GeForce RTX 5090 Now Easily Available in China; Blower Style Conversion Prevents Connector Melting Issues

The Chinese market appears to have a consistent supply of the GeForce RTX 5090, despite the ban on its export due to U. S.policies. This situation raises interesting questions about availability and deployment, particularly in sectors leveraging artificial intelligence (AI).

NVIDIA RTX 5090: Availability and Configuration Insights from China

Recently, we highlighted the surge of NVIDIA GeForce RTX 5090 graphics cards being smuggled into China, as they are officially prohibited from exportation. These GPUs are being repurposed into dual-slot blower-style cards, specifically designed for AI workloads integrated into server environments. A Bilibili channel has elaborated on these developments, indicating a marked ease of access to the RTX 5090 within the region.

User explaining power supply and configuration for a gaming server setup. — Source: Bilibili

The card is not only readily available on various major e-commerce platforms, but its steady supply means it can readily meet enterprise demand. The RTX 5090 is noted for its superior AI performance, especially when compared to its restricted counterpart, the RTX 5090 D. While news of the upcoming RTX 5090 D V2 suggests a new model will be available soon, enterprises seem to prioritize the faster, more capable RTX 5090, which boasts 32 GB VRAM as opposed to the 24 GB of the RTX 5090 D V2.

Technician examining graphics card with annotated VRM and pin connectors. — The blower-style RTX 5090 features its 16-pin connector positioned on the right. Image Credit: @unikoshardware

Insights from the Bilibili channel 阿健座谈 suggest that while the RTX 5090 is engineered for high performance, its original design isn’t necessarily ideal for server deployment. Server cases often require a more compact design, which has led to a reconfiguration where custom thicker GPUs are transformed into dual-slot graphics cards equipped with a blower-style cooler. Notably, this modification includes a shift of the GPU chip and VRAM modules to a newer printed circuit board (PCB) that accommodates the 16-pin connector, now relocated to the right side.

Close-up of computer hardware with cables and labeled measurements for installation guidance. — Regular RTX 5090 designs lack the necessary clearance, making them incompatible with server cases. Image Credit: @unikoshardware

The configuration change addresses a crucial issue; standard server cases often do not allow the 35mm clearance needed for the 16-pin connector, which can prevent connections from overheating. By relocating the connector, space is conserved, and the risk of melting is significantly diminished. Reports indicate that enterprises utilizing nearly a thousand RTX 5090 and RTX 4090 cards have experienced no instances of connector failures so far, reinforcing the effectiveness of this new design. Insights from EVGA’s Kingpin suggest that the connector’s positioning plays a vital role in thermal management; connectors closer to VRM can overheat, while those on the right edge of the PCB maintain cooler operation.

Not only do these adjustments promote stability within server environments, but they also mitigate the risks associated with cable bends that could occur at the connector’s end, enhancing both performance and longevity.

For more information on the developments regarding the RTX 5090, visit @unikoshardware.

Source & Images