This card features two GPUs and 12 GB of video memory.
This week during GTC 2014, Nvidia CEO Jen-Hsun Huang revealed a monster of a graphics card, the GeForce GTX Titan Z. This card features 5,760 CUDA cores, 12 GB of on-board frame buffer memory, and 8 TeraFLOPS of performance.
"If you're in desperate need for a supercomputer and you need one really close by and handy, and you want it to fit under your desk, we have just the friend for you," Huang said in his keynote.
The new Titan Z features two GK110 chips that are tuned to run at the same clock speed, and backed by dynamic power balancing so that the GPUs don't create a performance bottleneck. They're also engineered for multi-monitor gaming and 5K media.
"That performance is delivered in a card that is cool and quiet, rather than hot and loud," states Nvidia's blog on Monday. "Low-profile components and ducted baseplate channels minimize turbulence and improves acoustic quality."
Huang compared the new Titan Z to Google Brain, which features 1,000 servers packing 2,000 CPUs (16,000 cores), 600 kWatts and a hefty price tag of $5,000,000 USD. A solution using Titan Z would only need three GPU-accelerated servers with 12 Nvidia GPUs total.
In a demo, the company revealed simulation and graphics running simultaneously. Plankton illuminated the water, exposing a number of glowing jellyfish and an oncoming whale. Once the creature was completely visible in the scene, the whale jumped up out of the water and came crashing back down, causing believable ripples on the surface. What was really cool was to pan around the whale in mid splash in real time.
When the whale came reaching out of the water, it generated the waves simulation, their 2D grids, their 3D grids, their particle systems, their mist generators and foam generators, all of which was rendered in real time. As the whale landed, the weight pushed around a small raft. Very, very impressive stuff.
The GeForce GTX Titan Z currently has a meaty price tag of $2,999 USD. Additional information will be provided to the press at a later date.