All NVIDIA GeForce GTX 1080 Specifications Revealed, Launch Presentation Leaked

While we knew quite a lot about NVIDIA’s GeForce GTX 1080 graphics card, but not all details we revealed at the card’s unveiling. Well ahead of the launch the official launch slide presentation has been leaked to the web. These slides reveal quite a lot about the card that we did not know about! Let’s get started!

NVIDIA Pascal GP104 Block Diagram

Above is the official block diagram for the Pascal GP104 GPU that powers the GTX 1080. In total we have 2560 CUDA cores, 160 TMUs, and 64 ROPs. The GP104 is split into four blocks (Graphics Processing Clusters), each GPC has 5 Streaming Multiprocessors (SM) and each SM has two block of 64 CUDA cores. The die area of GP104 GPU is 314 mm2, which puts it right between the GM204 and GM206. Even though it is smaller than the GM204 it actually have 2 billion more transistors more than GM204.

8 GB of GDDR5X Memory

The GTX 1080 is the first commercially available graphics card to make use of GDDR5X memory. GDDR5X has higher bandwidth and shorter interfaces than GDDR5. The memory on the GTX 1080 runs at 1250 MHz or 10 GHz (GDDR5X-effective). The GP104 has eight 32-bit memory controllers (256-bit total), so you maximum bandwidth should be around 320 GB/s.

Pascal Memory Compression

The GP104 brings with it new compression technology. The fourth generation delta color compression is said to deliver 1.7 times of GTX 980 effective bandwidth. You can see the difference in compression between Maxwell and Pascal below.

NVIDIA-GeForce-GTX-1080-Maxwell-Memory-Compression NVIDIA-GeForce-GTX-1080-Pascal-Memory-Compression-2

Display Support and Video Support

As far as display outputs you have 3 DisplayPort 1.4, HDMI 2.0B, and DVI. The GTX 1080 supports a maximum resolution of 7680 x 4320 @ 60 Hz (requires 2 DP 1.3 connectors). As far as video support we have some new things, check out the graph below.



Now during the GTX 1080 unveiling Jen-Hsun said that thee GTX 1080 is faster than GTX 1080 SLI, but this is only true when Virtual Reality is considered. In real-word gaming scenario’s the performance is almost comparable (reviews should confirm this), but it is definitely not faster, at least not in all games.


GPU Boost 3.0



Preemption in VR – Pascal




Simultaneous Multi-Projection



Asynchronous Computing



About Author