AMD Radeon RX 9070 XT vs NVIDIA GeForce GTX 660 Ti Comparison

NVIDIA GeForce GTX 660 Ti and AMD Radeon RX 9070 XT sit at opposite ends of the hardware timeline, with the former launching in August 2015 as a Kepler-based part and the latter arriving in March 2025 on RDNA 4.0. The benchmark database records two shared tests for these cards, and in both, the RX 9070 XT dominates, but the comparison is far more nuanced than a simple score gap. The GTX 660 Ti holds a 57th percentile ranking among all GPUs, while the RX 9070 XT sits at the 55th percentile, a quirk driven by the different benchmark suites each card runs and the average score calculations that include tests the older card never executed. This analysis breaks down where each card wins, what separates their architectures, and how the recorded data informs a practical choice.

Where Each One Wins

The RX 9070 XT wins every head-to-head benchmark recorded in the database. In Geekbench OpenCL, it scores 17,428 against the GTX 660 Ti's 15,113, a 13.3% lead. In Geekbench Vulkan, the margin is enormous: 65,879 versus 15,012, a 77.2% advantage. The Vulkan result is the standout, reflecting not just raw compute but modern API efficiency, as the RX 9070 XT supports Vulkan 1.4 while the GTX 660 Ti is limited to Vulkan 1.2.175.

The GTX 660 Ti has no wins in the shared benchmark set. Its strengths lie elsewhere, in legacy DirectX workloads and its lighter power footprint, but the database does not record a single test where it outpaces the RX 9070 XT. For users running Vulkan-based games or compute workloads, the RX 9070 XT is the only rational pick. For those stuck on older DirectX 9 or 10 titles, the RX 9070 XT still posts Passmark scores of 353 in DirectX 9 and 154 in DirectX 10, but those numbers are not directly comparable to the GTX 660 Ti because the older card lacks Passmark entries in the database.

Architecture Differences

The architectural gap is generational. The GTX 660 Ti uses the GK104 chip on a 28 nm TSMC process, packing 3,540 million transistors into a 294 mm² die with a transistor density of 12.0M per mm². The RX 9070 XT uses Navi 48 on a 4 nm TSMC process, with 53,900 million transistors in a 357 mm² die, yielding a density of 151.0M per mm². That is a 12.6x density increase, enabling far more compute resources in a similar physical footprint.

The GTX 660 Ti has 1,344 shading units, 112 texture mapping units, and 24 raster operation pipelines. The RX 9070 XT scales to 4,096 shading units, 256 TMUs, and 128 ROPs. The RX 9070 XT also introduces 64 ray tracing cores, a feature the GTX 660 Ti lacks entirely. Clock speeds reflect the process advantage: the GTX 660 Ti runs at 915 MHz base and 980 MHz boost, while the RX 9070 XT runs at 1,660 MHz base, 2,400 MHz game clock, and 2,970 MHz boost. The FP32 throughput tells the story: 2.634 TFLOPS for the GTX 660 Ti versus 48.66 TFLOPS for the RX 9070 XT, an 18.5x gap. The RX 9070 XT also matches FP16 at 48.66 TFLOPS with 1:1 ratio, while the GTX 660 Ti has no recorded FP16 capability.

Memory architecture diverges completely. The GTX 660 Ti uses 2 GB of GDDR5 on a 192-bit bus with 144.2 GB/s bandwidth. The RX 9070 XT uses 16 GB of GDDR6 on a 256-bit bus with 644.6 GB/s bandwidth, a 4.5x bandwidth increase and 8x capacity increase. The RX 9070 XT supports PCIe 5.0 x16, while the GTX 660 Ti is limited to PCIe 3.0 x16. Display outputs also differ: the GTX 660 Ti offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2; the RX 9070 XT offers 1x HDMI 2.1b and 3x DisplayPort 2.1a, meaning modern high-refresh displays are only fully supported on the newer card.

Head-to-Head Benchmarks

The Geekbench OpenCL result is the closest contest. The GTX 660 Ti scores 15,113, and the RX 9070 XT scores 17,428, a 13.3% delta. This test exercises general compute workloads, and while the RX 9070 XT wins, the margin is modest compared to other tests. That suggests OpenCL performance scales with raw FP32 and memory bandwidth, but the older Kepler architecture still holds its own in some compute patterns, likely due to its simpler scheduling and lower overhead.

The Geekbench Vulkan result is a rout. The RX 9070 XT scores 65,879 against 15,012, a 77.2% lead. Vulkan is a low-level API that rewards modern hardware with explicit control over queues, memory, and synchronization. The RX 9070 XT's RDNA 4.0 architecture, with its dedicated ray tracing cores and 1.4 Vulkan support, simply overwhelms the GTX 660 Ti's Kepler design, which was never built for such workloads. For any Vulkan-based game, the RX 9070 XT delivers roughly four times the performance, which translates to playable frame rates at high settings versus slideshow territory on the GTX 660 Ti.

The database also records RX 9070 XT scores in other suites, though no GTX 660 Ti equivalents exist. Passmark G3D gives the RX 9070 XT 26,795, and Passmark GPU Compute gives 15,864. In 3DMark Steel Nomad DX12, it scores 7,260. These numbers place the RX 9070 XT well above its nearest rivals in the database, which include AMD Radeon HD 7770M (13,536 average, 0.1% delta) and NVIDIA GeForce GTX 570 (13,515 average, 0.2% delta). That proximity is misleading, as those rivals are ancient parts, but it reflects the RX 9070 XT's 55th percentile ranking, which is dragged down by the inclusion of Passmark DirectX 9, 10, and 12 tests where modern cards often score low due to emulation overhead.

Specification Differences

The two cards differ in nearly every measurable specification. Process node: 28 nm versus 4 nm. Transistors: 3,540 million versus 53,900 million. Die size: 294 mm² versus 357 mm². Transistor density: 12.0M versus 151.0M per mm². Base clock: 915 MHz versus 1,660 MHz. Boost clock: 980 MHz versus 2,970 MHz. Memory size: 2 GB versus 16 GB. Memory type: GDDR5 versus GDDR6. Bus width: 192-bit versus 256-bit. Bandwidth: 144.2 GB/s versus 644.6 GB/s. Shading units: 1,344 versus 4,096. TMUs: 112 versus 256. ROPs: 24 versus 128. Ray tracing cores: none versus 64. Pixel rate: 27.44 GPixel/s versus 380.2 GPixel/s. Texture rate: 109.8 GTexel/s versus 760.3 GTexel/s. FP32: 2.634 TFLOPS versus 48.66 TFLOPS. TDP: 150 W versus 304 W. Power connectors: 2x 6-pin versus 2x 8-pin. Suggested PSU: 450 W versus 700 W. Bus interface: PCIe 3.0 x16 versus PCIe 5.0 x16. DirectX support: 12 (11_0) versus 12 Ultimate (12_2). Vulkan support: 1.2.175 versus 1.4. The GTX 660 Ti is end-of-life, while the RX 9070 XT is active. The RX 9070 XT's launch MSRP was 599 USD, and the GTX 660 Ti's launch MSRP was 299 USD, though those figures are historical data points, not current market indicators.

FAQ

Q: Which card is faster in Vulkan workloads?

A: The RX 9070 XT scores 65,879 in Geekbench Vulkan versus the GTX 660 Ti's 15,012, a 77.2% advantage. The RX 9070 XT also supports Vulkan 1.4, while the GTX 660 Ti is capped at Vulkan 1.2.175.

Q: How much memory does each card have?

A: The GTX 660 Ti has 2 GB of GDDR5 on a 192-bit bus, while the RX 9070 XT has 16 GB of GDDR6 on a 256-bit bus. The RX 9070 XT offers 644.6 GB/s bandwidth against 144.2 GB/s for the older card.

Q: Does the RX 9070 XT support ray tracing?

A: Yes, it has 64 dedicated ray tracing cores. The GTX 660 Ti has no ray tracing cores at all.

Q: What is the power draw difference?

A: The GTX 660 Ti has a TDP of 150 W with a suggested 450 W PSU, while the RX 9070 XT has a TDP of 304 W with a suggested 700 W PSU. The RX 9070 XT requires 2x 8-pin connectors versus 2x 6-pin on the older card.

Q: Which card has better DirectX support?

A: The RX 9070 XT supports DirectX 12 Ultimate (12_2), while the GTX 660 Ti only supports DirectX 12 (11_0). This means the newer card can run modern DirectX 12 Ultimate features, while the older card cannot.

Q: How do the average benchmark scores compare?

A: The GTX 660 Ti has an average benchmark score of 15,063, placing it at the 57th percentile. The RX 9070 XT has an average score of 13,543, placing it at the 55th percentile, but this is skewed by the RX 9070 XT's inclusion of Passmark DirectX 9, 10, and 12 tests where it scores low (353, 154, and 78 respectively), dragging its average down despite its dominant head-to-head wins.

The Verdict

The data is unambiguous for modern workloads: the RX 9070 XT is the superior card by every recorded metric. It wins both shared benchmarks, offers 18.5x the FP32 throughput, 4.5x the memory bandwidth, and adds ray tracing, PCIe 5.0, and modern API support. Anyone building a system for current games, Vulkan-based applications, or compute tasks should choose the RX 9070 XT without hesitation. Its 304 W TDP and 700 W PSU requirement are the only practical hurdles, but those are standard for high-end cards in its class.

The GTX 660 Ti remains relevant only as a legacy part. Its 150 W TDP and 450 W PSU requirement make it easier to slot into older builds, and its 57th percentile ranking is actually higher than the RX 9070 XT's 55th, but that is an artifact of the benchmark suite, not a real performance advantage. For DirectX 9 or 10 titles from its era, the GTX 660 Ti is serviceable, but the RX 9070 XT still posts higher scores in those specific Passmark tests (353 in DirectX 9, 154 in DirectX 10), albeit without a direct GTX 660 Ti comparison. The verdict is clear: if you are choosing between these two today, the RX 9070 XT is the only defensible option. The GTX 660 Ti belongs in a museum or a retro build, not in a modern gaming rig.