The company's OEMs release their own-brand models as well

Feb 18, 2014 14:20 GMT  ·  By

As we've been saying it would happen for the past several weeks, or months really, NVIDIA has finally introduced the Maxwell architecture, even if it postponed the real high-end chip until fall.

The company has launched the GeForce GTX 750, based on the GM107-300 GPU, and the GeForce GTX 750 Ti, powered by the GM107-400.

Maxwell promises 35% better performance per core than Kepler, as well as a twofold improvement to performance/watt rating.

Also, compared to Fermi, the power efficiency is four times better, which means that the efficiency rating has gone up by a factor of four in as many years.

But, in the end, DIY customers are mostly interested in the specifications of the graphics cards, so we suppose we may as well provide them.

Or confirm them really, since rumors and leaks made sure to expose most of the information by last week.

First off, the GeForce GTX 750 Ti has 640 CUDA cores (GM107-400), two 64-bit memory controllers (for a total 128-bit interface) and either 1 GB or 2 GB GDDR5 VRAM operating at 5.4 GHz. Most of NVIDIA's partners have chosen 2 GB.

GPU Boost 2.0 technology is included, which means that while the base clock is of 1020 MHz, the card can jump to 1085 MHz when needed.

Further overclocking is possible via whatever software OEMs include in the driver. NVIDIA said in its press brief that it managed to push the card to 1271 MHz.

The GeForce GTX 750 has only 512 CUDA cores (GM107-300 GPU), but the same memory interface. 1 GB of GDDR5 is available.

The reference clocks are of 1020 MHz and 1085 MHz like on the GTX 750 Ti, but the lower number of CUDA cores means lower overall performance despite that.

The GTX 750 (55W TDP) and 750 Ti (60W) will replace the GTX 650 Ti. Their prices should be of around $120 / €120 and $150 / €150.

Photo Gallery (2 Images)

NVIDIA GeForce GTX 750 Ti
NVIDIA GeForce GTX 750
Open gallery