It will be powered by a GM206 graphics processing unit

Dec 29, 2014 13:43 GMT  ·  By

The GeForce series of graphics cards is in an odd situation right now, expected to not advance that much by 2016. But the GeForce GTX 960 is still set for release in January 2015, so here are the things we know about it from all the reports that reached the web thus far.

The GeForce GTX 960 is expected to be the first video board based on NVIDIA's Maxwell architecture that actually qualifies for the mainstream market. Barely.

True, you might say that the GeForce GTX 750 (Ti) already covers that base. But the GPU acting as their heart, the GM107, was a pretty hasty retrofit of a technology that was supposed to be designed on TSMC's 20nm node.

True, the other Maxwell chips revealed since then have been redesigned for 28nm as well, but NVIDIA could at least have taken some time while it bled Kepler dry.

The GeForce GTX 980 and 970 were the results of Maxwell being properly reengineered for an older, larger fabrication process. And they've been doing well, beating even GTX 780 Ti boards and the Titan, as well as AMD's Radeon R9 series.

The GeForce GTX 960 is expected to cover the bulk of the consumer market, the mainstream as it were, or at least the upper level of that massive customer base.

The GeForce GTX 960

High-end video cards like the GeForce GTX 980 and 780 Ti sell in much smaller numbers than the mid-range cards, despite their superior abilities. The price is the obvious reason for this, as are the power requirements.

The GeForce GTX 960 will only barely fall below the high-end mark, but if it manages to secure a $250 - $300 / €250 - €300 price, it could become quite a bargain.

After all, from all reports and leaks so far, the adapter is set to boast a Maxwell GM206 GPU with 1,280 CUDA cores and a memory interface of 192 bits.

True, it's possible that the chip will actually have a 128-bit interface and 2 GB of GDDR5 VRAM backing it, rather than 3 GB, but even then the performance should be considerable for its product class.

The CUDA cores would be stacked in 10 Maxwell Streaming Multiprocessors (SMM), while power would be provided through two 6-pin ports. Without the advantages of 20nm or 16nm transistor size, and the power efficiency that comes with it, we hesitate to hope for a single-port power delivery.

Availability

Since the GeForce GTX 960 is expected to come out during the CES 2015 trade show (Consumer Electronics Show, January 6-9), the sales should begin immediately afterwards, or maybe a week or so later.