NVIDIA’s GTX 650 Ti and GTX 660 may be popular graphics cards in their own respective markets; there is a significant performance and price difference between them. The new GTX 650 Ti Boost may have a mouthful of a name but, like AMD’s own HD 7790, it is meant to plug one of the most obvious gaps within the GeForce lineup with a card that can offer high framerates at 1080P without costing more than $175. Somewhat ironically, this is exactly what AMD’s newest card sets out to accomplish but NVIDIA’s approach is very different. We hinted at a green team broadside in the HD 7790 review and this is it.
Instead of neatly slotting the GTX 650 Boost into a convenient position without modifying their current lineup’s pricing structure, NVIDIA will be shuffling things around a bit. The GTX 650 Ti Boost will start at $169 for the 2GB version and $149 for an upcoming 1GB SKU while incorporating $75 of in game credits for certain Free to Play titles. Naturally, this necessitates some movement for the GTX 650 Ti which is being lowered to a mere $129 and the GTX 650 goes to $109 while even the GTX 660 2GB has moved to a new $199 level.
With NVIDIA suddenly targeting areas where AMD previously dominated, this causes a world of issues for the Radeon product stack. They just introduced their HD $149 7790, a card which competes quite well against the GTX 650 Ti but suddenly finds itself being an overpriced outsider. This isn’t an envious situation for card which hasn’t even become available to purchase yet, while NVIDIA's latest addition will be hard launched today. To make matters even worst, the GTX 650 Ti Boost is actually meant to compete against the HD 7850, a GPU that currently goes for about $15 to $20 more.
The GTX 650 Ti Boost uses the same 28nm GK106 core as its sibling, the GTX 650 Ti, and has one of its SMs disabled as well but that’s where the similarities with the lower end version come to an end. Unlike the standard GTX 650 Ti, this one makes use of three 64-bit memory controllers for a 192-bit wide interface. Since the 1:1 ratio between ROPs and memory controllers remains in place, this also allows for an increase in ROPs to 24.
While the memory layout will grant the GTX 650 Ti Boost a significant bandwidth advantage over its predecessor, NVIDIA’s addition of GeForce Boost into this equation is the real differentiating factor. Boost gives the engine clock an ability to run at much higher frequencies provided there’s enough thermal and power overhead to do so. The previous version of the GTX 650 Ti didn’t have this ability, meaning it was constrained to its reference clock at all times.
The GTX 650 Ti Boost takes some clock speed mojo from NVIDIA’s GTX 660 and melds it together with a few tricks out of the GTX 650 Ti’s architecture. This leads to its base, boost and memory frequencies mirroring those found on the $199 card while the actual CUDA core layout uses the same 768 design as its less expensive brother. The clock speeds on our sample routinely hit the 1100MHz mark when taking advantage of GeForce Boost’s headroom. The resulting Frankenstein-like creation melds the best of both worlds without stepping on the feet of any other NVIDIA cards. Expect board partners’ products to hit a number of price points and clock speeds above these marks.
Mixed memory sizes also make a comeback with the three controllers being in charge of 1GB or 2GB. The first GTX 650 Ti Boost SKU being introduced will be the 2GB version while the 1GB product should be available sometime in the first two weeks of April. We’ll have a review of the 1GB in the coming weeks but expect it to be a perfect fit for the 1080P crowd.
Due to its clock speeds, memory allotment and the use of a partially disabled GK106 core, the Boost’s TDP comes in closer to the GTX 660. This shouldn’t be too much of an issue since 140W is still low enough to ensure its use in SFF systems while still retaining a good perf per watt ratio in the mid-range market.
NVIDIA’s reference GTX 650 Ti Boost looks exactly like the GTX 660 and GTX 660 Ti with a push / pull type heatsink configuration and a long black shroud. It also uses a single 6-pin power connector along with SLI compatibility which is something the non-Boost version doesn’t have. Just don’t expect too many retail products to reflect this configuration since most will come with custom heatsinks.
Flipping the card over we can see that NVIDIA retained their ridiculously small PCB, hopefully allowing board partners to created designs which are tailored to the HTPC and SFF market. We can also see that NVIDIA has added a pair of slits behind the fan overhang, ensuring the cooling assembly is never starved for airflow.
Instead of neatly slotting the GTX 650 Boost into a convenient position without modifying their current lineup’s pricing structure, NVIDIA will be shuffling things around a bit. The GTX 650 Ti Boost will start at $169 for the 2GB version and $149 for an upcoming 1GB SKU while incorporating $75 of in game credits for certain Free to Play titles. Naturally, this necessitates some movement for the GTX 650 Ti which is being lowered to a mere $129 and the GTX 650 goes to $109 while even the GTX 660 2GB has moved to a new $199 level.
With NVIDIA suddenly targeting areas where AMD previously dominated, this causes a world of issues for the Radeon product stack. They just introduced their HD $149 7790, a card which competes quite well against the GTX 650 Ti but suddenly finds itself being an overpriced outsider. This isn’t an envious situation for card which hasn’t even become available to purchase yet, while NVIDIA's latest addition will be hard launched today. To make matters even worst, the GTX 650 Ti Boost is actually meant to compete against the HD 7850, a GPU that currently goes for about $15 to $20 more.

The GTX 650 Ti Boost uses the same 28nm GK106 core as its sibling, the GTX 650 Ti, and has one of its SMs disabled as well but that’s where the similarities with the lower end version come to an end. Unlike the standard GTX 650 Ti, this one makes use of three 64-bit memory controllers for a 192-bit wide interface. Since the 1:1 ratio between ROPs and memory controllers remains in place, this also allows for an increase in ROPs to 24.
While the memory layout will grant the GTX 650 Ti Boost a significant bandwidth advantage over its predecessor, NVIDIA’s addition of GeForce Boost into this equation is the real differentiating factor. Boost gives the engine clock an ability to run at much higher frequencies provided there’s enough thermal and power overhead to do so. The previous version of the GTX 650 Ti didn’t have this ability, meaning it was constrained to its reference clock at all times.

The GTX 650 Ti Boost takes some clock speed mojo from NVIDIA’s GTX 660 and melds it together with a few tricks out of the GTX 650 Ti’s architecture. This leads to its base, boost and memory frequencies mirroring those found on the $199 card while the actual CUDA core layout uses the same 768 design as its less expensive brother. The clock speeds on our sample routinely hit the 1100MHz mark when taking advantage of GeForce Boost’s headroom. The resulting Frankenstein-like creation melds the best of both worlds without stepping on the feet of any other NVIDIA cards. Expect board partners’ products to hit a number of price points and clock speeds above these marks.
Mixed memory sizes also make a comeback with the three controllers being in charge of 1GB or 2GB. The first GTX 650 Ti Boost SKU being introduced will be the 2GB version while the 1GB product should be available sometime in the first two weeks of April. We’ll have a review of the 1GB in the coming weeks but expect it to be a perfect fit for the 1080P crowd.
Due to its clock speeds, memory allotment and the use of a partially disabled GK106 core, the Boost’s TDP comes in closer to the GTX 660. This shouldn’t be too much of an issue since 140W is still low enough to ensure its use in SFF systems while still retaining a good perf per watt ratio in the mid-range market.

NVIDIA’s reference GTX 650 Ti Boost looks exactly like the GTX 660 and GTX 660 Ti with a push / pull type heatsink configuration and a long black shroud. It also uses a single 6-pin power connector along with SLI compatibility which is something the non-Boost version doesn’t have. Just don’t expect too many retail products to reflect this configuration since most will come with custom heatsinks.

Flipping the card over we can see that NVIDIA retained their ridiculously small PCB, hopefully allowing board partners to created designs which are tailored to the HTPC and SFF market. We can also see that NVIDIA has added a pair of slits behind the fan overhang, ensuring the cooling assembly is never starved for airflow.
Last edited: