MSI GeForce RTX 5090 SUPRIM Liquid SOC 32G

The MSI GeForce RTX 5090 SUPRIM Liquid SOC 32G is MSI's flagship single-GPU graphics card built on NVIDIA's Blackwell architecture, featuring 32GB of GDDR7 memory and the full GB202 GPU die. As part of the SUPRIM Liquid SOC series, it ships with a factory-overclocked boost clock and an integrated all-in-one liquid cooling solution, enabling sustained peak compute throughput in thermally constrained workstation environments where conventional air-cooled blower or open-air designs would throttle under continuous load.

The RTX 5090 GPU at the heart of this card delivers 21,760 CUDA cores, fifth-generation Tensor Cores for accelerated AI inference and training, and fourth-generation RT Cores for real-time ray tracing. The 512-bit wide GDDR7 memory interface provides substantial memory bandwidth suited to large-model inference, high-resolution simulation, and professional visualization workloads that saturate the memory subsystem of previous-generation cards. NVIDIA's DLSS 4 with Multi Frame Generation is supported, enabling AI-assisted rendering for interactive visualization pipelines.

Targeted at AI workstation integrators, VFX studios, and enterprise compute teams requiring maximum single-GPU performance, the SUPRIM Liquid SOC 32G occupies a dual-slot PCIe form factor at the display bracket while routing thermal dissipation externally through its AIO radiator, making it compatible with standard ATX chassis that include radiator mounting points. The card connects via PCIe 5.0 x16 and draws power through a 16-pin (12V-2x6) connector configuration, requiring a high-capacity power supply appropriate to the card's TDP.

Ideal for

• Large language model inference and fine-tuning on AI workstations where 32GB GDDR7 capacity enables running multi-billion-parameter models without offloading to system RAM
• Professional VFX and 3D animation rendering using GPU-accelerated renderers such as NVIDIA Iray, Chaos V-Ray GPU, and Blender Cycles, benefiting from full Blackwell RT Core throughput
• Scientific simulation and computational fluid dynamics leveraging CUDA and the high memory bandwidth of the 512-bit GDDR7 bus for large dataset processing
• AI-assisted medical imaging and diagnostic workstation applications requiring sustained, thermally stable GPU compute over extended sessions
• Real-time virtual production and LED volume rendering pipelines where DLSS 4 and high rasterization throughput support interactive frame rates at 8K resolutions
• Generative AI content creation workflows including image synthesis and video diffusion models that benefit from the combination of high VRAM capacity and Tensor Core throughput

Technical specifications

Available from Omnixon Global. Submit an RFQ and our team will confirm configuration and availability for your order.