AMD Radeon Pro 450 vs NVIDIA GeForce GTX 670 Comparison

Head-to-Head Benchmarks

The benchmark data records three head-to-head comparisons between the NVIDIA GeForce GTX 670 and the AMD Radeon Pro 450. The GTX 670 takes two of the three tests, with decisive margins in compute-oriented workloads, while the Radeon Pro 450 claims a significant victory in Apple's Metal API.

The largest win for the GTX 670 comes in the Geekbench OpenCL test, where it scores 15,341 against the Radeon Pro 450's 9,002. That is a 70.4% advantage, the single biggest gap in any test. In Vulkan, the GTX 670 again leads, scoring 15,553 versus 10,518, a 47.9% difference. These are not narrow margins; they represent substantial performance gaps in cross-platform compute and modern graphics APIs.

The Radeon Pro 450 fights back in Geekbench Metal, scoring 12,893 versus the GTX 670's 7,424. That puts the AMD card 42.4% ahead in this specific test. Metal is Apple's native graphics and compute API, so this result is particularly relevant for macOS users. The data suggests the Radeon Pro 450 is strongly optimized for Metal workloads, while the GTX 670, despite its age, holds up better in OpenCL and Vulkan.

Looking at average benchmark scores across all recorded tests, the GTX 670 posts 12,773 against the Radeon Pro 450's 10,804. That gives the NVIDIA card an overall lead of roughly 18% in the aggregate. The GTX 670 sits at the 52nd percentile of all GPUs in the database, while the Radeon Pro 450 lands at the 49th percentile. Both are mid-pack performers, but the GTX 670 edges ahead in overall standing.

The nearest rival data confirms the GTX 670's positioning. It sits within 0.6% of the AMD FirePro W5100 (12,847 average score) and within 0.5% of the AMD Radeon RX 7600M XT (12,710). The AMD Radeon Pro 455 is nearly identical at 12,831, just 0.4% behind the GTX 670's average. The Radeon Pro 450, by contrast, is closest to the NVIDIA Quadro K2200 (10,761), sitting just 0.4% ahead of it, and within 1.7% of the AMD Radeon RX 6600S (10,629). These proximity figures show that the GTX 670 competes in a slightly higher performance tier than the Radeon Pro 450.

Architecture Differences

The two cards come from different manufacturing generations and design philosophies. The GTX 670 uses the GK104 chip, built on NVIDIA's Kepler architecture, fabricated on a 28 nm process at TSMC. The Radeon Pro 450 uses the Baffin chip, based on AMD's GCN 4.0 architecture, produced on a 14 nm process at GlobalFoundries. The process node difference is stark: 28 nm versus 14 nm, which explains the Radeon Pro 450's much smaller die size of 123 mm² compared to the GTX 670's 294 mm².

Transistor counts tell a similar story. The GTX 670 packs 3,540 million transistors into its larger die, yielding a density of 12.0 million transistors per mm². The Radeon Pro 450 has 3,000 million transistors in a much smaller area, achieving 24.4 million per mm², roughly double the density. This is a direct consequence of the newer manufacturing process.

The compute resources differ substantially. The GTX 670 has 1,344 shading units, 112 texture mapping units, and 32 render output units. The Radeon Pro 450 has 640 shading units, 40 TMUs, and 16 ROPs. In every count, the GTX 670 has roughly double the hardware resources. This explains its dominance in OpenCL and Vulkan, where raw shader throughput matters.

Clock speeds are not fully specified for the Radeon Pro 450 (base and boost clocks are not recorded), but the memory clocks are listed. The GTX 670 runs its GDDR5 memory at 1502 MHz, effective 6 Gbps, while the Radeon Pro 450 runs at 1270 MHz, effective 5.1 Gbps. The GTX 670 also has a wider memory bus: 256-bit versus 128-bit. This yields memory bandwidth of 192.3 GB/s for the GTX 670 versus 81.28 GB/s for the Radeon Pro 450, a 2.4x advantage for NVIDIA.

Both cards have 2 GB of GDDR5 memory, so capacity is equal. The GTX 670 supports DirectX 12 (11_0) while the Radeon Pro 450 supports DirectX 12 (12_0), meaning the AMD card has a newer feature level. Both support OpenGL 4.6, but the Radeon Pro 450 has Vulkan 1.3 support versus the GTX 670's 1.2.175, giving AMD a newer API version.

The power profiles are dramatically different. The GTX 670 has a 170 W TDP, requires a dual-slot cooler, and needs two 6-pin power connectors plus a 450 W suggested PSU. The Radeon Pro 450 has a 35 W TDP and comes as an MXM module, a mobile form factor. The GTX 670 is a desktop card drawing nearly five times the power, which explains its higher performance ceiling.

Where Each One Wins

The GTX 670 is the clear winner for compute-heavy workloads outside of Apple's ecosystem. Its 70.4% lead in OpenCL and 47.9% lead in Vulkan make it the stronger choice for cross-platform applications, gaming under Vulkan, and general GPU compute tasks. The doubling of shading units, TMUs, and ROPs gives it a hardware advantage that the Radeon Pro 450 cannot overcome in these tests.

The Radeon Pro 450 wins decisively in Metal, with a 42.4% advantage. This makes it the better option for macOS-specific workflows, where Metal is the primary graphics and compute API. The data indicates that AMD's GCN 4.0 architecture is well-tuned for Metal, and the Radeon Pro 450's placement in the Radeon Pro Mac series suggests it was designed with Apple platforms in mind.

For gaming under Vulkan, the GTX 670 is the pick. Its Vulkan score of 15,553 is 47.9% higher than the Radeon Pro 450's 10,518. For OpenCL compute, whether for rendering, simulation, or data processing, the GTX 670's 15,341 versus 9,002 is a commanding lead. For Metal-based applications, particularly on macOS, the Radeon Pro 450's 12,893 versus 7,424 makes it the better fit.

The GTX 670 also wins on overall average score, 12,773 versus 10,804, and sits at a higher percentile (52nd versus 49th). The Radeon Pro 450's main advantages are power efficiency (35 W versus 170 W) and newer API support (DirectX 12_0, Vulkan 1.3). If the workload is Metal-centric or power-constrained, the Radeon Pro 450 is the sensible choice. For everything else in these benchmarks, the GTX 670 leads.

FAQ

Q: Which card has the higher average benchmark score?

A: The NVIDIA GeForce GTX 670 has an average benchmark score of 12,773, while the AMD Radeon Pro 450 averages 10,804. The GTX 670 leads by roughly 18% in the aggregate.

Q: How big is the GTX 670's lead in OpenCL?

A: The GTX 670 scores 15,341 in Geekbench OpenCL versus the Radeon Pro 450's 9,002, a 70.4% advantage. This is the largest margin in any recorded test between the two.

Q: Is the Radeon Pro 450 better in any benchmark?

A: Yes. The Radeon Pro 450 wins the Geekbench Metal test with a score of 12,893 against the GTX 670's 7,424, a 42.4% lead. This makes it the better choice for Metal-based applications.

Q: How do their memory bandwidths compare?

A: The GTX 670 has a 256-bit memory bus with 192.3 GB/s bandwidth, while the Radeon Pro 450 has a 128-bit bus with 81.28 GB/s. The GTX 670 offers more than double the memory bandwidth.

Q: What are their power requirements?

A: The GTX 670 has a TDP of 170 W and requires a 450 W suggested PSU with two 6-pin power connectors. The Radeon Pro 450 has a TDP of 35 W and uses an MXM module form factor with no separate power connectors listed.

Q: Which card supports newer APIs?

A: The Radeon Pro 450 supports DirectX 12 (12_0) and Vulkan 1.3, while the GTX 670 supports DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

The Verdict

The data points to a clear split based on use case. The NVIDIA GeForce GTX 670 is the stronger overall performer, with a higher average score, a higher percentile ranking, and decisive wins in OpenCL and Vulkan. It is the card to choose for general GPU compute, Vulkan-based gaming, or any workload that relies on cross-platform APIs. Its 70.4% lead in OpenCL and 47.9% lead in Vulkan are too large to ignore.

The AMD Radeon Pro 450 is the better choice for macOS users and Metal-centric workflows. Its 42.4% lead in Metal is substantial, and its 35 W TDP makes it far more power-efficient than the GTX 670's 170 W. For portable devices or systems where power draw and heat are primary concerns, the Radeon Pro 450 is the practical option. Its newer API support (DirectX 12_0, Vulkan 1.3) also gives it a future-proofing edge.

The GTX 670 sits at the 52nd percentile of all GPUs, with nearest rivals including the AMD Radeon Pro 455 and NVIDIA GeForce GTX 590, all within 0.6% of its average score. The Radeon Pro 450 sits at the 49th percentile, with closest rivals like the NVIDIA Quadro K2200 and GeForce GTX 560 Ti. This places the GTX 670 in a slightly higher performance class.

In practical terms, the GTX 670 is the better all-rounder for desktop users who need raw compute and gaming performance. The Radeon Pro 450 is the better specialist for Apple platforms and power-sensitive environments. Neither card is current, but the data shows the GTX 670 holds a broader performance advantage, while the Radeon Pro 450 wins specifically where Metal matters.

Specification Differences

| Specification | NVIDIA GeForce GTX 670 | AMD Radeon Pro 450 |

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

| Architecture | Kepler | GCN 4.0 |

| Process Node | 28 nm | 14 nm |

| Transistors | 3,540 million | 3,000 million |

| Die Size | 294 mm² | 123 mm² |

| Transistor Density | 12.0M / mm² | 24.4M / mm² |

| Memory Clock | 1502 MHz (6 Gbps effective) | 1270 MHz (5.1 Gbps effective) |

| Memory Bus Width | 256 bit | 128 bit |

| Memory Bandwidth | 192.3 GB/s | 81.28 GB/s |

| Shading Units | 1344 | 640 |

| TMUs | 112 | 40 |

| ROPs | 32 | 16 |

| Pixel Rate | 27.44 GPixel/s | 12.80 GPixel/s |

| Texture Rate | 109.8 GTexel/s | 32.00 GTexel/s |

| FP32 Performance | 2.634 TFLOPS | 1,024.0 GFLOPS |

| FP16 Performance | Not recorded | 1,024.0 GFLOPS (1:1) |

| TDP | 170 W | 35 W |

| Slot Width | Dual-slot | MXM Module |

| Power Connectors | 2x 6-pin | None listed |

| Suggested PSU | 450 W | Not listed |

| Bus Interface | PCIe 3.0 x16 | PCIe 3.0 x8 |

| Display Outputs | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 | Portable Device Dependent |

| DirectX Support | 12 (11_0) | 12 (12_0) |

| Vulkan Support | 1.2.175 | 1.3 |

| Dimensions | 241 mm x 111 mm x 38 mm | Not recorded |

| Release Date | 2012-05-09 | 2016-10-29 |

| Launch MSRP | 399 USD | Not recorded |