Quick/Comparisons
The Intel Arc B370's 7-point lead in gaming performance is significant, thanks to its Celestial architecture and hardware ray tracing support.
The Intel Arc B370's 3-point lead in workstation performance is due to its Panther Lake codename and System Shared memory features.
The Intel Arc B370's 11-point lead in AI/ML performance is substantial, making it a better choice for tasks that rely on this capability.
17 | Gaming | 10 |
24 | Workstation | 21 |
20 | AI/ML |
Compare with something else?
Compare Intel Arc B370 with another →9 |
100 | Energy Efficiency | 73 |
33 | Quick Comparison Final Score | 23 |
| Panther Lake | Graphics Processor | TU117 |
| — | Cores | 1024 |
| — | TMUs / ROPs | 64 / 32 |
| — | Blender GPU | 142.83 |
| 38 | Forza Horizon 5 | — |
| 54 | The Witcher 3 | — |
| 61 | Counter-Strike 2 | — |
| 52 | Far Cry 6 | — |
| 24 | Hogwarts Legacy | — |
| 35 | Call of Duty | — |
| 27 | Ghost of Tsushima | — |
| 33 | Cyberpunk 2077 | — |
| 50 | Tomb Raider | — |
| 42 | Average FPS | — |
| 54 | 1080p High | — |
| 42 | 1080p Ultra | — |
| 29 | 1440p Ultra | — |
| 15 | 4K Ultra | — |
| High | Margin of Error | — |
| 48273 | GB6 Compute Score | 35511 |
| 117.4 img/sec | Background Blur | 61.9 img/sec |
| 60.6 img/sec | Face Detection | 28.6 img/sec |
| 1.39 Gpixels/sec | Horizon Detection | 1.2 Gpixels/sec |
| 1.95 Gpixels/sec | Edge Detection | 1.67 Gpixels/sec |
| 2.45 Gpixels/sec | Gaussian Blur | 1.9 Gpixels/sec |
| 0.66 Gpixels/sec | Feature Matching | 0.5 Gpixels/sec |
| 144.8 Gpixels/sec | Stereo Matching | 174.2 Gpixels/sec |
| 7272.6 FPS | Particle Physics | 4869.6 FPS |
| OpenCL | API | OpenCL |
| 8227 | G3D Mark Score | 4752 |
| 800 | G2D Mark | 401 |
| 64 FPS | DirectX 11 | 35 FPS |
| 41 FPS | DirectX 12 | 30 FPS |
| 3327 Ops/s | GPU Compute | 2256 Ops/s |
| 10834 | GB6 ML Single Precision | 4976 |
| 20482 | GB6 ML Half Precision | 4623 |
| 7349 | GB6 ML Quantized | 3647 |
| 5389 | Image Classification (SP) | 2063 |
| 18899 | Image Segmentation (HP) | 3228 |
| 8326 | Image Super Resolution (Q) | 4420 |
| 25377 | Face Detection (HP) | 5053 |
| 16174 | Pose Estimation (Q) | 14361 |
| 1763 | Text Classification (SP) | 1515 |
| 4140 | Machine Translation (HP) | 2212 |
| 7138 | Object Detection (SP) | 2394 |
| 20034 | Depth Estimation (Q) | 8645 |
| 84880 | Style Transfer (SP) | 33780 |
| ONNX | Framework | ONNX |
| DirectML | Backend | DirectML |
| — | Pixel Fill Rate | 53 GPixel/s |
| — | Texture Fill Rate | 107 GTexel/s |
| — | FLOPS (FP32) | 3.4 TFLOPS |
| System Shared | Memory Type | GDDR6 |
| — | Memory Size | 4 GB |
| — | Memory Clock | 1500 MHz |
| — | Effective Memory Speed | 12000 Mbps |
| — | Bus | 64-bit |
| No | ECC | No |
| — | Memory Bandwidth | 96 GB/s |
| PCIe 5.0 | Interface | PCIe 3.0 x16 |
| — | TGP | 30 W |
| TSMC | Manufacturing | TSMC |
| 3 nm | Fabrication Process | 12 nm |
| — | Die Size | 200 mm² |
| — | Transistor Count | 4.7 billion |
| — | Transistor Density | 23.5 MTr/mm² |
| 12.1 | DirectX | 12 |
| 1.4 | Vulkan | 1.3 |
| 4.6 | OpenGL | 4.6 |
| 3 | OpenCL | 3 |
| — | CUDA | 7.5 |
| Yes | Ray Tracing | No |
| No | DLSS | No |
| 2.1 | DisplayPort | 1.4 |
| Intel | Vendor | Nvidia |
| Integrated | Build | Discrete |
| January 5, 2026 | Released | May 1, 2022 |
| Laptop | Case | Laptop |
| Generic | Purpose | Professional |
| Mid-range | Segment | Entry-level |
| Celestial | Architecture | Turing |
| Panther Lake | GPU Codename | TU117 |
| -Intel Core Ultra 7 356Hor above | Recommended CPU | -Intel Core i7 1265Uor above |
| 5045 | Steel Nomad Lite Score | 1984 |
| 6023 | Time Spy | 2273 |
| — | Solar Bay | — |
| 3578 | Port Royal | — |
| 14710 | Fire Strike | 5443 |
| — | Wild Life Extreme | 4481 |
| — | Night Raid | 28589 |
| — | Base Clock | 1065 MHz |
| 2400 MHz | Boost Clock | 1665 MHz |
| — | Shading Units | 1024 |
| — | Texture Mapping Units (TMUs) | 64 |
| — | Render Output Units (ROPs) | 32 |
| 10 | Compute Units (Pipelines) | 16 |
| — | Tensor Cores | — |
| — | Ray-tracing Cores | — |
| 64KB per cluster | L1 Cache | 64KB per cluster |
| 16MB shared | L2 Cache | 1MB shared |
| 2 IPC | Instructions Per Cycle | 2 IPC |