NVIDIA GeForce GTX 660 Ti vs NVIDIA GeForce RTX 3050 OEM Comparison

The GeForce RTX 3050 OEM and the GeForce GTX 660 Ti are separated by a decade of GPU architecture, and the benchmark data reflects that chasm clearly. The RTX 3050 OEM is an Ampere-era card built on an 8 nm Samsung process, while the GTX 660 Ti is a Kepler-era card on TSMC's 28 nm node. This head-to-head is less a contest and more a demonstration of how far the product line has progressed, with the newer card dominating every measured workload.

Head-to-Head Benchmarks

The data available for direct comparison covers two compute-oriented tests, and the RTX 3050 OEM wins both by a massive margin. In Geekbench OpenCL, the RTX 3050 OEM scores 60,740, which is 301.9% higher than the GTX 660 Ti's 15,113. That is not a marginal improvement; it is a quadrupling of raw compute throughput. The Geekbench Vulkan result tells the same story: the RTX 3050 OEM posts 57,103 against the GTX 660 Ti's 15,012, a delta of 280.4%. The GTX 660 Ti has no wins in this comparison; the scorecard is 2-0 in favor of the newer card.

These two tests are the only ones both cards share in the fact pack, but they are indicative of the broader gulf in performance. The RTX 3050 OEM's average benchmark score across all tests is 15,199, while the GTX 660 Ti's average is 15,063. That overall average is close, but it is misleading because the GTX 660 Ti only has two benchmark entries, both of which are the same Geekbench tests where it loses badly. The RTX 3050 OEM has nine benchmark entries, including Passmark scores that cover DirectX 9 through 12, 2D, 3D, and compute workloads. Its Passmark G3D score alone is 11,857, and its GPU compute score is 5,779. The GTX 660 Ti has no Passmark entries in the data, so its average is pulled from the two Geekbench scores alone.

When placed against their respective nearest rivals, both cards sit at the 57th percentile of all GPUs, which is a curious coincidence. The RTX 3050 OEM's closest competitor is the AMD Radeon RX 7600, with a delta of just 0.2% in favor of the RTX 3050 OEM (15,199 vs. 15,171). The GTX 660 Ti's nearest rival is the AMD Radeon Pro 560X, which edges it by 0.1% (15,082 vs. 15,063). Notably, the two cards themselves are listed as rivals to each other: the RTX 3050 OEM is 0.9% ahead of the GTX 660 Ti in the latter's rival list, and the GTX 660 Ti is 0.9% behind the RTX 3050 OEM in the former's list. That 0.9% delta reflects the average scores, but the individual Geekbench tests show the RTX 3050 OEM is anywhere from 280% to 302% faster. The average is dragged down because the GTX 660 Ti only has those two losing scores, while the RTX 3050 OEM's nine scores include less demanding legacy tests.

The Verdict

From the data, the choice is unambiguous for anyone needing raw performance in compute or Vulkan workloads: the RTX 3050 OEM is the superior card. It is roughly three to four times faster in both Geekbench OpenCL and Vulkan tests. The GTX 660 Ti, despite having a higher TDP at 150 W versus the RTX 3050 OEM's 130 W, cannot compete in any measured metric. The RTX 3050 OEM also has 8 GB of GDDR6 memory versus the GTX 660 Ti's 2 GB of GDDR5, and it supports modern APIs like DirectX 12 Ultimate and Vulkan 1.4, whereas the GTX 660 Ti is limited to DirectX 12 (11_0) and Vulkan 1.2.175.

The only scenario where the GTX 660 Ti makes sense is if you are building a period-correct system from 2012 or need a card that fits a specific legacy PCIe 3.0 x16 slot configuration. But even then, the RTX 3050 OEM, which uses PCIe 4.0 x8, will run in most modern boards. Given that the RTX 3050 OEM is end-of-life and the GTX 660 Ti is also end-of-life, there is no future-proofing argument for the older card. The RTX 3050 OEM is the pick for virtually any use case, from gaming to compute tasks, based strictly on the benchmark results.

Architecture Differences

The two cards come from different architectural eras. The RTX 3050 OEM uses the GA106 chip built on Ampere architecture, manufactured by Samsung on an 8 nm process. It packs 12,000 million transistors into a 276 mm² die, yielding a transistor density of 43.5 million per square millimeter. The GTX 660 Ti uses the GK104 chip on Kepler architecture, built by TSMC on a 28 nm process, with 3,540 million transistors spread over a larger 294 mm² die, giving a density of just 12.0 million per square millimeter. That density difference is a direct result of the process node gap.

The RTX 3050 OEM has 2,304 shading units, 72 texture mapping units, and 32 ROPs. It also includes 18 ray tracing cores and 72 tensor cores, features the GTX 660 Ti lacks entirely. The GTX 660 Ti has 1,344 shading units, 112 TMUs, and 24 ROPs. Clock speeds tell a similar story: the RTX 3050 OEM runs at a base of 1515 MHz and boosts to 1755 MHz, while the GTX 660 Ti is at 915 MHz base and 980 MHz boost. The RTX 3050 OEM's FP32 compute is 8.087 TFLOPS, and it has FP16 at the same 8.087 TFLOPS with a 1:1 ratio. The GTX 660 Ti manages only 2.634 TFLOPS FP32 and has no FP16 data listed.

Memory architectures diverge as well. The RTX 3050 OEM uses 8 GB of GDDR6 on a 128-bit bus, delivering 224.0 GB/s of bandwidth. The GTX 660 Ti uses 2 GB of GDDR5 on a 192-bit bus, which provides 144.2 GB/s. The RTX 3050 OEM's memory clock is 1750 MHz (14 Gbps effective), while the GTX 660 Ti's is 1502 MHz (6 Gbps effective). Pixel and texture rates follow the same pattern: the RTX 3050 OEM outputs 56.16 GPixel/s and 126.4 GTexel/s, versus the GTX 660 Ti's 27.44 GPixel/s and 109.8 GTexel/s.

FAQ

Q: Which card has more raw compute power?

A: The RTX 3050 OEM delivers 8.087 TFLOPS of FP32 performance, which is more than triple the GTX 660 Ti's 2.634 TFLOPS.

Q: Do both cards support ray tracing?

A: No. The RTX 3050 OEM has 18 dedicated ray tracing cores, while the GTX 660 Ti has no ray tracing hardware at all.

Q: What is the memory capacity difference?

A: The RTX 3050 OEM has 8 GB of GDDR6 memory, while the GTX 660 Ti has 2 GB of GDDR5, a fourfold difference in capacity.

Q: How do their power requirements compare?

A: The RTX 3050 OEM has a TDP of 130 W and requires a 300 W power supply, while the GTX 660 Ti has a higher TDP of 150 W and needs a 450 W PSU.

Q: Which card supports newer DirectX features?

A: The RTX 3050 OEM supports DirectX 12 Ultimate (12_2), whereas the GTX 660 Ti is limited to DirectX 12 (11_0).

Q: Are their average benchmark scores similar?

A: The averages are close (15,199 for the RTX 3050 OEM vs. 15,063 for the GTX 660 Ti), but this is because the GTX 660 Ti only has two Geekbench entries, both of which it loses by over 280%.

Where Each One Wins

The RTX 3050 OEM wins in every category where data exists. In Geekbench OpenCL, it is 301.9% ahead, and in Geekbench Vulkan, it is 280.4% ahead. For compute-heavy tasks like GPGPU processing or any workload that leverages OpenCL, the RTX 3050 OEM is the clear choice. Its 72 tensor cores also make it suitable for AI-adjacent workloads, though no specific AI benchmarks are in the data. The 8 GB of GDDR6 memory means it can handle larger textures and datasets than the GTX 660 Ti's 2 GB frame buffer, which is a practical advantage in modern games and creative applications.

The GTX 660 Ti has no benchmark wins, but its specification sheet shows it is not without merits in a historical context. It has a wider 192-bit memory bus, which at the time helped offset its lower bandwidth. Its 112 TMUs outnumber the RTX 3050 OEM's 72, though the RTX 3050 OEM's higher clock speeds more than compensate in texture rate (126.4 GTexel/s vs. 109.8 GTexel/s). The GTX 660 Ti's 28 nm process is older, but it uses a larger die (294 mm²) despite having far fewer transistors, which suggests it was a less efficient design. For anyone building a system today, the RTX 3050 OEM is the only rational choice from this data. The GTX 660 Ti is a relic that can only be justified for retro builds or as a display adapter for very old software, and even then, its lack of modern API support is a significant limitation.

Specification Differences

The two cards differ in nearly every specification field. The RTX 3050 OEM uses a GA106 chip on Ampere architecture with an 8 nm Samsung process, while the GTX 660 Ti uses a GK104 chip on Kepler architecture with a 28 nm TSMC process. Transistor counts are 12,000 million versus 3,540 million, and die sizes are 276 mm² versus 294 mm², giving transistor densities of 43.5M/mm² and 12.0M/mm² respectively. The RTX 3050 OEM has base and boost clocks of 1515 MHz and 1755 MHz, versus 915 MHz and 980 MHz on the GTX 660 Ti. Memory differs: 8 GB GDDR6 on a 128-bit bus with 224.0 GB/s bandwidth versus 2 GB GDDR5 on a 192-bit bus with 144.2 GB/s. Shading units are 2304 versus 1344, TMUs are 72 versus 112, and ROPs are 32 versus 24. The RTX 3050 OEM adds 18 RT cores and 72 tensor cores, which the GTX 660 Ti does not have. Pixel rate is 56.16 GPixel/s versus 27.44 GPixel/s, and texture rate is 126.4 GTexel/s versus 109.8 GTexel/s. FP32 compute is 8.087 TFLOPS versus 2.634 TFLOPS. TDP is 130 W versus 150 W, power connectors are 1x 8-pin versus 2x 6-pin, and suggested PSU is 300 W versus 450 W. Bus interface is PCIe 4.0 x8 versus PCIe 3.0 x16. Display outputs are 1x HDMI 2.1 and 3x DisplayPort 1.4a versus 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. API support: DirectX 12 Ultimate (12_2) versus DirectX 12 (11_0), OpenGL 4.6 on both, and Vulkan 1.4 versus Vulkan 1.2.175. The RTX 3050 OEM has a launch MSRP of none listed; the GTX 660 Ti had a launch MSRP of 299 USD.