Gainward GTX Titan
Under the Hood
Let us refresh your memory for a moment. On the left is the new GK110 chip that beats at the heart of the Titan. On the right, the GTX680. So say this is a monumental leap in computing power is understating things. In every department the Titan lives up to its name. 48 Raster Processors compared to the 32 on the GTX680. 6GB of GDDR5 compared to 2GB. 1152 more CUDA Cores. In every conceivable number the Titan dwarfs its predecessor.
So what does all this mean? A good place to start is the GigaFlops number. A measure of pure calculative performance the Kepler GK110 has the capability to perform 4,500 billion floating point operations per second. If that doesn't seem like a lot to you, then compare it to the current CPU king, the Intel Core i7-3960X, with its measily 316 billion, and the insane power of the nVidia effort starts to become clear.
For comparison, the HD7970 has 3800 GigaFlops and 4.3 billion transistors..
There are a lot of tweaks and clever things going on beneath the hood of the GK110, but we've never been ones to baffle you with meaningless science here at OC3D. Certainly not for the sake of just making our reviews needlessly lengthy. You're busy people. What we care about is the end product and how fast it is, and we're sure you're the same. So perhaps the biggest change to the way that the Titan works is the adjustment in the GPU Boost. Whereas the GTX6xx series adjusted the clockspeed of the GPU relative to the amount of power that was being drawn, the GPU Boost 2.0 on the GTX Titan takes the temperature of the card as the limiting factor.
This might seem like a small change, but if the thermals are the only limit, and you can water cool them or even go sub-zero, then hopefully you're all smiling like us at the practical applications of this tweak. Lastly nVidia believe that the Titan should give a substantial performance boost over its predecessor, but that's what we're here for.
So enough preamble, let's take a look at today's card, the Gainward GTX Titan.