3DMark Benchmark (DirectX, Raytracing)
3DMark is a benchmark program that determines the performance of certain components of a computer and then reports the performance as a numerical value.

The NVIDIA RTX 6000 Ada has 18176 shader units and these clock a maximum of 2.54 GHz which results in an FP32 computing power of 92.15 TFLOPS.

The INNO3D GeForce RTX 4070 SUPER ICHILL X3 has 7168 shader units and these clock at a maximum of 2.51 GHz and achieve an FP32 computing power of 35.91 TFLOPS.
In FP32 performance, the NVIDIA RTX 6000 Ada is 157 % faster than the INNO3D GeForce RTX 4070 SUPER ICHILL X3.
| GPU | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Based on | NVIDIA RTX 6000 Ada | NVIDIA GeForce RTX 4070 SUPER |
| GPU Chip | AD102 | AD104-350-A1 |
| Streaming Multiprocessors | 142 | 56 |
| Shader | 18176 | 7168 |
| Render Output Units | 192 | 80 |
| Texture Units | 568 | 224 |
| Raytracing Cores | 142 | 56 |
| Memory | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Memory Size | 48 GB | 12 GB |
| Memory Type | GDDR6 | GDDR6X |
| Memory Clock | 2.500 GHz | 1.313 GHz |
| Memory Speed | 20.0 Gbps | 21.0 Gbps |
| Memory bandwidth | 960 GB/s | 504 GB/s |
| Memory Interface | 384 bit | 192 bit |
| Clock Speeds | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Base Clock | 2.175 GHz | 1.980 GHz |
| Boost Clock | 2.535 GHz | 2.505 GHz |
| Overclocking | Yes | Yes |
| Thermal Design | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| TDP | 300 W | 220 W |
| Tjunction max | 93 °C | 90 °C |
| PCIe-Power | 1 x 16-Pin | 1 x 16-Pin |
| Cooler & Fans | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Fan-Type | Radial | Axial |
| Fan 1 | 1 x 70 mm | 3 x 98 mm |
| Cooler-Type | Air cooling | Air cooling |
| Noise (Idle) | 0 dB | 0 dB |
| Connectivity | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Max. Displays | 4 | 4 |
| HDCP | 2.3 | 2.3 |
| HDMI | -- | 1x HDMI 2.1a |
| DisplayPort | 4x DP 1.4a | 3x DP 1.4a |
| Featureset | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Max. resolution | 7680x4320 | 7680x4320 |
| DirectX | 12_2 (Ultimate) | 12_2 (Ultimate) |
| Raytracing | Yes | Yes |
| DLSS / FSR | Yes | Yes |
| LED | No LED lighting | Addressable LED |
| Supported Video Codecs | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| h264 | Decode / Encode | Decode / Encode |
| h265 / HEVC | Decode / Encode | Decode / Encode |
| VP8 | Decode | Decode |
| VP9 | Decode | Decode |
| AV1 | Decode / Encode | Decode |
| Dimensions | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| Length | 267 mm | 334 mm |
| Height | 112 mm | 148 mm |
| Width | -- | 62 mm |
| Width (Slots) | 2 Slots | 3 Slots |
| SFF-Ready | No | No |
| Additional data | NVIDIA RTX 6000 Ada | INNO3D GeForce RTX 4070 SUPER ICHILL X3 |
|---|---|---|
| GPU Interface | PCIe 4.0 x 16 | PCIe 4.0 x 16 |
| Release date | Q1/2023 | Q1/2024 |
| Launch Price | 6,799 $ | 599 $ |
| Structure size | 5 nm | 5 nm |
| Documents | data sheet | data sheet |
3DMark is a benchmark program that determines the performance of certain components of a computer and then reports the performance as a numerical value.
Cyberpunk 2077 was developed by CD Projekt Red and is based on the development studio's own REDengine 4. Ray tracing is switched off in our benchmarks.
The Last of Us Part One is a game released by Sony in June 2013 exclusively for the Playstation and released for the PC in early 2023. The benchmark values here were determined at high details.
Battlefield 5 is a visually stunning game that is ideal as a graphics card benchmark. We test the game with maximum details on Windows 10.
Geekbench 6 is a cross-platform benchmark for main processors, which also carries out 3 different graphics benchmarks and outputs them in the form of a numerical value.
The theoretical computing power of the graphics card with single precision (32 bit) in TFLOPS indicates how many trillion FP32 floating point operations the graphics card (GPU) can perform per second.