AMD Radeon RX 5500M vs NVIDIA GeForce GTX 670 Comparison

The AMD Radeon RX 5500M and the NVIDIA GeForce GTX 670 represent two distinct eras of GPU design, separated by seven years of architectural evolution. The benchmark data shows a decisive victory for the AMD part across all shared tests, with the RX 5500M delivering a 578.3% advantage in Geekbench Metal, a 152.4% lead in Geekbench OpenCL, and a 129.5% edge in Geekbench Vulkan. While the GTX 670 was a high-end desktop part in its day, the RX 5500M's modern RDNA architecture and process node advantages render the comparison lopsided, though the older card retains specific characteristics that matter in legacy scenarios.

Head-to-Head Benchmarks

The most striking result is in Geekbench Metal, where the RX 5500M scores 50,359 against the GTX 670's 7,424. That 578.3% delta is not incremental; it reflects a generational shift in compute throughput and API efficiency. The AMD card's Metal performance is nearly seven times higher, which positions it as a viable option for modern macOS-style workloads, whereas the GTX 670's score indicates it was never designed for such APIs.

In Geekbench OpenCL, the RX 5500M posts 38,725 versus 15,341 for the GTX 670, a 152.4% advantage. This gap is substantial but narrower than Metal, suggesting the older Kepler architecture still has some general-purpose compute capability. The RX 5500M's FP32 throughput of 4.632 TFLOPS versus 2.634 TFLOPS for the GTX 670 explains much of this difference, though the AMD card's higher memory bandwidth (224.0 GB/s vs 192.3 GB/s) also contributes to its OpenCL dominance.

The Geekbench Vulkan result shows 35,693 for the RX 5500M against 15,553 for the GTX 670, a 129.5% lead. Vulkan support is listed for both cards, but the RX 5500M runs Vulkan 1.4 while the GTX 670 is capped at 1.2.175. The newer API version, combined with RDNA 1.0's optimized front-end, allows the AMD card to extract more performance from the same workload. Interestingly, the GTX 670's Vulkan score is nearly identical to its OpenCL score (15,553 vs 15,341), indicating that Kepler's compute units handle both APIs with similar efficiency, whereas the RX 5500M shows a clear preference for Metal over Vulkan.

Across all three head-to-head tests, the RX 5500M wins 3-0. The average benchmark score reinforces this: the AMD card averages 13,356 across its full benchmark suite, while the GTX 670 averages 12,773. The RX 5500M sits at the 54th percentile of all GPUs, slightly ahead of the GTX 670's 52nd percentile. In terms of nearest rivals, the RX 5500M's closest competitor is the AMD FirePro M6100 with a 0% delta, while the GTX 670's nearest rival is the AMD Radeon Pro 455 at -0.4%, placing both cards in a similar mid-pack performance tier despite the massive head-to-head deltas.

FAQ

Q: Why does the RX 5500M have such a massive lead in Geekbench Metal?

A: The RX 5500M scores 50,359 in Metal versus 7,424 for the GTX 670, a 578.3% delta. This is partly due to the AMD card's newer RDNA 1.0 architecture, which was designed with modern API efficiency in mind, and its 7 nm process node that allows higher clock speeds (1375 MHz base, 1645 MHz boost) compared to the GTX 670's 915 MHz base and 980 MHz boost.

Q: Which card has better raw compute performance?

A: The RX 5500M delivers 4.632 TFLOPS of FP32 compute, while the GTX 670 manages 2.634 TFLOPS. This 75.9% advantage in raw throughput is reflected in the OpenCL benchmark, where the RX 5500M leads 38,725 to 15,341.

Q: Is the GTX 670 still competitive in any benchmark?

A: The data shows no benchmark where the GTX 670 wins. However, its PassMark scores for DirectX 9 (100) and DirectX 10 (39) suggest it retains some legacy API performance, though no direct head-to-head comparison exists for those tests in this data set.

Q: How do the cards compare in memory bandwidth?

A: The RX 5500M has 224.0 GB/s of bandwidth from its 4 GB GDDR6 memory on a 128-bit bus, while the GTX 670 has 192.3 GB/s from 2 GB GDDR5 on a 256-bit bus. Despite the narrower bus, the AMD card's faster GDDR6 memory (14 Gbps effective vs 6 Gbps effective) gives it a 16.5% bandwidth advantage.

Q: What explains the difference in average benchmark scores?

A: The RX 5500M averages 13,356 across all its benchmarks, while the GTX 670 averages 12,773. The AMD card's percentile rank of 54 versus 52 for NVIDIA indicates both are near the middle of the performance distribution, but the RX 5500M's newer architecture yields higher scores in modern API tests.

Q: Which card has better API support?

A: The RX 5500M supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 670 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The AMD card's higher DirectX feature level and newer Vulkan version give it better compatibility with current software.

Where Each One Wins

The AMD Radeon RX 5500M wins in every measurable category from the head-to-head benchmarks. Its 578.3% Metal advantage makes it the clear choice for any workload leveraging Apple's graphics API, which is critical for mobile content creation on macOS-based laptops. The 152.4% OpenCL lead positions it well for general-purpose GPU computing, including video encoding and scientific simulations that use OpenCL as their backend. The 129.5% Vulkan advantage covers modern cross-platform games and applications that prioritize Vulkan for its low overhead. Its 4 GB of GDDR6 memory also provides more capacity than the GTX 670's 2 GB, which matters for texture-heavy scenes at higher resolutions.

The NVIDIA GeForce GTX 670 has no benchmark wins in this comparison, but its specification sheet reveals niche strengths. Its 256-bit memory bus, though paired with slower GDDR5, still provides a wider path for data movement, and its 112 texture mapping units exceed the RX 5500M's 88 TMUs. The GTX 670's dual-slot design and 2x 6-pin power connectors indicate it was built for desktop expansion, whereas the RX 5500M is portable-device dependent with no power connectors. For legacy DirectX 9 and DirectX 10 workloads, the GTX 670's PassMark scores of 100 and 39, respectively, suggest it retains functional capability, even if those tests are not part of the head-to-head data.

Specification Differences

The two cards differ fundamentally in nearly every specification. The RX 5500M uses a 7 nm process from TSMC, while the GTX 670 uses 28 nm from the same foundry. This translates to a transistor density of 40.5M / mm² for AMD versus 12.0M / mm² for NVIDIA, despite the RX 5500M having 6,400 million transistors on a 158 mm² die versus 3,540 million on a 294 mm² die.

Memory configurations diverge sharply: the RX 5500M has 4 GB of GDDR6 on a 128-bit bus with 224.0 GB/s bandwidth, while the GTX 670 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s. Clock speeds favor AMD significantly: 1375 MHz base and 1645 MHz boost versus 915 MHz base and 980 MHz boost. The RX 5500M also has a game clock of 1448 MHz, a feature the GTX 670 lacks entirely.

Power requirements are inverted: the RX 5500M draws 85 W TDP with no power connectors, while the GTX 670 draws 170 W TDP with 2x 6-pin connectors and a suggested 450 W PSU. The RX 5500M uses PCIe 4.0 x8, whereas the GTX 670 uses PCIe 3.0 x16. Display outputs differ as well: the RX 5500M is portable-device dependent, while the GTX 670 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2.

Architecture Differences

The RX 5500M is built on RDNA 1.0 architecture, part of the Radeon RX 5000 series, using the Navi 14 chip. It features 1408 shading units, 88 TMUs, and 32 ROPs, with a pixel rate of 52.64 GPixel/s and texture rate of 144.8 GTexel/s. Its FP16 performance is 9.265 TFLOPS with a 2:1 ratio to FP32, enabling mixed-precision workloads. The architecture supports DirectX 12 (12_1), the highest feature level in this comparison.

The GTX 670 is based on Kepler architecture, part of the GeForce 600 generation, using the GK104 chip. It has 1344 shading units, 112 TMUs, and 32 ROPs, with a pixel rate of 27.44 GPixel/s and texture rate of 109.8 GTexel/s. Its FP32 performance is 2.634 TFLOPS, and it has no listed FP16 capability. The Kepler architecture supports DirectX 12 (11_0), a lower feature level that lacks some modern rendering features like conservative rasterization.

The RX 5500M's RDNA 1.0 architecture was designed for efficiency and modern API utilization, while Kepler predates many of these optimizations. The AMD card's 7 nm process enables higher clocks and lower power simultaneously, whereas the GTX 670's 28 nm process limits both. The RX 5500M's transistor density of 40.5M / mm² versus 12.0M / mm² for the GTX 670 highlights the architectural leap, though the GTX 670's larger die area (294 mm² vs 158 mm²) shows it was a physically bigger chip. The RX 5500M's generation is Navi Mobile (RX 5000M), with a predecessor of Polaris Mobile, while the GTX 670's predecessor is GeForce 500 and successor is GeForce 700, placing the two cards at opposite ends of their respective product lifecycles.