It has been less than a week since a major GPU launch, AMD's Radeon HD 7790, and we already have one from NVIDIA, the GeForce GTX 650 Ti Boost. When AMD announced the Radeon HD 7950 Boost Edition in August 2012, reacting to stiff competition from NVIDIA's GeForce GTX 660 Ti, there wasn't much to lose if a consumer overlooked the "Boost" component while making the purchase, just higher clock speeds and PowerTune with Boost. The GeForce GTX 650 Ti Boost, launched today, is an extremely different beast to the original GTX 650 Ti it shares most of its name with.
NVIDIA launched the original GeForce GTX 650 Ti in October 2012 to capture the $150 price-point. The card comprehensively beat AMD's $120 Radeon HD 7770 GHz Edition in performance, which forced AMD to carve a $180 1 GB variant out of its Radeon HD 7850 that is armed with 2 GB of memory for around $210. Even as AMD carried on for six months with a gaping hole between the HD 7770 and HD 7850—a hole filled by the HD 7790 just last week—NVIDIA dug itself a hole of its own between the $150 GeForce GTX 650 Ti and the GeForce GTX 660, giving the two HD 7850 variants a free hand in the sub-$200 market, but the company has now rolled out the GeForce GTX 650 Ti Boost to fill exactly that gap.
Strategically priced at $170 for the 2 GB variant, with the possibility of a $150 1 GB variant in the near future, the new GeForce GTX 650 Ti Boost has its eyes set on rattling AMD's sub-$200 stronghold. So what exactly gives the "Boost" superhero powers over the GTX 650 Ti? The answer lies in the silicon. The original GTX 650 Ti uses a significantly cut down version of the 28 nm GK106 silicon. It only features 80 percent of the chip's available streaming multiprocessors, a 33% narrower memory bus, unimpressive sub-gigahertz GPU clock speed, and lacks GPU Boost—a well-implemented dynamic overclocking technology. The GeForce GTX 650 Ti Boost doesn't tinker with the streaming multiprocessor count, but utilizes the full 192-bit wide GDDR5 memory interface, GPU Boost with core clock speeds crossing the 1 GHz mark far too often, and a memory bandwidth of 144 GB/s. These three subtle changes, along with a PCB/cooler design similar to that of the GeForce GTX 660, gives the card a more premium-feel in the sub-$200 market.
The GeForce GTX 650 Ti Boost, based on the "Kepler" GPU micro-architecture, features four of the silicon's five streaming multiprocessors, which amounts to 768 CUDA cores, 64 texture memory units (TMUs), 24 raster operations processors (ROPs), and a 192-bit wide memory interface. 2 GB is the standard memory amount for now. The core features a nominal clock speed of 980 MHz that can boost itself to 1033 MHz; most real-world gaming and creativity apps should have no problems letting the chip run at that speed. The memory is clocked at 6008 MHz (GDDR5-effective), which translates into a decent 144 GB/s memory bandwidth.
In this review, we test the ZOTAC GTX 650 Ti Boost, which comes with a tiny 13 MHz overclock on the GPU and reference memory clocks. Instead of the NVIDIA heatsink, ZOTAC chose to use their dual-fan cooler we've seen on other designs from the company. What's great is that their card comes at no price premium. It, like the reference design, costs $169.