AMD Radeon RX 570X vs NVIDIA GeForce GTX 670 Comparison

The AMD Radeon RX 570X and the NVIDIA GeForce GTX 670 are separated by six years of architecture evolution, and the benchmark data reflects a decisive generational shift. The RX 570X holds a commanding lead in the only directly comparable compute test, while the GTX 670 relies on its older Kepler design’s strengths in different API workloads. Below is a breakdown of their measured performance, architectural divergence, and specification-level differences.

Head-to-Head Benchmarks

The single head-to-head benchmark available—Geekbench OpenCL—paints a stark picture. The AMD Radeon RX 570X scores 38,939 points, while the NVIDIA GeForce GTX 670 manages 15,341 points. This translates to a deltaPct of -60.6%, meaning the RX 570X outperforms the GTX 670 by roughly 154% in this compute-oriented test. The margin is enormous, and it underscores how much raw parallel throughput has advanced between 2012 and 2018.

In other benchmark categories, the two cards do not share direct comparisons, but their overall averages tell a similar story. The RX 570X posts an average benchmark score of 13,871, which places it in the 55th percentile of all GPUs. The GTX 670’s average is 12,773, sitting at the 52nd percentile. That 1,098-point gap in average score is modest compared to the OpenCL blowout, suggesting the GTX 670 remains competitive in certain legacy workloads where its driver optimizations or specific architecture traits shine.

Looking at the GTX 670’s individual results, its Geekbench Metal score of 7,424 and Geekbench Vulkan score of 15,553 indicate that it can still produce respectable numbers in API-specific tests. However, the RX 570X does not have corresponding Metal or Vulkan entries in the data, so a direct comparison in those APIs is impossible. The OpenCL result, though, is the clearest available signal: the RX 570X is in a different performance class for general-purpose compute.

The nearest rivals for each card further contextualize their standing. The GTX 670’s closest competitor is the AMD FirePro W5100, with an average score of 12,847 and a deltaPct of -0.6, meaning the GTX 670 trails it by less than 1%. The RX 570X, meanwhile, sits just 0.4% behind the NVIDIA RTX A2000 Mobile (13,821) and 0.4% ahead of the AMD Radeon 660M (13,812). These tight margins indicate that both cards are well-positioned within their respective performance tiers, despite the massive gap between them.

Where Each One Wins

The RX 570X dominates in raw compute performance, as evidenced by its OpenCL score. This makes it the clear choice for tasks that leverage general-purpose GPU computing, such as rendering, scientific simulations, or any workload that can utilize OpenCL acceleration. Its 5.095 TFLOPS of FP32 performance and 5.095 TFLOPS of FP16 (1:1) further reinforce this advantage, providing nearly double the floating-point throughput of the GTX 670’s 2.634 TFLOPS.

The GTX 670, by contrast, shows relative strength in its average benchmark score relative to its OpenCL showing. While its OpenCL score is only 39.4% of the RX 570X’s, its average score is 92.1% of the RX 570X’s average. This suggests that in non-OpenCL workloads—particularly those relying on Metal or Vulkan, where the GTX 670 scores 7,424 and 15,553 respectively—the older card narrows the gap considerably. For users running legacy applications or specific games that favor Kepler’s architecture, the GTX 670 may still deliver acceptable performance despite its age.

In terms of memory, the RX 570X offers 8 GB of GDDR5, which is four times the GTX 670’s 2 GB. This capacity advantage is critical for modern games and applications with large texture sets, though the GTX 670’s 192.3 GB/s bandwidth is not drastically lower than the RX 570X’s 224.0 GB/s. The RX 570X also has a higher memory clock at 7 Gbps effective versus 6 Gbps effective, further widening the performance gap in memory-bound scenarios.

Architecture Differences

The architectural divide between these two GPUs is fundamental. The GTX 670 uses NVIDIA’s Kepler architecture on a 28 nm process node from TSMC, with a chip designated GK104. It packs 3,540 million transistors into a 294 mm² die, yielding a transistor density of 12.0M per mm². Kepler was designed for efficiency and high clock speeds, which explains its relatively modest core count of 1,344 shading units, 112 TMUs, and 32 ROPs.

The RX 570X employs AMD’s GCN 4.0 architecture on a 14 nm process node from GlobalFoundries, with the Polaris 20 chip. This newer process allows AMD to fit 5,700 million transistors into a smaller 232 mm² die, achieving a transistor density of 24.6M per mm²—more than double the GTX 670’s density. The RX 570X scales up to 2,048 shading units, 128 TMUs, and 32 ROPs, providing substantially more parallel execution resources.

These architectural differences manifest directly in computed rates. The RX 570X achieves a pixel rate of 39.81 GPixel/s and a texture rate of 159.2 GTexel/s, compared to the GTX 670’s 27.44 GPixel/s and 109.8 GTexel/s. The RX 570X’s FP16 support at a 1:1 ratio with FP32 is another significant divergence, as the GTX 670 has no FP16 capability listed. This makes the RX 570X more suitable for workloads that benefit from reduced-precision arithmetic.

API support also differs. The RX 570X supports DirectX 12 (12_0), Vulkan 1.3, and OpenGL 4.6, while the GTX 670 is limited to DirectX 12 (11_0), Vulkan 1.2.175, and OpenGL 4.6. The newer Vulkan version and full DirectX 12 feature level give the RX 570X better compatibility with modern titles and compute APIs.

Specification Differences

The two cards differ across nearly every major specification category, reflecting their generational gap. The most striking difference is memory capacity: the RX 570X offers 8 GB of GDDR5, while the GTX 670 provides only 2 GB. Both use a 256-bit bus, but the RX 570X’s higher memory clock (7 Gbps effective vs. 6 Gbps effective) results in 224.0 GB/s bandwidth versus 192.3 GB/s.

Clock speeds favor the RX 570X as well, with a base clock of 1168 MHz and boost of 1244 MHz, compared to the GTX 670’s 915 MHz base and 980 MHz boost. This higher clock, combined with more shading units, gives the RX 570X its substantial compute advantage. Power consumption, however, tells a different story: the RX 570X has a TDP of 150 W, which is lower than the GTX 670’s 170 W, despite delivering significantly more performance. The RX 570X also requires only a single 6-pin power connector, while the GTX 670 needs two 6-pin connectors.

Display outputs differ as well. The GTX 670 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the RX 570X provides 1x DVI, 1x HDMI 2.0b, and 3x DisplayPort 1.4a. The newer HDMI and DisplayPort standards on the RX 570X support higher resolutions and refresh rates. Both cards are dual-slot designs with a length of 241 mm, and both use a PCIe 3.0 x16 interface. The GTX 670’s dimensions are fully specified at 111 mm height and 38 mm width, while the RX 570X only lists length. The GTX 670 was released on May 9, 2012, with a launch MSRP of 399 USD; the RX 570X launched on April 10, 2018, with no MSRP listed. Both are end-of-life products, with the GTX 670 part of the GeForce 600 generation and the RX 570X part of the Polaris (RX 500X) generation.

FAQ

Q: Which GPU has better raw compute performance?

A: The AMD Radeon RX 570X is decisively ahead, scoring 38,939 in Geekbench OpenCL versus the GTX 670’s 15,341, a deltaPct of -60.6% in favor of the RX 570X.

Q: How do their memory specifications compare?

A: The RX 570X offers 8 GB of GDDR5 with 224.0 GB/s bandwidth, while the GTX 670 has 2 GB of GDDR5 with 192.3 GB/s bandwidth. Both use a 256-bit bus, but the RX 570X’s memory runs at 7 Gbps effective versus 6 Gbps effective.

Q: What is the transistor and die size difference?

A: The RX 570X packs 5,700 million transistors into a 232 mm² die using a 14 nm process, while the GTX 670 has 3,540 million transistors on a 294 mm² die using a 28 nm process. This gives the RX 570X a transistor density of 24.6M per mm² versus 12.0M per mm².

Q: Which card consumes less power?

A: The RX 570X has a lower TDP of 150 W compared to the GTX 670’s 170 W, and it requires only a single 6-pin power connector versus the GTX 670’s two 6-pin connectors.

Q: How do their API supports differ?

A: The RX 570X supports DirectX 12 (12_0) and Vulkan 1.3, while the GTX 670 is limited to DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

Q: What are their average benchmark scores and percentiles?

A: The RX 570X has an average benchmark score of 13,871, placing it in the 55th percentile of all GPUs. The GTX 670 averages 12,773, ranking in the 52nd percentile.