AMD Radeon RX 550X vs NVIDIA GeForce GTX 670 Comparison

Head-to-Head Benchmarks

The benchmark data shows a decisive overall victory for the NVIDIA GeForce GTX 670, which wins both recorded head-to-head tests against the AMD Radeon RX 550X. The margin is substantial in both compute and graphics API workloads, though the gap narrows considerably when moving from OpenCL to Vulkan.

In the Geekbench OpenCL test, the GTX 670 scores 15,341 points against the RX 550X's 9,662 points. That translates to a 58.8% advantage for the NVIDIA card, a massive lead that reflects the large difference in raw shading throughput between the two designs. The GTX 670's FP32 rating of 2.634 TFLOPS versus the RX 550X's 1,211.4 GFLOPS explains much of this gap, as OpenCL workloads tend to scale closely with raw compute resources. The GTX 670 also fields 1,344 shading units, 112 texture mapping units, and 32 ROPs, while the RX 550X is limited to 512 shading units, 32 TMUs, and 16 ROPs. These are not close specifications, and the benchmark results confirm that the architectural disparity translates directly into performance.

The Vulkan test narrows the gap but still favors the GTX 670. Here the NVIDIA card scores 15,553 points, while the AMD card manages 11,299 points. The delta is 37.6%, still a commanding lead but noticeably smaller than the OpenCL margin. This is interesting because the RX 550X supports a newer Vulkan version (1.3) compared to the GTX 670's 1.2.175, and the AMD card's GCN 4.0 architecture was designed with modern API overhead reduction in mind. Despite that, the GTX 670's superior raw throughput carries the day. The data suggests that the RX 550X's architectural efficiency improvements are not enough to overcome the GTX 670's roughly two-to-one advantage in shading units and texture rate.

Looking at the broader database context, the GTX 670's average benchmark score of 12,773 places it in the 52nd percentile of all GPUs tracked. Its closest rivals include the AMD Radeon Pro 455 at 12,831 (0.4% faster), the NVIDIA GeForce GTX 590 at 12,830 (0.4% faster), and the AMD FirePro W5100 at 12,847 (0.6% faster). The GTX 670 also sits just ahead of the AMD Radeon RX 7600M XT, which scores 12,710 (0.5% slower). This positions the GTX 670 as a solidly mid-pack performer even years after its release.

The RX 550X, by contrast, averages 10,481 across its recorded benchmarks, landing in the 49th percentile. Its nearest rivals are tightly clustered: the AMD Radeon RX 6600S at 10,629 (1.4% faster), the AMD Radeon R9 M275X at 10,582 (1% faster), the NVIDIA Tesla C2075 at 10,400 (0.8% slower), and the AMD Radeon RX 6500M at 10,362 (1.2% slower). The RX 550X sits in the middle of this group, indicating that its performance level is competitive with other entry-level and older mid-range parts, but it is clearly a tier below the GTX 670.

FAQ

Q: Which GPU wins the head-to-head benchmark comparison?

A: The NVIDIA GeForce GTX 670 wins both recorded tests. It leads by 58.8% in Geekbench OpenCL and by 37.6% in Geekbench Vulkan.

Q: How do the average benchmark scores compare between the two cards?

A: The GTX 670 has an average benchmark score of 12,773, while the RX 550X averages 10,481. This places the GTX 670 in the 52nd percentile of all GPUs, compared to the RX 550X's 49th percentile.

Q: Which card has higher memory bandwidth?

A: The GTX 670 has significantly higher memory bandwidth at 192.3 GB/s, using a 256-bit bus. The RX 550X offers 96.00 GB/s over a 128-bit bus. However, the RX 550X has double the memory capacity at 4 GB versus 2 GB.

Q: What are the closest rivals to each card in the database?

A: For the GTX 670, the AMD FirePro W5100 (12,847, 0.6% faster) and AMD Radeon Pro 455 (12,831, 0.4% faster) are the nearest competitors. For the RX 550X, the AMD Radeon RX 6600S (10,629, 1.4% faster) and AMD Radeon R9 M275X (10,582, 1% faster) sit closest.

Q: Which card supports newer graphics API versions?

A: The RX 550X 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.

Q: What is the power requirement difference between the two cards?

A: The GTX 670 has a TDP of 170 W and requires a 450 W suggested power supply with two 6-pin connectors. The RX 550X has a TDP of 50 W, needs no power connectors, and works with a 250 W suggested power supply.

Architecture Differences

The two GPUs come from different architectural generations and foundries, which explains much of their performance and efficiency divergence. The GTX 670 is built on NVIDIA's Kepler architecture, using the GK104 chip manufactured by TSMC on a 28 nm process. The RX 550X uses AMD's GCN 4.0 architecture, with the Lexa chip produced by GlobalFoundries on a 14 nm node. This process advantage gives the RX 550X a much higher transistor density: 21.4 million transistors per square millimeter versus 12.0 million for the GTX 670. The RX 550X packs 2,200 million transistors into a 103 mm² die, while the GTX 670 contains 3,540 million transistors across a much larger 294 mm² die.

The Kepler architecture in the GTX 670 was designed around high shader counts and wide memory interfaces. Its 1,344 shading units, 112 TMUs, and 32 ROPs deliver a pixel rate of 27.44 GPixel/s and a texture rate of 109.8 GTexel/s. The GCN 4.0 architecture in the RX 550X takes a different approach, emphasizing efficiency and modern API features. Its 512 shading units, 32 TMUs, and 16 ROPs produce a pixel rate of 18.93 GPixel/s and a texture rate of 37.86 GTexel/s. Notably, the RX 550X supports FP16 compute at a 1:1 ratio with FP32 (1,211.4 GFLOPS for both), while the GTX 670 has no recorded FP16 capability. This makes the RX 550X potentially more versatile for workloads that utilize half-precision arithmetic.

The RX 550X also brings a more modern memory architecture, despite its narrower bus. It supports 4 GB of GDDR5 across a 128-bit interface, while the GTX 670 uses 2 GB across 256 bits. The RX 550X's memory runs at 1500 MHz (6 Gbps effective), slightly lower than the GTX 670's 1502 MHz (6 Gbps effective), but the bandwidth difference comes entirely from the bus width. The GTX 670's 192.3 GB/s is exactly double the RX 550X's 96.00 GB/s, reflecting the 256-bit versus 128-bit interface.

In terms of connectivity, the RX 550X is newer and supports more modern display standards, including HDMI 2.0b and DisplayPort 1.4a. The GTX 670 offers HDMI 1.4a and DisplayPort 1.2, along with two DVI outputs. The RX 550X also uses a PCIe 3.0 x8 interface, while the GTX 670 uses PCIe 3.0 x16.

Specification Differences

The two cards differ across almost every major specification category. Process node: the GTX 670 uses 28 nm from TSMC, while the RX 550X uses 14 nm from GlobalFoundries. Transistor count: 3,540 million for the GTX 670 versus 2,200 million for the RX 550X. Die size: 294 mm² versus 103 mm². Transistor density: 12.0M per mm² versus 21.4M per mm².

Clock speeds favor the AMD card. The RX 550X has a base clock of 1100 MHz and a boost clock of 1183 MHz, compared to the GTX 670's 915 MHz base and 980 MHz boost. Memory clocks are nearly identical: 1502 MHz (6 Gbps effective) for the GTX 670 and 1500 MHz (6 Gbps effective) for the RX 550X.

Memory configuration differs in capacity and bus width. The GTX 670 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s bandwidth. The RX 550X has 4 GB of GDDR5 on a 128-bit bus with 96.00 GB/s bandwidth.

Compute resources heavily favor the GTX 670: 1,344 shading units versus 512, 112 TMUs versus 32, and 32 ROPs versus 16. Pixel rate is 27.44 GPixel/s versus 18.93 GPixel/s. Texture rate is 109.8 GTexel/s versus 37.86 GTexel/s. FP32 output is 2.634 TFLOPS versus 1,211.4 GFLOPS. The RX 550X adds FP16 support at 1,211.4 GFLOPS, which the GTX 670 lacks.

Power and cooling: the GTX 670 has a 170 W TDP, requires two 6-pin power connectors, and needs a 450 W suggested power supply. The RX 550X has a 50 W TDP, requires no power connectors, and needs only a 250 W suggested power supply. Both are dual-slot cards, but the GTX 670 is much larger at 241 mm (9.5 inches) long, 111 mm (4.4 inches) high, and 38 mm (1.5 inches) wide. The RX 550X measures just 145 mm (5.7 inches) in length, with no recorded height or width.

API support: the GTX 670 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The RX 550X supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. Bus interface is PCIe 3.0 x16 for the GTX 670 and PCIe 3.0 x8 for the RX 550X.

Where Each One Wins

The GTX 670 is the clear performance winner in every recorded test. Its strengths lie in raw compute throughput, memory bandwidth, and texture fill rate. The 58.8% OpenCL advantage and 37.6% Vulkan advantage make it the better choice for any application that depends on shader performance or high-bandwidth memory access. The 256-bit memory bus and 192.3 GB/s bandwidth give it a significant edge in bandwidth-sensitive workloads, while its 112 TMUs and 109.8 GTexel/s texture rate handle high-resolution texture filtering with ease. The GTX 670's 52nd percentile ranking and average score of 12,773 place it in the middle of the GPU performance distribution, making it a viable option for 1080p gaming and general compute tasks even at end-of-life.

The RX 550X wins on efficiency and modern features. Its 50 W TDP is less than a third of the GTX 670's 170 W, and it requires no external power connectors. This makes it suitable for small form factor systems or upgrades where power delivery is limited. The 4 GB memory capacity is double the GTX 670's 2 GB, which can be advantageous in modern games or applications that exceed 2 GB of video memory usage, even if the overall bandwidth is lower. The RX 550X also supports newer display outputs (HDMI 2.0b, DisplayPort 1.4a), DirectX 12 (12_0), and Vulkan 1.3, giving it better forward compatibility with current software and display technology. Its 49th percentile ranking and average score of 10,481 indicate it is only slightly below the GTX 670 in overall standing, despite the large head-to-head deltas.

For workloads that prioritize FP16 compute, the RX 550X's 1:1 FP16 capability is a distinct advantage. The GTX 670 has no recorded FP16 performance, so any application leveraging half-precision arithmetic will run more efficiently on the AMD card. The RX 550X's smaller size (145 mm versus 241 mm) also makes it easier to fit into compact cases.

The GTX 670 is the choice for maximum performance in legacy applications and any workload that can utilize its high shading throughput and memory bandwidth. The RX 550X is the choice for energy-conscious builds, modern API support, and scenarios where 4 GB of memory is necessary. The data does not show any test where the RX 550X outperforms the GTX 670, so the decision ultimately comes down to whether the AMD card's efficiency and feature set outweigh the NVIDIA card's substantial performance lead.