NVIDIA’s upcoming GeForce RTX 5090 graphics card is set to feature a significantly larger GPU compared to the current RTX 4090. The flagship model of the GeForce RTX 50 "Blackwell" series will boast a GB202 chip with a die size of 744mm², marking a 22% increase over the RTX 4090’s 609mm².
In terms of chip design, the NVIDIA GeForce RTX 5090 is shaping up to be a powerhouse. Using TSMC’s 5nm process node, this GPU promises substantial performance improvements while incorporating a large die size of 24x31mm (744 mm²) and a packaging area of 63x65mm (3528 mm²). For context, the RTX 4090 had a die size of 609mm² and packaging of 2601mm² (51x51mm). While a larger die doesn’t inherently mean higher power consumption or increased heat, it’s important to note that this die is still smaller than the 754mm² Turing TU102 chip from the first-gen RTX 20 series, which was pioneering in its use of Tensor and RT cores.
Looking specifically at the next-gen NVIDIA GeForce RTX 5090, it will feature the top-tier PG144/145-SKU30 PCB design with the GB202-300-A1 core. Out of 192 Streaming Multiprocessors (SMs), 170 will be enabled, offering 21,760 cores. This reflects about an 11.4% core reduction compared to the total, slightly higher than RTX 4090’s 11.1% reduction from its full AD102 die.
On the memory front, the RTX 5090 won’t disappoint, coming equipped with 32 GB of GDDR7 memory through a 512-bit interface. Operating at speeds of 28 Gbps, the card will deliver up to an impressive 1792 GB/s bandwidth. Enhanced memory compression and a larger L3 cache will further augment the GPU’s bandwidth. Despite its powerful specs, the card is rated with a 600W Total Board Power (TBP), though actual gaming consumption will be lower. Earlier reports suggest a 2-slot cooler design for the Founders Edition of the RTX 5090.
Expect NVIDIA to unveil the RTX 5090 at the upcoming CES 2025 event. More details are likely to emerge as the event approaches. Here’s a quick glance at the preliminary specs for the entire GeForce RTX 50 lineup:
-
NVIDIA GeForce RTX 5090: 21760 cores, 32 GB GDDR7, 512-bit memory, 1792 GB/s bandwidth, 600W TBP.
-
NVIDIA GeForce RTX 5080: 10752 cores, 16 GB GDDR7, 256-bit memory, 1024 GB/s bandwidth, 400W TBP.
- NVIDIA GeForce RTX 5070 and RTX 5060: Specifications are yet to be determined.
Each of these models will feature the new 12V-2×6 (16-Pin) power interface. Now, which of these RTX 50 series GPUs are you most excited about? Share your thoughts on social media or stay tuned for the latest updates!
