AMD Radeon HD 8790M vs NVIDIA GeForce GTX 670M Comparison

The Verdict

The NVIDIA GeForce GTX 670M is the stronger GPU for general compute workloads, based on the recorded benchmark data. It wins the only head-to-head test, Geekbench OpenCL, by a substantial 29.8% margin over the AMD Radeon HD 8790M. The GTX 670M also holds a higher percentile ranking at 38, compared to the Radeon HD 8790M's 33, meaning it sits above a larger share of all GPUs in the database. Its average benchmark score of 6513 versus 5691 confirms this advantage in raw measured performance.

The AMD Radeon HD 8790M, however, is not without merit. It carries a newer architecture and a smaller, more efficient process node, and it is the only one of the two with a Vulkan benchmark score recorded (6365). For users specifically targeting Vulkan-based applications, the Radeon HD 8790M has measurable capability that the GTX 670M cannot demonstrate in the database. If your workload relies on Vulkan, the AMD part is the safer choice. If your work is OpenCL-centric, the GTX 670M leads clearly.

Choose the GTX 670M for higher raw throughput in OpenCL tasks and a better overall standing in the database. Choose the Radeon HD 8790M for Vulkan support and a more modern feature set, accepting a measurable drop in OpenCL performance.

Where Each One Wins

The GTX 670M wins in pure compute throughput. Its OpenCL score of 6513 outpaces the Radeon HD 8790M's 5017 by 29.8%. That gap is decisive: the NVIDIA part delivers roughly 30% more performance in this specific workload. The GTX 670M also wins on memory bandwidth, pushing 72.00 GB/s over a 192-bit bus, compared to the Radeon's 64.00 GB/s over a 128-bit bus. Higher bandwidth favors the GTX 670M in memory-intensive tasks such as large data transfers or texture streaming.

The Radeon HD 8790M wins in architectural efficiency and API support. It is built on a 28 nm process versus the GTX 670M's 40 nm node, and its transistor density is more than double: 12.3M per mm² against 5.9M per mm². That density advantage means the Radeon packs more logic into a much smaller die, 77 mm² versus 332 mm². It also supports Vulkan 1.2.170, while the GTX 670M lists no Vulkan support at all. The Radeon's base and boost clocks are recorded at 850 MHz and 900 MHz, giving it a defined clock structure that the GTX 670M lacks in the database.

In shading unit count, the Radeon has more: 384 versus 336. Yet the GTX 670M still wins in pixel rate and texture rate. The NVIDIA part achieves 8.372 GPixel/s and 33.49 GTexel/s, while the Radeon manages 7.200 GPixel/s and 21.60 GTexel/s. So the GTX 670M turns its hardware into higher throughput per clock, despite having fewer shaders.

Architecture Differences

The two GPUs come from different architectural generations. The GTX 670M uses NVIDIA's Fermi 2.0 architecture on the GF114 chip, fabricated by TSMC on a 40 nm process. It packs 1,950 million transistors into a 332 mm² die, yielding a transistor density of 5.9M per mm². The Radeon HD 8790M uses AMD's GCN 1.0 architecture on the Mars chip, also built by TSMC but on a 28 nm process. It contains 950 million transistors on a 77 mm² die, giving a density of 12.3M per mm². The Radeon is far smaller and denser, a sign of a more modern manufacturing process.

The memory subsystems differ as well. The GTX 670M has 1536 MB of GDDR5 on a 192-bit bus, producing 72.00 GB/s of bandwidth. The Radeon HD 8790M has 2 GB of GDDR5 on a 128-bit bus, producing 64.00 GB/s. The NVIDIA part offers higher bandwidth but less total capacity. The Radeon's memory clock is listed at 1000 MHz (4 Gbps effective), while the GTX 670M's memory clock is 750 MHz (3 Gbps effective). The Radeon runs its memory faster, but the narrower bus limits overall throughput.

Compute resources differ in configuration. The GTX 670M has 336 shading units, 56 texture mapping units, and 24 ROPs. The Radeon HD 8790M has 384 shading units, 24 TMUs, and only 8 ROPs. The Radeon has more shaders but far fewer texture units and ROPs, which explains its lower texture and pixel rates. The GTX 670M's FP32 output is 803.7 GFLOPS versus the Radeon's 691.2 GFLOPS, a 16% lead in raw floating-point throughput.

API support diverges clearly. Both support DirectX 12, but the GTX 670M lists it as 12 (11_0) while the Radeon lists 12 (11_1). OpenGL is identical at 4.6. The Radeon adds Vulkan 1.2.170; the GTX 670M has no Vulkan entry. This is a significant feature difference for any application that can use Vulkan.

The bus interfaces also differ: the GTX 670M uses MXM-B (3.0), while the Radeon uses MXM-A (3.0). Both are MXM modules with no power connectors and portable-device-dependent display outputs. The GTX 670M has a recorded TDP of 75 W; the Radeon's TDP is not listed in the database.

FAQ

Q: Which GPU has the higher OpenCL score?

A: The NVIDIA GeForce GTX 670M scores 6513 in Geekbench OpenCL, while the AMD Radeon HD 8790M scores 5017. The GTX 670M is 29.8% ahead in that benchmark.

Q: Does the AMD Radeon HD 8790M support Vulkan?

A: Yes. The Radeon HD 8790M has a recorded Geekbench Vulkan score of 6365 and lists Vulkan 1.2.170 support. The GTX 670M has no Vulkan support listed in the database.

Q: What is the memory bandwidth difference?

A: The GTX 670M provides 72.00 GB/s over a 192-bit bus, while the Radeon HD 8790M provides 64.00 GB/s over a 128-bit bus. The NVIDIA part has 12.5% more bandwidth.

Q: How do transistor counts compare?

A: The GTX 670M contains 1,950 million transistors on a 332 mm² die. The Radeon HD 8790M contains 950 million transistors on a 77 mm² die. The Radeon uses a 28 nm process versus the GTX 670M's 40 nm process.

Q: Which GPU has more shading units?

A: The Radeon HD 8790M has 384 shading units, compared to the GTX 670M's 336. Despite having fewer shaders, the GTX 670M achieves higher FP32 throughput at 803.7 GFLOPS versus 691.2 GFLOPS.

Q: What is the release timeline difference?

A: The GTX 670M was released on 2012-03-21, and the Radeon HD 8790M was released on 2013-03-31. The Radeon is roughly a year newer. Both are end-of-life products.

Head-to-Head Benchmarks

The only direct benchmark comparison in the database is Geekbench OpenCL. The GTX 670M scores 6513, and the Radeon HD 8790M scores 5017. That is a delta of 29.8%, a wide margin that clearly favors the NVIDIA part. In practical terms, the GTX 670M delivers nearly one-third more performance in this compute test. For any OpenCL-based workload, whether video encoding, physics simulation, or data processing, the GTX 670M is the stronger choice.

The Radeon HD 8790M does have a Vulkan score of 6365, but there is no corresponding Vulkan result for the GTX 670M to compare. That absence makes a direct head-to-head Vulkan comparison impossible. What the data shows is that the Radeon can run Vulkan workloads and produce a competitive score. The GTX 670M cannot claim any Vulkan capability in the database.

When looking at derived performance metrics, the GTX 670M also leads in pixel and texture throughput. Its pixel rate of 8.372 GPixel/s exceeds the Radeon's 7.200 GPixel/s by 16.3%. Its texture rate of 33.49 GTexel/s is 55% higher than the Radeon's 21.60 GTexel/s. These are not head-to-head benchmark runs, but they are recorded specifications that indicate where each GPU has a hardware advantage.

The FP32 numbers reinforce the same story. The GTX 670M outputs 803.7 GFLOPS, the Radeon 691.2 GFLOPS. The NVIDIA part is 16.3% ahead in raw floating-point compute. Combined with its higher bandwidth, the GTX 670M is simply the more capable compute device in most scenarios.

The Radeon's advantages are architectural. It has a smaller die (77 mm² vs 332 mm²), a denser transistor layout (12.3M / mm² vs 5.9M / mm²), and a more recent release date (2013-03-31 vs 2012-03-21). It also supports Vulkan 1.2.170 and a slightly higher DirectX feature level (12 (11_1) vs 12 (11_0)). These are meaningful for compatibility with newer software, but they do not translate into wins in the recorded compute benchmarks.

Specification Differences

| Specification | NVIDIA GeForce GTX 670M | AMD Radeon HD 8790M |

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

| Architecture | Fermi 2.0 | GCN 1.0 |

| Chip | GF114 | Mars |

| Process Node | 40 nm | 28 nm |

| Transistors | 1,950 million | 950 million |

| Die Size | 332 mm² | 77 mm² |

| Transistor Density | 5.9M / mm² | 12.3M / mm² |

| Base Clock | Not listed | 850 MHz |

| Boost Clock | Not listed | 900 MHz |

| Memory Clock | 750 MHz (3 Gbps effective) | 1000 MHz (4 Gbps effective) |

| Memory Size | 1536 MB | 2 GB |

| Memory Type | GDDR5 | GDDR5 |

| Memory Bus Width | 192 bit | 128 bit |

| Memory Bandwidth | 72.00 GB/s | 64.00 GB/s |

| Shading Units | 336 | 384 |

| TMUs | 56 | 24 |

| ROPs | 24 | 8 |

| Pixel Rate | 8.372 GPixel/s | 7.200 GPixel/s |

| Texture Rate | 33.49 GTexel/s | 21.60 GTexel/s |

| FP32 | 803.7 GFLOPS | 691.2 GFLOPS |

| TDP | 75 W | Not listed |

| Bus Interface | MXM-B (3.0) | MXM-A (3.0) |

| DirectX | 12 (11_0) | 12 (11_1) |

| OpenGL | 4.6 | 4.6 |

| Vulkan | Not listed | 1.2.170 |

| Release Date | 2012-03-21 | 2013-03-31 |

The GTX 670M leads in bandwidth, texture rate, pixel rate, FP32, and die size. The Radeon HD 8790M leads in transistor density, shading units, memory capacity, and API features. The recorded data shows a clear compute performance winner, but the Radeon offers better modern software support. For a builder choosing between these two end-of-life mobile GPUs, the GTX 670M is the performance pick, while the Radeon HD 8790M is the compatibility pick, especially for Vulkan.