Quick/Comparisons
The GeForce GTX 1660 Ti takes the lead in gaming performance, thanks to its Turing architecture and 6GB GDDR6 memory.
Both GPUs offer strong workstation performance, with the GeForce GTX 1660 Ti's 29 points and Intel Arc B390's 25 points being very close.
The Intel Arc B390's 22 points in AI/ML outpace the GeForce GTX 1660 Ti's 18 points, thanks to its dedicated AI tensor cores and Celestial architecture.
20 | Gaming | 16 |
29 | Workstation | 25 |
18 | AI/ML |
Compare with something else?
Compare GeForce GTX 1660 with another →22 |
44 | Energy Efficiency | 100 |
23 | Quick Comparison Final Score | 33 |
| TU116 | Graphics Processor | Panther Lake |
| 1536 | Cores | 1536 |
| 96 / 48 | TMUs / ROPs | 48 / 24 |
| 796.71 | Blender GPU | 1268.79 |
| 52 | Forza Horizon 5 | 38 |
| 69 | The Witcher 3 | 55 |
| 96 | Counter-Strike 2 | 60 |
| 64 | Far Cry 6 | 49 |
| 37 | Hogwarts Legacy | 21 |
| 47 | Call of Duty | 32 |
| 32 | Ghost of Tsushima | 25 |
| 42 | Cyberpunk 2077 | 33 |
| 74 | Tomb Raider | 52 |
| 57 | Average FPS | 41 |
| 74 | 1080p High | 52 |
| 57 | 1080p Ultra | 41 |
| 41 | 1440p Ultra | 29 |
| 21 | 4K Ultra | 14 |
| Low | Margin of Error | Low |
| 72340 | GB6 Compute Score | 64546 |
| 124.8 img/sec | Background Blur | — |
| 69.4 img/sec | Face Detection | — |
| 2.67 Gpixels/sec | Horizon Detection | — |
| 3.48 Gpixels/sec | Edge Detection | — |
| 3.52 Gpixels/sec | Gaussian Blur | — |
| 0.92 Gpixels/sec | Feature Matching | — |
| 355.9 Gpixels/sec | Stereo Matching | — |
| 9094.4 FPS | Particle Physics | — |
| OpenCL | API | — |
| 12661 | G3D Mark Score | 9029 |
| 779 | G2D Mark | 786 |
| 103 FPS | DirectX 11 | 70 FPS |
| 51 FPS | DirectX 12 | 50 FPS |
| 5074 Ops/s | GPU Compute | 3702 Ops/s |
| 7620 | GB6 ML Quantized | 7879 |
| 4272 | Image Classification (SP) | 5487 |
| 6435 | Image Segmentation (HP) | 16864 |
| 9635 | Image Super Resolution (Q) | 8788 |
| 10867 | Face Detection (HP) | 25029 |
| 41061 | Pose Estimation (Q) | 18852 |
| 1737 | Text Classification (SP) | 2115 |
| 3301 | Machine Translation (HP) | 4266 |
| 5600 | Object Detection (SP) | 7449 |
| 18914 | Depth Estimation (Q) | 22207 |
| 83427 | Style Transfer (SP) | 95039 |
| ONNX | Framework | ONNX |
| DirectML | Backend | DirectML |
| 10788 | GB6 ML Single Precision | 11871 |
| 8753 | GB6 ML Half Precision | 20120 |
| 85 GPixel/s | Pixel Fill Rate | 60 GPixel/s |
| 170 GTexel/s | Texture Fill Rate | 120 GTexel/s |
| 5.4 TFLOPS | FLOPS (FP32) | 7.7 TFLOPS |
| GDDR6 | Memory Type | System Shared |
| 6 GB | Memory Size | — |
| 1500 MHz | Memory Clock | — |
| 12000 Mbps | Effective Memory Speed | — |
| 192-bit | Bus | — |
| No | ECC | No |
| 288 GB/s | Memory Bandwidth | — |
| PCIe 3.0 x16 | Interface | PCIe 5.0 |
| 120 W | TGP | — |
| TSMC | Manufacturing | TSMC |
| 12 nm | Fabrication Process | 3 nm |
| 284 mm² | Die Size | — |
| 6 billion | Transistor Count | — |
| 21.13 MTr/mm² | Transistor Density | — |
| 12 | DirectX | 12.1 |
| 1.3 | Vulkan | 1.4 |
| 4.6 | OpenGL | 4.6 |
| 3 | OpenCL | 3 |
| 7.5 | CUDA | — |
| No | Ray Tracing | Yes |
| No | DLSS | No |
| 1.4a | DisplayPort | 2.1 |
| Nvidia | Vendor | Intel |
| Discrete | Build | Integrated |
| February 22, 2019 | Released | January 5, 2026 |
| Desktop | Case | Laptop |
| Gaming | Purpose | Gaming |
| Entry-level | Segment | Mid-range |
| Turing | Architecture | Celestial |
| TU116 | GPU Codename | Panther Lake |
| -Intel Core i5 9500or above | Recommended CPU | -Intel Core Ultra X9 388Hor above |
| 5944 | Steel Nomad Lite Score | 4242 |
| 6339 | Time Spy | 6595 |
| 20534 | Solar Bay | 26399 |
| 1650 | Port Royal | 3886 |
| 16653 | Fire Strike | 16770 |
| 13005 | Wild Life Extreme | 11707 |
| 76361 | Night Raid | 68498 |
| 1500 MHz | Base Clock | 300 MHz |
| 1770 MHz | Boost Clock | 2500 MHz |
| 1536 | Shading Units | 1536 |
| 96 | Texture Mapping Units (TMUs) | 48 |
| 48 | Render Output Units (ROPs) | 24 |
| 24 | Compute Units (Pipelines) | 12 |
| No | Tensor Cores | — |
| No | Ray-tracing Cores | — |
| 64KB per cluster | L1 Cache | 64KB per cluster |
| 1536KB shared | L2 Cache | 16MB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |