NVIDIA's GeForce RTX 5090D, designed for the Chinese market, will come with significant restrictions impacting its AI and crypto-mining performance. These limitations aim to prevent the card from being used in multi-GPU setups commonly used for professional workloads.
Restricted Performance for AI and Crypto
The RTX 5090D will have a performance cut-off for AI and crypto-related applications. After approximately three seconds of heavy workload, the GPU will automatically reduce its performance, likely through reduced clock speeds or software throttling. Power limits will also be restricted, further impacting performance for professional computing tasks.
No Multi-GPU Support
Unlike previous models, the GeForce RTX 5090D will not support multi-GPU configurations. This means users will be unable to combine multiple RTX 5090D cards for increased performance, a practice that was previously common for those trying to achieve performance similar to that of AI accelerators. These restrictions are in place even for Linux systems.
Implications for Gamers
While the limitations are focused on AI and professional computing, these restrictions will have no impact on the gaming performance of the RTX 5090D. Gamers can expect the same level of performance as they would from the regular RTX 5090, which will release alongside the 5090D.
RTX 5090D Specifications
The GeForce RTX 5090D will utilize the GB202 GPU die, featuring 21,760 CUDA Cores. It will also be equipped with 32 GB of GDDR7 memory running at 28 Gbps, providing a total bandwidth of 1792 GB/s. AIB (Add-in-Board) models from manufacturers like GALAX have already surfaced, indicating the product is ready for release.
Source: Chiphell
