I made a bold prediction before the launch of Nvidia's (NVDA) GTX Titan. Despite the fact that this special-edition, ultra-high end, boutique card costs $999 and is seemingly only accessible to the most affluent of gamers - e-tailers can't keep this thing from flying off the shelves. According to nowinstock.com, the GTX Titan is uniformly out of stock across the web, and whenever stock does show up, it is all consumed within an hour as gamers (among others) participate in the F5 Olympics to snatch one (or more) of those cards. In this article, I explain just why demand is so strong and will continue to be for this high ASP, high margin product.
The Gamer's Dream
The GTX Titan is based on Nvidia's "GK110" silicon, the same that was designed for and powers many of the world's supercomputers, including the world's fastest (the "Titan"). It is the world's fastest gaming card according to Anandtech,
No one should be surprised then when we proclaim that GeForce GTX Titan has unquestionably reclaimed the single-GPU performance crown for NVIDIA. It's simply in a league of its own right now, reaching levels of performance no other single-GPU card can touch.
A lot of folks will claim that it's not worth spending $999 on the world's "fastest" GPU since 2x Radeon 7970GE or even 2x GeForce GTX 680 will be faster for the same price, but multi-GPU solutions are a nightmare. Driver compatibility problems, inconsistent frame times, and more heat than most computers can handle. To illustrate the problem (in particular with AMD's (AMD) solution), I appeal to the fine work done at PCPerspective:
See this chart? It's a chart of frame times, or how long it takes to actually render a frame. The Nvidia GTX 680, the 2x GTX 680's in SLI, and the single AMD Radeon 7970 GE all seem pretty consistent. However, the dual Radeon 7970GE is very wildly inconsistent, which means that in "real world" scenarios the dual AMD GPU solution will seem very inconsistent/not smooth, even if it can bust out higher "raw" frames per second.
Nvidia's dual GPU solution is markedly better, but it's still nowhere near as consistent as a single GPU from either AMD or Nvidia, which is why gamers generally prefer a single faster GPU rather than two slower ones combined if possible. They will routinely buy more than one of the highest end card to achieve maximum performance unattainable by a single card. And the folks with the cash to splurge on two $500 cards have no problem buying at least one $999 card.
The best part, though? Gamers with deep pockets who need more performance than the best single GPU can provide will probably buy more than one of these.
The Scientist's Dream
The problem for the gamer boys and girls is that GTX Titan isn't just a gamer card. Sure, it's the fastest one available and the really hardcore folks will be buying as many as they can, but there is a whole different class of folks that would consider the $999 price tag a "steal" for what they are getting. I'm talking about GTX Titan as a high performance floating point accelerator.
According to Anandtech,
If you're doing GPU computing, are invested in CUDA, and need a fast compute card, then Titan is the compute card CUDA developers and researchers have been dreaming of.
In the benchmarks, not only is the card very efficient, but it is also generally faster than anything else by a mile. The appetite for high performance compute accelerators, especially in the scientific/visualization world, is quite voracious. Now, Nvidia sells higher end GK110 parts known as the "K20" with support for additional features like ECC and other virtualization features that customers with stricter requirements and deeper pockets need. But for a scientist who needs the power but doesn't need the additional features on the $2,000+ K20, the Titan is actually a steal at $999.
AMD's decision to focus on smaller, cheaper to produce GPUs is the right one for the company given its condition, but it leaves a lot of room at the very top of the heap for Nvidia to introduce a very high margin, high ASP card essentially unopposed. The gamers who need to have the very best will continue to buy this in droves, and the scientists who want a lot of compute on the cheap will also be buying this in droves.