single-thread benchmark (listed twice in the data as passmark_single_thread and passmark_singlethread) shows 3,265 versus 2,320, a 28.9% advantage. Physics follows at 28.1% (2,893 versus 2,080). Prime number finding is the largest single-core margin at 53.6% (304 versus 141), which is particularly notable because it requires both fast integer arithmetic and efficient branch prediction—areas where Zen 5 excels. Data encryption is the closest contest, with the EPYC winning by just 0.9% (17,790 versus 17,625), effectively a statistical tie. The EPYC's wins are smaller in absolute terms than the Xeon's, but they come from a chip with one-third the cores and half the threads.

When comparing to the nearest rivals, the Xeon's average score of 58,323 places it 0.1% behind the AMD Ryzen 7 9850X3D (58,386) and 0.3% behind the Intel Xeon w5-2545 (58,504), while leading the Intel Core i9-14900 (58,115) by 0.4% and the AMD Ryzen AI 9 HX 470 (58,826) by 0.9%. The EPYC's average of 57,555 puts it 0.1% ahead of the AMD Ryzen 9 9900X (57,498) and 0.3% ahead of the AMD EPYC 7313 (57,399), while trailing the Intel Core i9-14900 by 1% and the Intel Xeon Platinum 8260M by 1.3%. Both processors sit at the 92nd percentile of all CPUs, indicating that despite their different architectures, they deliver comparable overall performance in the benchmark database.

FAQ

Q: Which processor has more cores and threads?

A: The Intel Xeon Platinum 8260M has 24 cores and 48 threads, while the AMD EPYC 9015 has 8 cores and 16 threads. This gives the Xeon a 3:1 advantage in both core and thread count.

Q: How do their clock speeds compare?

A: The EPYC 9015 has a higher base clock of 3.60 GHz versus 2.40 GHz for the Xeon, and a higher boost clock of 4.10 GHz versus 3.90 GHz. This helps explain the EPYC's 28.9% single-thread advantage.

Q: Which processor wins in multi-threaded workloads?

A: The Xeon Platinum 8260M wins the multithread benchmark with 34,592 versus 30,689, a 12.7% margin. It also dominates data compression (62.6% ahead), integer math (39.7% ahead), and random string sorting (74.1% ahead).

Q: What memory technologies do they support?

A: The Xeon supports DDR4 memory with ECC, while the EPYC supports DDR5 with a twelve-channel memory bus and 576.0 GB/s bandwidth. The EPYC also provides PCIe Gen 5 with 128 lanes.

Q: Which processor is more recent?

A: The EPYC 9015 was released in October 2024, while the Xeon Platinum 8260M was released in December 2018. The EPYC is built on a 4 nm process from TSMC, compared to the Xeon's 14 nm process from Intel.

Q: How close are their overall benchmark scores?

A: The Xeon has an average benchmark score of 58,323, while the EPYC scores 57,555. That is a 1.3% difference, with both processors at the 92nd percentile of all CPUs.

The Verdict

The data presents a clear choice based on workload requirements. The Intel Xeon Platinum 8260M is the processor for parallel throughput. Its 24 cores and 48 threads deliver substantial wins in data compression (62.6% ahead), random string sorting (74.1% ahead), and integer math (39.7% ahead). For server workloads that scale with thread count—database operations, batch processing, virtualization hosts, or any heavily parallelized compute—the Xeon's benchmark results are unambiguous. The multithread score of 34,592 versus 30,689 confirms that even in mixed workloads, the core advantage translates to real performance gains.

The AMD EPYC 9015 is the choice for single-thread performance and per-core efficiency. Its 28.9% single-thread lead and 53.6% prime number finding advantage indicate that latency-sensitive applications, high-frequency trading, or workloads with limited parallelism will perform better on the EPYC. The physics benchmark win (28.1%) also suggests that simulation and modeling tasks with strict per-core requirements favor AMD. The EPYC's smaller power envelope (125 W TDP versus 165 W) and modern DDR5 memory with 576.0 GB/s bandwidth add to its appeal for dense, power-conscious deployments.

The overall average scores are nearly identical, and both processors sit at the 92nd percentile. The decision hinges entirely on whether the workload is thread-bound or latency-bound. The Xeon's six wins in throughput benchmarks outweigh the EPYC's five wins in per-core tests in terms of raw margin, but the EPYC achieves its results with one-third the cores, making it a more efficient design per core. For new deployments where software is optimized for high clock speeds and modern instructions, the EPYC 9015's Zen 5 architecture is compelling. For existing infrastructure that leverages many threads and can tolerate the older process node, the Xeon Platinum 8260M remains a strong performer. The benchmark database shows two excellent processors, each optimized for different halves of the server workload spectrum.