ASUS GTX 750 Ti OC 2 GB Review 21

ASUS GTX 750 Ti OC 2 GB Review




Here it is, NVIDIA's GeForce GTX 750 Ti, one of the first GPUs based on its next-generation "Maxwell" GPU architecture. First reports of the chip being based on "Maxwell" hit us by surprise as we presumed that a new micro-architecture is invariably pegged to NVIDIA's foundry partner, TSMC, to launch its next-generation 20 nanometer silicon fabrication process. The fact that TSMC's 20 nm node isn't in a position to ship out such complex ASICs as graphics processors right now and that NVIDIA wouldn't want its new architecture to in any way be hit by unforeseen issues arising out of the move to a new process may have led to the decision to use the existing 28 nanometer process.

The GeForce GTX 750 Ti and GTX 750 (also launched today) are based on the 28 nm GM107 (GeForce-Maxwell 107) silicon. This 1.87 billion-transistor GPU features a basic component hierarchy that's not too different from that of "Kepler," but sees some under-the-hood changes in the design of its key parallel processing sub-unit, the streaming multiprocessor "Maxwell" (SMM). The GTX 750 Ti is positioned between the GTX 760 and GTX 660, which is a sizable gap to fill. NVIDIA is promising some huge performance-per-watt gains with "Maxwell." Given that the GM107 is based on the same 28 nm process as the GeForce Kepler series, there's only one way NVIDIA can deliver on its promise - by developing a better overall architecture.

Today we are reviewing the ASUS GTX 750 Ti OC. This will be our first review of a GTX 750 Ti with a 6-pin power connector.

ASUS has also overclocked the card out of the box and uses their DirectCU dual-fan thermal solution on the card, but without heatpipes.

The card is currently available online for $160, which is a $10 price premium over the reference design.

ASUS GTX 750 Ti Market Segment Analysis
GTX 650 Ti
HD 7790
HD 7850
GTX 750
R7 260X
650 Ti Boost
GTX 750 Ti
750 Ti OC
R7 265
GTX 660
HD 7870
R9 270X
GTX 660 Ti
GTX 760
Shader Units768896102451289676864064010249601280128013441152
Graphics ProcessorGK106BonairePitcairnGM107BonaireGK106GM107GM107PitcairnGK106PitcairnPitcairnGK104GK104
Memory Size1024 MB1024 MB2048 MB1024 MB2048 MB2048 MB2048 MB2048 MB2048 MB2048 MB2048 MB2048 MB2048 MB2048 MB
Memory Bus Width128 bit128 bit256 bit128 bit128 bit192 bit128 bit128 bit256 bit192 bit256 bit256 bit192 bit256 bit
Core Clock925 MHz1000 MHz860 MHz1020 MHz+1100 MHz980 MHz+1020 MHz+1072 MHz+925 MHz980 MHz+1000 MHz1050 MHz915 MHz+980 MHz+
Memory Clock1350 MHz1500 MHz1200 MHz1250 MHz1625 MHz1502 MHz1350 MHz1350 MHz1400 MHz1502 MHz1200 MHz1400 MHz1502 MHz1502 MHz


NVIDIA's primary design goal with the GM107 is to up the performance-per-watt game. If NVIDIA achieves a significant performance-per-watt gain with "Maxwell" over "Kepler," it can trade that gain for performance on bigger "Maxwell" chips bolstered by the 20 nm process. It's this fact that makes the GK107 academically interesting.

As we mentioned earlier, the GM107 essentially features the same component hierarchy as previous generation "Kepler" GPUs, but introduces changes to the design of the streaming multiprocessor (SMM), the parallel-processing sub-unit of the GPU. At its outermost ring, the GM107 features the GigaThread Engine, a component that marshals data and instructions between the graphics processing cluster (GPC), the raster operations processors (ROPs), the L3 cache, the memory controllers, the bus interface, and the display I/O.

Several GPCs can typically be routed to a GigaThread engine, but being a mid-range GPU, the GM107 features just one. This GPC features a raster engine that handles high-level assembly of data and instructions and five streaming multiprocessors (SMMs), which is where the number crunching takes place. Unlike the streaming multiprocessors (SMXs) of "Kepler" GPUs, which feature an incoherent group of 192 CUDA cores, the SMM features four groups of 32 cores each, which totals 128 per SMM. The SMM shares a Polymorph Engine that features components such as the tessellator, fetch, setup, transform, and stream output with the four groups. The four groups of 32 CUDA cores, each, feature dedicated warp schedulers and registers, with a texture cache cushioning transfers between the groups and TMUs. The GM107 hence features a total of 640 CUDA cores and 48 TMUs. At a higher level, the chip features 16 color ROPs and a 128-bit wide GDDR5 memory interface.


Package Front
Package Back


You will receive:
  • Graphics card
  • Driver CD + documentation
Our Patreon Silver Supporters can read articles in single-page format.
Discuss(21 Comments)