AMD Radeon RX 570X vs NVIDIA GeForce GTX 660 Ti Comparison

The benchmark data places the NVIDIA GeForce GTX 660 Ti and the AMD Radeon RX 570X in adjacent performance tiers, yet their underlying specifications and single available head-to-head result tell a story of generational divergence. The AMD Radeon RX 570X holds the only direct comparison win, but the GTX 660 Ti’s average benchmark score is actually higher, creating a nuanced picture that requires careful parsing of the numbers.

Head-to-Head Benchmarks

The single head-to-head benchmark result is decisive in one direction. In the Geekbench OpenCL test, the AMD Radeon RX 570X scores 38,939 points, while the NVIDIA GeForce GTX 660 Ti scores 15,113 points. This produces a deltaPct of -61.2% from the AMD card’s perspective, meaning the RX 570X outperforms the GTX 660 Ti by a substantial margin in this compute-oriented workload. The raw score difference is 23,826 points, which is more than 2.5 times the NVIDIA card’s total output.

However, the aggregate benchmark picture complicates this narrative. The GTX 660 Ti’s average benchmark score across all tests is 15,063, derived from its two Geekbench results (OpenCL and Vulkan). The RX 570X’s average is 13,871, which includes three tests: Geekbench OpenCL (38,939), Passmark G2D (750), and Passmark G3D (1,923). The NVIDIA card’s average is 8.6% higher, despite losing the OpenCL head-to-head badly. This discrepancy suggests the RX 570X’s performance is highly workload-dependent, excelling in OpenCL but lagging in other metrics like the Passmark G2D score of 750.

Looking at the nearest rivals for each card provides additional context. The GTX 660 Ti sits at the 57th percentile of all GPUs, with its closest competitor being the AMD Radeon Pro 560X at an average score of 15,082, a delta of -0.1%. The RX 570X holds the 55th percentile, with the NVIDIA RTX A2000 Mobile as its nearest rival at 13,821 (deltaPct 0.4%). Notably, the RX 570X’s rival list includes the AMD Radeon RX 7900 XT at 13,745, which is 0.9% behind, indicating that the RX 570X’s average score places it in a mixed performance neighborhood where high-end and low-end products converge.

FAQ

Q: Which card wins the only direct benchmark comparison?

A: The AMD Radeon RX 570X wins the Geekbench OpenCL test with a score of 38,939 against the NVIDIA GeForce GTX 660 Ti’s 15,113, a deltaPct of -61.2% favoring the AMD card.

Q: Does the GTX 660 Ti have a higher average benchmark score?

A: Yes. The GTX 660 Ti averages 15,063 across its benchmarks, while the RX 570X averages 13,871, making the NVIDIA card approximately 8.6% higher on aggregate.

Q: What is the memory configuration difference between the two cards?

A: The GTX 660 Ti has 2 GB of GDDR5 memory on a 192-bit bus with 144.2 GB/s bandwidth, while the RX 570X has 8 GB of GDDR5 on a 256-bit bus with 224.0 GB/s bandwidth.

Q: How do their compute capabilities compare in FP32?

A: The RX 570X delivers 5.095 TFLOPS of FP32 performance, whereas the GTX 660 Ti provides 2.634 TFLOPS, meaning the AMD card has roughly 93% more raw single-precision compute throughput.

Q: Which card is more recent in its release cycle?

A: The GTX 660 Ti was released on 2012-08-15, while the RX 570X came later on 2018-04-10, a gap of approximately five years and eight months.

Q: Do both cards use the same power supply recommendation?

A: Yes, both cards list a suggested PSU of 450 W, and both have a TDP of 150 W, though the GTX 660 Ti requires two 6-pin power connectors while the RX 570X needs only one.

Architecture Differences

The architectural divide between these two GPUs is stark. The GTX 660 Ti uses NVIDIA’s Kepler architecture on the GK104 chip, fabricated on TSMC’s 28 nm process node. This chip contains 3,540 million transistors on a 294 mm² die, yielding a transistor density of 12.0M per mm². The RX 570X, by contrast, employs AMD’s GCN 4.0 architecture on the Polaris 20 chip, built on GlobalFoundries’ 14 nm process. This newer node packs 5,700 million transistors into a smaller 232 mm² die, achieving a much higher transistor density of 24.6M per mm².

Core counts differ substantially. The GTX 660 Ti has 1,344 shading units, 112 texture mapping units, and 24 ROPs. The RX 570X scales this up to 2,048 shading units, 128 TMUs, and 32 ROPs. This 52.4% advantage in shading units and corresponding increases in TMUs and ROPs explain much of the RX 570X’s compute superiority. The RX 570X also supports FP16 compute at a 1:1 ratio with FP32 (both 5.095 TFLOPS), whereas the GTX 660 Ti has no FP16 capability listed.

API support also reflects the generational gap. The RX 570X supports DirectX 12 (12_0) and Vulkan 1.3, while the GTX 660 Ti is limited to DirectX 12 (11_0) and Vulkan 1.2.175. Both cards support OpenGL 4.6. Display outputs differ as well: the RX 570X offers one DVI, one HDMI 2.0b, and three DisplayPort 1.4a connections, while the GTX 660 Ti provides two DVI, one HDMI 1.4a, and one DisplayPort 1.2.

Specification Differences

The two cards diverge on nearly every measurable specification. Clock speeds show the RX 570X running at 1168 MHz base and 1244 MHz boost, compared to the GTX 660 Ti’s 915 MHz base and 980 MHz boost. Memory clocks also favor AMD: 1750 MHz (7 Gbps effective) versus 1502 MHz (6 Gbps effective). The RX 570X’s memory subsystem is wider and faster, with 8 GB on a 256-bit bus delivering 224.0 GB/s, versus 2 GB on a 192-bit bus at 144.2 GB/s.

Pixel and texture fill rates follow the same pattern. The RX 570X achieves 39.81 GPixel/s and 159.2 GTexel/s, while the GTX 660 Ti manages 27.44 GPixel/s and 109.8 GTexel/s. FP32 throughput is nearly double on the AMD card: 5.095 TFLOPS versus 2.634 TFLOPS. The RX 570X also lists FP16 at 5.095 TFLOPS (1:1), which the GTX 660 Ti lacks entirely.

Physical and power characteristics are similar but not identical. Both cards are dual-slot, 241 mm (9.5 inches) long, with a 150 W TDP and a suggested 450 W PSU. The GTX 660 Ti requires two 6-pin power connectors, whereas the RX 570X needs only one. The GTX 660 Ti carries a launch MSRP of 299 USD, while the RX 570X has no launch MSRP listed. The GTX 660 Ti is part of the GeForce 600 generation with a GeForce 700 successor, while the RX 570X belongs to the Polaris (RX 500X) generation with a Vega successor.

Where Each One Wins

The RX 570X wins decisively in raw compute and memory-bound workloads. Its Geekbench OpenCL score of 38,939 is more than double the GTX 660 Ti’s 15,113, making it the clear choice for OpenCL acceleration tasks. The AMD card also wins on memory capacity and bandwidth, offering four times the VRAM (8 GB versus 2 GB) and 55.3% more bandwidth (224.0 GB/s versus 144.2 GB/s). Its higher pixel rate (39.81 GPixel/s versus 27.44 GPixel/s) and texture rate (159.2 GTexel/s versus 109.8 GTexel/s) suggest advantages in fill-rate-limited scenarios.

The GTX 660 Ti wins on aggregate benchmark consistency. Its average score of 15,063 exceeds the RX 570X’s 13,871, and its nearest rivals include the Radeon Pro 560X (-0.1%) and RTX 3050 OEM (-0.9%), indicating it competes with a newer generation of mid-range parts. The NVIDIA card also achieves a higher percentile ranking at 57 versus the RX 570X’s 55. In Vulkan, the GTX 660 Ti scores 15,012, which is close to its OpenCL result, suggesting more uniform performance across different API workloads. The RX 570X’s Passmark G2D score of 750 and G3D score of 1,923 pull its average down, indicating weaker performance in those specific tests.

The Verdict

The data presents a clear trade-off. The AMD Radeon RX 570X is the superior choice for users prioritizing raw compute throughput, large memory buffers, and modern API support. Its 5.095 TFLOPS FP32 performance, 8 GB VRAM, and Vulkan 1.3 compatibility make it the more future-proof option for compute-heavy applications. The RX 570X’s single head-to-head win with a -61.2% deltaPct over the GTX 660 Ti in OpenCL is unambiguous evidence of its compute dominance.

The NVIDIA GeForce GTX 660 Ti appeals to those who value consistent performance across a broader range of benchmarks. Its higher average score (15,063 versus 13,871) and better percentile ranking (57 versus 55) suggest that in mixed workloads, it may hold its own better than the RX 570X’s uneven profile. The GTX 660 Ti also offers dual-link DVI outputs, which the RX 570X lacks, potentially mattering for legacy display setups.

For a buyer choosing strictly from this data, the decision hinges on workload type. Compute-heavy tasks with large datasets favor the RX 570X overwhelmingly. General-purpose gaming or varied workloads with less VRAM pressure might see the GTX 660 Ti perform more consistently, as its Vulkan score of 15,012 nearly matches its OpenCL score, whereas the RX 570X’s Passmark G3D score of 1,923 represents a relatively weak showing. The RX 570X’s lower power connector requirement (one 6-pin versus two) and newer architecture make it the more modern engineering choice, but the GTX 660 Ti’s higher aggregate score cannot be ignored. Ultimately, the RX 570X wins on raw capability, while the GTX 660 Ti wins on benchmark balance.