PCI Express graphics card MM4070A6-8G

HDMImedical imaging

Product overview
The MM4070A6-8G is a compact MXM 3.1 Type A GPU module based on the NVIDIA Ada Lovelace architecture. It integrates a GeForce RTX4070 Max-Q-class GPU with 8GB GDDR6 memory and 4608 CUDA cores, delivering high-performance graphics and compute in a low-power 65W profile suited for embedded workstations and industrial systems.

Key features

• NVIDIA Ada Lovelace architecture — NVIDIA GeForce RTX4070 Max-Q GPU
• 8GB GDDR6, 128-bit memory bus, 256GB/s bandwidth
• 4608 CUDA cores; 144 Tensor Cores; 36 RT Cores
• MXM 3.1 Type A form factor (70×82 mm), PCIe 4.0 x8 interface
• Multi-display outputs: 2×DP 1.4++ , HDMI 2.1, eDP; supports up to DP 7680×4320@120Hz and HDMI 8K@60Hz
• 65W TDP; supports Windows 10/11 and Linux

Technical specifications

• Model: MM4070A6-8G (RTX4070m)
• GPU architecture: NVIDIA Ada Lovelace
• GPU: NVIDIA GeForce RTX4070 Max-Q
• Memory: 8GB GDDR6, 128-bit, 256GB/s
• Core frequency (Base/Boost): 735 / 1230 MHz
• CUDA Cores: 4608; Tensor Cores: 144; RT Cores: 36
• TMUs: 144; ROPs: 48; SM count: 36
• Interface: MXM 3.1 Type A (70×82 mm), PCIe 4.0 x8
• Power: TDP 65W; Supply voltage DC 12–19V, 3.3V & 5V ±5%
• Dimensions: 70 × 82 mm
• Display outputs: 2×DP 1.4++, HDMI 2.1, eDP
• Maximum resolution: DP 7680×4320@120Hz; HDMI 8K@60Hz
• FP32 performance: 15.7 TFLOPS; FP64: 250.1 GFLOPS
• APIs: DirectX 12 Ultimate (12_2), OpenGL 4.6, OpenCL 3.0, Vulkan 1.3, Shader Model 6.7
• CUDA version: 8.9

Operating conditions & compatibility

• Operating temperature: 0°C to 45°C; Storage temperature: -20°C to 75°C
• Operating humidity: 0%–95% non-condensing; Storage humidity: 10%–90%
• OS support: Windows 10/11, Linux

Applications

• Embedded workstations, industrial PCs and compact servers
• Machine vision, edge AI inference, image processing and visualization
• Medical imaging, simulation workstations and specialized OEM systems