GPU Comparison

Head-to-Head Benchmarks

The only direct benchmark comparison available in the data is Geekbench OpenCL, and it delivers a clear result. The AMD Radeon Instinct MI25 scores 68,562 against the Intel Arc Pro A60's 63,485. That is an 8% advantage for the AMD part, a meaningful margin in compute workloads. The data shows the MI25 winning the sole head-to-head test, giving it a 1-0 record in direct competition.

Context from the nearest rivals sharpens the picture. The MI25 sits just 0.4% behind the Intel Arc A770 (68,809) and 0.6% behind the NVIDIA CMP 90HX (69,000). It trails the AMD Radeon Pro WX 8200 by 1.9% (69,870) and the NVIDIA Quadro P6000 by 2% (69,986). Those are tight gaps — the MI25 is essentially trading blows with cards that are typically considered upper-midrange to high-end for their respective generations.

The Intel Arc Pro A60, meanwhile, sits in a different neighborhood. Its 63,485 OpenCL score places it just 0% away from the NVIDIA GeForce RTX 4090 (60,347) — a statistical tie that looks odd given the RTX 4090's reputation, but the numbers are what they are. The A60 is 0.3% ahead of the AMD Radeon Pro Vega 48 (60,140) and 2.5% behind the AMD Radeon Pro W6600M (61,896). It is 2.8% ahead of the AMD Radeon PRO V710 (58,657). The A60's overall average benchmark score of 60,326 combines its OpenCL result with a Vulkan score of 57,166, pulling its average down relative to its OpenCL peak.

The 8% delta between the two cards in OpenCL is the headline number. It is not a landslide, but it is consistent — the MI25 outperforms the A60 in the one test where both have data. The A60's Vulkan score of 57,166 has no corresponding MI25 result, so cross-API comparisons are not possible from this pack. The data only supports a direct comparison in OpenCL, and there the AMD card wins.

Where Each One Wins

The MI25 wins in raw compute throughput. Its FP32 peak of 12.29 TFLOPS stands well above the A60's 8.397 TFLOPS — a 46% advantage in theoretical single-precision performance. That translates directly to the OpenCL benchmark result, where the MI25's lead is 8%. The gap between theoretical FP32 and real-world benchmark is instructive: the MI25 does not fully convert its compute advantage into benchmark wins, but it still leads.

The A60 wins in efficiency and feature set. Its TDP of 130 W is less than half the MI25's 300 W, and it requires only a 300 W suggested PSU versus 700 W for the AMD card. The A60 is single-slot and needs no external power connectors, while the MI25 is dual-slot with 2x 8-pin connectors. For dense deployments or workstations with limited power budgets, the A60 is clearly the more practical choice.

The A60 also wins on API support. It supports DirectX 12 Ultimate (12_2), while the MI25 only reaches DirectX 12 (12_1). The A60 supports Vulkan 1.4; the MI25 is on Vulkan 1.3. Both support OpenGL 4.6. The A60 also brings 16 ray tracing cores, a feature the MI25 lacks entirely. For modern graphics workloads or applications that leverage ray tracing, the A60 has a structural advantage that no amount of compute throughput can overcome on the MI25.

Memory configuration favors the MI25 on capacity and bandwidth. The AMD card has 16 GB of HBM2 across a 2048-bit bus, delivering 436.2 GB/s. The A60 has 12 GB of GDDR6 on a 192-bit bus, delivering 384.0 GB/s. The MI25 leads by 4 GB of capacity and 52.2 GB/s of bandwidth — a 13.6% bandwidth advantage. But the A60's 16 Gbps effective memory speed is far higher than the MI25's 1704 Mbps effective, reflecting the different memory architectures. HBM2 offers wider buses; GDDR6 offers faster clocks per pin.

Architecture Differences

The two cards come from fundamentally different design philosophies. The MI25 is built on AMD's GCN 5.0 architecture (Vega 10 chip), fabricated on a 14 nm process at GlobalFoundries. The A60 uses Intel's Xe-HPG architecture (DG2-256 chip), built on TSMC's 6 nm node. The process gap is enormous: 14 nm versus 6 nm means the Intel part packs transistors much more densely.

The numbers confirm this. The MI25 has 12,500 million transistors on a 495 mm² die, yielding a density of 25.3 million transistors per mm². The A60 has 11,500 million transistors on a 269 mm² die — a density of 42.8 million per mm². The Intel chip is 45.6% smaller but contains nearly as many transistors. That is the 6 nm advantage in action.

Shader configuration differs sharply. The MI25 has 4,096 shading units, 256 TMUs, and 64 ROPs. The A60 has 2,048 shading units, 128 TMUs, and 64 ROPs. The AMD card has exactly double the shaders and TMUs, but the same ROP count. Pixel rate tells the story: the A60's 131.2 GPixel/s is 36.7% higher than the MI25's 96.00 GPixel/s, despite the AMD card's larger shader array. The A60's boost clock of 2050 MHz versus the MI25's 1500 MHz boost explains part of that. Texture rate favors the MI25: 384.0 GTexel/s versus 262.4 GTexel/s, a 46.3% lead for AMD.

Ray tracing is exclusive to the A60, which carries 16 dedicated RT cores. The MI25 has none. This is a generation-defining difference — the MI25 predates mainstream ray tracing hardware, while the A60 is built for it. The A60 also supports DirectX 12 Ultimate, which mandates ray tracing and mesh shaders. The MI25's DirectX 12 (12_1) support lacks those features.

The bus interface differs as well. The A60 uses PCIe 4.0 x16, while the MI25 uses PCIe 3.0 x16. For data transfer to and from the host, the A60 has double the bandwidth available. Display outputs are another major split: the A60 has 4x DisplayPort 2.0 outputs, while the MI25 has no display outputs at all. The MI25 is a pure compute accelerator; the A60 is a workstation GPU that can drive monitors.

The Verdict

The data supports a clear split by workload. For pure compute performance, the MI25 wins. Its OpenCL score is 8% higher, its FP32 throughput is 46% higher, and it offers more memory capacity and bandwidth. If the task is number-crunching — machine learning inference, scientific simulation, or any FP32-heavy workload — the MI25 is the stronger card, despite being five years older in release date (2017 versus 2023).

For anything involving graphics output, ray tracing, or modern API features, the A60 is the only reasonable choice. The MI25 cannot output video at all; the A60 has 4x DisplayPort 2.0. The A60 supports ray tracing; the MI25 does not. The A60 runs on half the power (130 W versus 300 W), fits in a single slot, and needs no external power connectors. The A60's production status is active, while the MI25 is end-of-life.

The percentile rankings echo this nuance. The MI25 sits at the 90th percentile of all GPUs, while the A60 sits at the 88th. The MI25's average benchmark score of 68,562 is higher than the A60's 60,326 — a 13.6% gap in average performance. But the A60's Vulkan score of 57,166 drags down its average; in OpenCL alone, the gap is only 8%.

The release dates matter for context. The MI25 launched on June 26, 2017; the A60 on June 5, 2023. That six-year gap explains the architectural differences: the MI25 is a product of the pre-ray-tracing era, while the A60 is a modern workstation part. The MI25's predecessor is listed as FirePro Data Center, and it has no successor — AMD moved on from this architecture. The A60 has no listed predecessor or successor, suggesting it is a standalone entry in Intel's pro lineup.

FAQ

Q: Which card has the higher Geekbench OpenCL score?

A: The AMD Radeon Instinct MI25 scores 68,562 versus the Intel Arc Pro A60's 63,485 — an 8% advantage for AMD.

Q: Does the Intel Arc Pro A60 support ray tracing?

A: Yes. The A60 has 16 ray tracing cores and supports DirectX 12 Ultimate (12_2), which requires ray tracing support. The MI25 has no ray tracing cores.

Q: How do the memory configurations compare?

A: The MI25 has 16 GB of HBM2 on a 2048-bit bus with 436.2 GB/s bandwidth. The A60 has 12 GB of GDDR6 on a 192-bit bus with 384.0 GB/s bandwidth. The MI25 leads in capacity and bandwidth.

Q: What is the power consumption difference?

A: The MI25 has a 300 W TDP and requires a 700 W suggested PSU with 2x 8-pin connectors. The A60 has a 130 W TDP and requires a 300 W suggested PSU with no external power connectors.

Q: Can either card drive displays?

A: Only the A60. It has 4x DisplayPort 2.0 outputs. The MI25 has no display outputs at all and is a compute-only accelerator.

Q: How do the FP32 compute figures differ?

A: The MI25 delivers 12.29 TFLOPS, while the A60 delivers 8.397 TFLOPS. The AMD card has a 46% higher FP32 peak.

Specification Differences

| Specification | AMD Radeon Instinct MI25 | Intel Arc Pro A60 |

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

| Architecture | GCN 5.0 | Xe-HPG |

| Process Node | 14 nm | 6 nm |

| Foundry | GlobalFoundries | TSMC |

| Transistors | 12,500 million | 11,500 million |

| Die Size | 495 mm² | 269 mm² |

| Transistor Density | 25.3M / mm² | 42.8M / mm² |

| Base Clock | 1400 MHz | 900 MHz |

| Boost Clock | 1500 MHz | 2050 MHz |

| Memory Size | 16 GB | 12 GB |

| Memory Type | HBM2 | GDDR6 |

| Memory Bus Width | 2048 bit | 192 bit |

| Memory Bandwidth | 436.2 GB/s | 384.0 GB/s |

| Shading Units | 4096 | 2048 |

| TMUs | 256 | 128 |

| ROPs | 64 | 64 |

| RT Cores | None | 16 |

| Pixel Rate | 96.00 GPixel/s | 131.2 GPixel/s |

| Texture Rate | 384.0 GTexel/s | 262.4 GTexel/s |

| FP32 | 12.29 TFLOPS | 8.397 TFLOPS |

| FP16 | 24.58 TFLOPS (2:1) | 16.79 TFLOPS (2:1) |

| TDP | 300 W | 130 W |

| Slot Width | Dual-slot | Single-slot |

| Power Connectors | 2x 8-pin | None |

| Suggested PSU | 700 W | 300 W |

| Bus Interface | PCIe 3.0 x16 | PCIe 4.0 x16 |

| Display Outputs | No outputs | 4x DisplayPort 2.0 |

| DirectX | 12 (12_1) | 12 Ultimate (12_2) |

| Vulkan | 1.3 | 1.4 |

| Production Status | End-of-life | Active |

| Release Date | 2017-06-26 | 2023-06-05 |

| Geekbench OpenCL | 68,562 | 63,485 |

| Average Benchmark Score | 68,562 | 60,326 |

| Percentile (All GPUs) | 90 | 88 |