AMD EPYC 7F72 vs Intel Core Ultra 9 290HX Plus Comparison

Head-to-Head Benchmarks

The benchmark record shows a split decision between these two very different processors. The AMD EPYC 7F72 and Intel Core Ultra 9 290HX Plus each take a cluster of wins, and the margin of victory tells the real story.

Start with the most lopsided result in the entire dataset: Cinebench R23 single-core. The EPYC 7F72 scores 6328 against the Intel's 2356, a 168.6% advantage. That is not a small gap; it is a generational chasm in this specific test. The same pattern appears in Cinebench R15 single-core, where the EPYC leads 637 to 340, a 87.4% margin. For a 2020 server chip to beat a 2026 mobile flagship this decisively in single-threaded rendering is surprising, and it points to the EPYC's Zen 2 architecture being exceptionally well-tuned for this workload.

The Intel part hits back in Cinebench R20 single-core, scoring 2992 against the EPYC's 2657, an 11.2% win. It also dominates PassMark single-thread, scoring 4951 against 2384, a 51.8% advantage. So the two parts trade single-core victories depending on the benchmark generation and methodology.

In multi-core, the picture is mixed too. Cinebench R15 multicore goes to Intel: 5981 vs 4518, a 24.5% margin. Cinebench R20 multicore also favors Intel: 21198 vs 18828, an 11.2% gap. But Cinebench R23 multicore flips back to AMD: 44829 vs 39684, a 13% win. The Passmark multithread test goes to Intel again: 59439 vs 52740, an 11.3% advantage.

The two most telling workloads for real-world use are floating point math and physics. The Intel Core Ultra 9 290HX Plus crushes the EPYC in Passmark floating point math: 201773 vs 108437, a 46.3% lead. That is the biggest Intel win in the entire dataset. Meanwhile, the EPYC dominates Passmark physics: 6459 vs 3387, a 90.7% lead. Physics tests often stress scheduling and memory hierarchy, and the EPYC's 192 MB of shared L3 cache appears to help enormously.

Data compression also favors the EPYC: 808795 vs 658724, a 22.8% win. Data encryption goes to AMD as well: 56261 vs 50008, a 12.5% margin. Integer math is closer but still an AMD win: 181103 vs 164839, a 9.9% edge. Random string sorting is solidly AMD: 102436 vs 80327, a 27.5% win. Extended instructions go to Intel: 51290 vs 46936, an 8.5% win. Prime number finding is nearly tied, Intel edging it 519 vs 498, a 4% margin.

Where Each One Wins

The AMD EPYC 7F72 is the clear choice for workloads that depend on massive cache hierarchies and high core counts with simultaneous multithreading. The 12.5% win in data encryption, the 22.8% win in compression, and the 27.5% win in random string sorting all point to a processor that excels at data movement and large working sets. The 90.7% win in physics is the standout: this is a classic simulation and rendering workload, and the EPYC's 192 MB shared L3 cache is a decisive factor. The Cinebench R23 single-core win (168.6%) suggests that for certain types of legacy or single-threaded rendering tasks, the EPYC's architecture is uniquely fast.

The Intel Core Ultra 9 290HX Plus is the answer for mobile compute where you need peak single-thread responsiveness and heavy floating point work. The 51.8% win in Passmark single-thread, paired with the 46.3% win in floating point math, makes it the better option for scientific computing, financial modeling, or any workload that does heavy SIMD math. The Cinebench R15 and R20 multicore wins show it also holds its own in short rendering bursts. The extended instructions win by 8.5% suggests better support for modern vectorized code.

The split is clean: AMD wins 8 benchmarks, Intel wins 9. The EPYC takes the memory-bandwidth-hungry and cache-heavy tasks, the Intel takes the raw compute and single-thread tasks.

The Verdict

Choose the AMD EPYC 7F72 if your work looks like data compression, encryption, random string sorting, integer math, physics simulation, or Cinebench R23 (both single-core and multi-core). The 192 MB shared L3 cache is the clear differentiator, and its 90.7% physics win and 168.6% Cinebench R23 single-core win are too large to ignore. This is a processor for a server or workstation that sits in a rack and chews through data all day.

Choose the Intel Core Ultra 9 290HX Plus if your work looks like floating point math, Passmark single-thread, Cinebench R15 multi-core, or Cinebench R20 (both single and multi-core). The 46.3% floating point lead is the biggest single workload advantage in the Intel's column, and the 51.8% single-thread win makes it the better pick for interactive and latency-sensitive tasks. The 11.3% multithread win in Passmark also suggests better overall throughput in mixed workloads.

The database's average benchmark scores tell the same story: the EPYC 7F 72 sits at 85072 average with a 96th percentile against all CPUs, while the Intel sits at 79574 with a 95th percentile. The EPYC is the more powerful overall chip, but the Intel is very close, and it achieves that with a 55 W TDP versus the EPYC's 240 W.

FAQ

Q: Which processor has more cores and threads?

A: Both have 24 cores. The AMD EPYC 7F72 has 48 threads (2 threads per core), while the Intel Core Ultra 9 290HX Plus has only 24 threads (1 thread per core).

Q: Which one is faster in Cinebench R23 single-core?

A: The AMD EPYC 7F72 is overwhelmingly faster. It scores 6328 vs the Intel's 2356, a 168.6% lead.

Q: Which one wins in Passmark floating point math?

A: The Intel Core Ultra 9 290HX Plus is the clear winner. It scores 201773 vs the AMD's 108437, a 46.3% advantage.

Q: What is the difference in memory bandwidth?

A: The AMD EPYC 7F72 has eight-channel DDR4 memory with 204.8 GB/s bandwidth. The Intel Core Ultra 9 290HX Plus has dual-channel DDR5 memory with 102.4 GB/s bandwidth.

Q: Which processor has the larger L3 cache?

A: The AMD EPYC 7F72 has 192 MB of shared L3 cache. The Intel Core Ultra 9 290HX Plus has 36 MB of shared L3 cache.

Q: What is the production status and release date for each?

A: Both are marked Active. The AMD EPYC 7F72 was released on April 13, 2020. The Intel Core Ultra 9 290HX Plus was released on March 16, 2026.

Architecture Differences

The two processors come from fundamentally different design philosophies. The AMD EPYC 7F72 is built on Zen 2 architecture, codename Rome, on a 7 nm process at TSMC. The Intel Core Ultra 9 290HX Plus is based on the Arrow Lake-HX Refresh, in the Ultra 9 series, on a 3 nm process at TSMC. The process node advantage goes to Intel, but the transistor count tells a different story: the Intel has 17,800 million transistors in a 243 mm² die, while the AMD has 3,800 million transistors in a 74 mm² die. The Intel packs almost five times more transistors into a larger area.

The cache hierarchy is a major architectural split. The AMD has 96 KB of L1 per core, 512 KB of L2 per core, and a massive 192 MB of shared L3. The Intel has 192 KB of L1 per core, 3 MB of L2 per core, and 36 MB of shared L3. The Intel has larger per-core L1 and L2, but the AMD's L3 is over five times larger. This directly explains the AMD's wins in compression, encryption, and string sorting, workloads that benefit from keeping large datasets on-chip.

Memory architecture also diverges. The AMD uses DDR4 with an eight-channel memory bus and 204.8 GB/s bandwidth. The Intel uses DDR5 with a dual-channel bus and 102.4 GB/s bandwidth. The AMD has double the memory bandwidth, which contributes to its physics performance. The Intel has a newer memory standard, but half the channels.

PCIe support differs: the AMD has Gen 4, while the Intel has Gen 5 with 20 lanes from the CPU. The Intel also includes integrated Arc Xe-LPG Graphics with 64 execution units, while the AMD has none. Socket types are completely different: the AMD uses Socket SP3, the Intel uses BGA 2114. The AMD is a server/workstation part, the Intel is a mobile part. The Intel has an unlocked multiplier, the AMD does not.

Specification Differences

| Specification | AMD EPYC 7F72 | Intel Core Ultra 9 290HX Plus |

|---|---|---|

| Manufacturer | AMD | Intel |

| Cores / Threads | 24 / 48 | 24 / 24 |

| Base Clock | 3.20 GHz | 2.70 GHz |

| Boost Clock | 3.70 GHz | 5.50 GHz |

| TDP | 240 W | 55 W |

| Socket | AMD Socket SP3 | Intel BGA 2114 |

| Process Node | 7 nm | 3 nm |

| Foundry | TSMC | TSMC |

| Transistors | 3,800 million | 17,800 million |

| Die Size | 74 mm² | 243 mm² |

| L1 Cache | 96 KB (per core) | 192 KB (per core) |

| L2 Cache | 512 KB (per core) | 3 MB (per core) |

| L3 Cache | 192 MB (shared) | 36 MB (shared) |

| Memory Support | DDR4 | DDR5 |

| Memory Bus | Eight-channel | Dual-channel |

| Memory Bandwidth | 204.8 GB/s | 102.4 GB/s |

| ECC Memory | Yes | Yes |

| PCIe | Gen 4 | Gen 5, 20 Lanes (CPU only) |

| Integrated Graphics | None | Arc Xe-LPG Graphics 64EU |

| Market Segment | Server/Workstation | Mobile |

| Release Date | April 13, 2020 | March 16, 2026 |

| Multiplier Unlocked | No | Yes |

The boost clock difference is stark: the Intel runs at 5.50 GHz max versus the AMD's 3.70 GHz. That explains the Intel's single-thread wins in Passmark and Cinebench R20. The TDP difference is also dramatic: 240 W versus 55 W. The AMD uses over four times the power budget, but it is designed for a rack server, not a laptop. The Intel's integrated GPU is a significant feature if you need display output without a discrete card. Both support ECC memory, making both viable for error-sensitive workloads.