Specifications for the next graphics card in Nvidia’s Kepler-based lineup,the GTX 650 Ti,have leaked. Last week saw the introduction of two new Kepler graphics cards from Nvidia with the GeForce GTX 650 and GTX 660. Those cards filled in the massive void between the nearly-integrated-graphics-equivalent $99 GT 640 and the GTX 660 Ti at $300. Both cards were alluring for gamers as they offered up decent budget gaming performance for the prices. The GTX 650 Ti will sit between those two cards to fill in the $109 to $229 market. The card is based on the GK106 GPU like the pricier GTX 660,but it makes several compromises to get into the sub-$200 price.

While the GTX 650 and below cards are souped up GK107 cores,the GTX 650 Ti keeps the GK106 core of the GTX 660. As far as Nvidia’s naming conventions go,the GTX 650 Ti should be just under the GTX 660,but the leaked specifications reveal several compromises that suggest the “Ti” variant will be closer in hardware to the GK107-based GTX 650 than the GTX 660.

Mainly,Nvidia has taken the GK106 GPU core and disabled one GPC (Graphics Processing Cluster) and one memory controller. That results in two-fewer SMX units (for a total of three),a smaller 128-bit memory bus,only 48 texture units,and a total of 576 CUDA cores. It is also limited to 1GB of GDDR5 memory. For sake of comparison,the GTX 660 has 960 CUDA cores,all five SMX units,a 192-bit memory bus,80 texture units,and 2GB of GDDR5 memory.Clock speeds are also reduced to a GPU core clock speed of 960MHz,no GPU Boost support,and memory clock speed of 1350MHz. That is a noticeable drop from the 980MHz/1033MHz (boost) GPU clock speed and 1502MHz memory clock speed on the GTX 660. The following table shows all the specifications of the leaked GTX 650 Ti compared against the GTX 650 and GTX 660.

The GTX 650 Ti has a suggested price of between $130 and $220,which would put it right around the $169 mark,and when price is taken into consideration the Ti card’s odd middle-ground pricing reveals many compromises in performance. The card manages to best the $109 GTX 650 in CUDA cores,texture units,and memory bandwidth,but it has a higher TDP to match,and its price is too close to the GTX 660 for the hardware — which is closer to the GTX 650. It is unclear why Nvidia decided to disable so much of the GK106 GPU instead of only disabling a single SMX unit and allowing the card to have two whole GPCs while also (possibly) keeping the 192-bit memory interface. That would have allowed the card to be a nice transitional step between the vanilla 650 and vanilla 660,and the approximate $170 price could much more easily be justified.

Source: www.extremetech.com