![](http://images.eurogamer.net/2013/articles/1/6/5/5/8/8/3/139272561864.jpg/EG11/resize/405x-1)
So what does Maxwell bring to the table? Remarkably, Nvidia claims to have doubled performance per watt compared to its previous Kepler technology, thanks to a combination of factors where a new CUDA core configuration and a massive boost in L2 memory cache (2MB vs. 256K) appear to provide the most tangible benefits. The big news here is that Nvidia has achieved this without moving to a new fabrication node - just like Kepler, Maxwell is a 28nm chip. It's usually the case that new GPU technologies are rolled out alongside new manufacturing technologies, but the new 20nm process still isn't ready for show-time. Indeed, Nvidia has indicated to us that we won't see any 20nm cards this year, but that's OK - Maxwell isn't just about improved performance per watt, it's about getting more processing power from a smaller area of die-space too.
You don't quite realise just how much of an achievement Maxwell actually is until you first hold GTX 750 Ti in your hands. The GM107 chip on the new card is tiny, with a relatively minuscule heatsink and fan, oddly reminiscent of the stock coolers supplied in the box with Intel CPUs. The board itself is very, very small - about the size of an old Soundblaster audio card, and it requires no additional PCI Express power cables from the power supply, making it an ideal upgrade for cheaper PCs lacking the necessary PSU connections for enthusiast gaming.
Read more…
More...