NVIDIA GeForce RTX 5090 PCB Leak Reveals Massive GPU, GDDR7 Memory, and Robust VRMs
The hype surrounding NVIDIA's GeForce RTX 5090 just got a boost with fresh leaks detailing its fully-equipped PCB. This flagship GPU from the upcoming RTX 50 "Blackwell" series flaunts some impressive hardware that sets a new standard for next-gen graphics performance.
What the Leak Reveals
Shared on Baidu Forums, the leaked PCB showcases the Blackwell GB202-300-A1 GPU, surrounded by 16 Samsung GDDR7 memory modules and an intricate array of 28 VRMs, ensuring robust power delivery. The qualification sample GPU features a massive die, highlighting NVIDIA's push for performance and efficiency.
The GDDR7 memory modules, clocked at 28 Gbps, operate on a 512-bit bus, delivering an astonishing 1792 GB/s bandwidth—a 78% jump over the RTX 4090's 1008 GB/s. Paired with the enhanced VRM setup and a 16-pin 12V-2x6 power connector, the RTX 5090 is designed to handle a 600W TBP (Total Board Power).
Performance Expectations
The RTX 5090's GPU packs 21760 CUDA cores (a 33% increase over the RTX 4090), while its memory configuration ensures top-notch gaming and workstation performance. With launch details pointing to January or February 2025, NVIDIA seems poised to kick off the new year with a bang, likely showcasing the card at CES 2025.
With competitors like AMD's RDNA 4 series on the horizon, NVIDIA is aiming to maintain dominance in the high-end GPU space. The introduction of GDDR7 memory and groundbreaking power delivery architectures ensures the RTX 5090 will remain a premium choice for enthusiasts and professionals alike.
Are you excited about the next-gen RTX 5090, or do you think the industry should focus more on power efficiency than raw performance? Let us know your thoughts in the comments below!
