Nvidia Geforce GTX Titan The $999 GPU Supercomputer

Nvidia Geforce GTX Titan The $999 GPU Supercomputer

The Nvidia Geforce GTX Titan was released yesterday. The single GPU video card for the PC desktop is the fastest graphics you can buy at the moment.

According to Tech Spot, “Cray’s XK7 Titan supercomputer is powered by no less than 18,000 Nvidia Tesla K20x GPUs, which Nvidia is proud to say highly contributes to make the Titan the world’s fastest supercomputer.” As previously reported by The Inquisitr, the standalone $999 Nvidia Geforce Titan is still a behemoth, combining 2688 stream processors, 224 texture units, and 48 raster operating units to produce a theoretical floating point performance of 4.7 TeraFLOPS. The GPU features 6GB of 384-bit DDR5 memory alongside a GPU clock of 875MHz and an effective memory clock of 6GHz.

According to AnandTech, Nvidia does not want review sites to list benchmarks quite yet, but we can comment on several factors. The TDP power rating is only 250 watts for the Geforce Titan. In comparison, the GTX 690 needs 300 watts of power. This increased energy efficiency, along with better thermal and noise levels, should allow small form factor PCs to work in the Geforce Titan.

The other major performance factor is that the Geforce Titan will ship with pull performance double-precision computer power. This affects certain shader effects mathematics that should use FP64 instead of FP32. Previous video cards usually paid a hefty performance penalty for using FP64.

As a comparison, GTX 680 can offer just shy of 3.1 TeraFLOPS of FP32 performance, 128 GigaTexels of second texturing throughput, 32 GigaPixels second rendering throughput, and 192 gigabytes per second of memory bandwidth. Titan leaps to 4.7 Tera FLOPS of FP32 performance, 187 GigaTexels of second texturing throughput, 40 GgiaPixels of second rendering throughput, and 288 gigabytes per second of memory bandwidth. This gives Titan 46 percent more shader computing and texturing performance, 25 percent more pixel throughput, and a full 50 percent more memory bandwidth than a Nvidia GTX 680.

We can also compare these hardware specs against the rumored Playstation 4 and Xbox 720 hardware specifications. The PlayStation 4 GPU is expected to produce 1.84 TeraFLOPS. At this time the GPU of the Microsoft Xbox 720 is unknown but sources at CES claim the graphics architecture is only capable of outputting 1.23 TeraFLOPS. Other rumors have the two consoles being closer to the 4.7 TeraFLOPS the Geforce Titan can produce, but due to heat, size, and cost requirements it’s highly doubtful that will happen.

A GTX 680 can be purchased for under $500, so it could be argued that since the Geforce Titan will be $999 and available in limited quantities that it might be better to buy two GTX 680’s or 690’s and use them in SLI. But some people have complained about compatibility issues with SLI including microstuttering and rendering errors. Never mind that you have to purchase a more expensive power supply, motherboard, and a PC case that’ll fit multiple large video cards. So if you don’t already have a SLI setup then the price difference might be a tossup.

Do you plan to pick up the GeForce GTX Titan when it is finally released?