Intel Arc Pro B65 vs GeForce RTX 4090
VS
Intel Arc Pro B65
BMG-G21
Nvidia GeForce RTX 4090
AD102
We compared two discrete desktop graphics cards: the Intel Arc Pro B65 32 GB with 20 pipelines and 2560 shaders against the 3 years and 7 months older GeForce RTX 4090 24 GB that utilizes 128 pipelines and 16384 shaders. Here you will find complete details about specs, efficiency, performance tests, and more.
Please note that the tests on the Intel Arc Pro B65 are done on an engineering sample provided by our insiders. The data will be more accurate after the final version of this GPU is available.
Key differences
Reasons to consider the Intel Arc Pro B65
- Includes 8 GB more video memory
Reasons to consider the GeForce RTX 4090
- 6.7x higher maximum theoretical performance (82.6 vs 12.3 TFLOPS)
- Has 6.4x more shading units (16384 vs 2560)
- Features 352 more tensor cores for effective ML and AI workloads
- Supports Nvidia DLSS 3 technology
- Has 66% higher memory bandwidth: 1010 vs 608 GB/s
Gaming Performance
Frame rate comparison across popular AAA titles at different resolutionsGames
| Forza Horizon 5 | - | 275 199 193 155 |
| The Witcher 3 | - | 292 268 191 101 |
| Counter-Strike 2 | - | 339 238 188 90 |
| Far Cry 6 | - | 203 179 170 104 |
| Hogwarts Legacy | - | 200 163 132 74 |
| CoD: Modern Warfare III | - | 266 263 219 151 |
| Ghost of Tsushima | - | 188 152 137 91 |
| Cyberpunk 2077 | - | 226 208 143 66 |
| Shadow of the Tomb Raider | - | 302 284 222 118 |
| 1080p High | - | 255 |
| 1080p Ultra | - | 217 |
| 1440p Ultra | - | 177 |
| 4K Ultra | - | 106 |
| Margin of Error | High | Low |
Benchmarks
Graphics cards’ performance in recent benchmarking apps3D Mark
Steel Nomad Lite Score
| Time Spy | - | 36299 |
| Solar Bay | - | 187525 |
| Port Royal | - | 26127 |
| Fire Strike | - | 72432 |
| Wild Life Extreme | - | 85218 |
| Night Raid | - | 195813 |
GeekBench 6 GPU Compute
GB6 Compute Score
| Background Blur | - | 318.9 img/sec |
| Face Detection | - | 222.8 img/sec |
| Horizon Detection | - | 14.2 Gpixels/sec |
| Edge Detection | - | 21 Gpixels/sec |
| Gaussian Blur | - | 23.9 Gpixels/sec |
| Feature Matching | - | 2.38 Gpixels/sec |
| Stereo Matching | - | 1950 Gpixels/sec |
| Particle Physics | - | 48090.4 FPS |
| API | - | OpenCL |
Cinebench 2024 GPU
Cinebench 2024 GPU
Passmark Graphics
G3D Mark Score
| G2D Mark | - | 1303 |
| DirectX 11 | - | 323 FPS |
| DirectX 12 | - | 151 FPS |
| GPU Compute | - | 25991 Ops/s |
Recent User Tests
Intel Arc Pro B65
No benchmark results yet
GeForce RTX 4090
| Date | Benchmark | Result |
|---|---|---|
| 📘 2026-04-03 (afroman420IU) | Cinebench 2024 | 32578 |
Artificial Intelligence Tests
Performance in machine learning and artificial intelligence tasksGeekBench 6 ML
GB6 ML Single Precision
GB6 ML Half Precision
GB6 ML Quantized
| Image Classification (SP) | - | 15600 |
| Image Segmentation (HP) | - | 36657 |
| Image Super Resolution (Q) | - | 48114 |
| Face Detection (HP) | - | 78111 |
| Pose Estimation (Q) | - | 289639 |
| Text Classification (SP) | - | 4074 |
| Machine Translation (HP) | - | 6349 |
| Object Detection (SP) | - | 21186 |
| Depth Estimation (Q) | - | 75599 |
| Style Transfer (SP) | - | 601428 |
| Framework | - | ONNX |
| Backend | - | DirectML |
Specifications
Technical specifications of Intel Arc Pro B65 and GeForce RTX 4090General
| Vendor | Intel | Nvidia |
| Build | Discrete | Discrete |
| Released | March 25, 2026 | September 20, 2022 |
| Launch price (MSRP) | - | $1599 |
| Case | Desktop | Desktop |
| Purpose | Professional | Gaming |
| Segment | Mid-range | High-end |
| Architecture | Battlemage Pro Series | Ada Lovelace |
| GPU Codename | BMG-G21 | AD102 |
| Successor | - | - GeForce RTX 5090 |
| Recommended CPU | - | - Intel Core i9 14900K or above |
Desktop GPU rating (#3rd place)
Graphics Processing Unit
| Base Clock | 2000 MHz | 2235 MHz |
| Boost Clock | 2400 MHz | 2520 MHz |
| Shading Units | 2560 | 16384 |
| Texture Mapping Units (TMUs) | 160 | 512 |
| Render Output Units (ROPs) | 80 | 176 |
| Compute Units (Pipelines) | 20 | 128 |
| Tensor Cores | 160 | 512 |
| Ray-tracing Cores | 20 | 128 |
| L1 Cache | 256KB per cluster | 128KB per cluster |
| L2 Cache | 10MB shared | 72MB shared |
| Instructions Per Cycle | 2 IPC | 2 IPC |
Raw Performance
| Pixel Fill Rate | 192 GPixel/s | 444 GPixel/s |
| Texture Fill Rate | 384 GTexel/s | 1290 GTexel/s |
FLOPS (FP32)
Physical
| Interface | PCIe 5.0 x16 | PCIe 4.0 x16 |
| TGP | 200 W | 450 W |
| Manufacturing | TSMC | TSMC |
| Fabrication Process | 5 nm | 5 nm |
| Die Size | 271.95 mm² | 609 mm² |
| Transistor Count | 19.6 billion | 76.3 billion |
| Transistor Density | 72.07 MTr/mm² | 125.29 MTr/mm² |
Memory
| Memory Type | GDDR6 | GDDR6X |
| Memory Size | 32 GB | 24 GB |
| Memory Clock | 2375 MHz | 2625 MHz |
| Effective Memory Speed | 19000 Mbps | 21000 Mbps |
| Bus | 256-bit | 384-bit |
| ECC | No | No |
Memory Bandwidth
API
| DirectX | 12.2 | 12 |
| Vulkan | 1.3 | 1.3 |
| OpenGL | 4.6 | 4.6 |
| OpenCL | 3.0 | 3.0 |
| CUDA | No | 8.9 |
| Ray Tracing | Yes | Yes |
| DLSS | No | DLSS 3 |
| DisplayPort | 2.1a | 1.4a |
Cast your vote
Total votes: < 1