More Maxwell GPUs
Nvidia today added a couple of new professional level graphics cards to its Quadro family, the Quadro M4000 and M5000. Both cards are based on Nvidia's Maxwell 2 architecture, the same as found in its flagship Quadro M6000 card along with consumer desktop solutions like the GeForce GTX 980.
The Maxwell 2-based M4000 (PDF) and M5000 (PDF) replace Nvidia's previous generation Quadro K4200 and K5200, both of which are Kepler cards. While the new parts aren't as powerful as the M6000, they tote the same architectural upgrades over the Kepler-based Quadro cards.
Starting with the higher end of the two, the double-slot M5000 sports 2,048 CUDA cores and 8GB of GDDR5 memory on a 256-bit bus for 211GB/s of memory bandwidth. It has a boost clockspeed of around 1,050MHz while the memory runs at 6,600MHz. NVIDIA rates FP32 performance at 4.3 TFLOPs. Finally, it has a 150W TDP.
The M4000 is a single-slot solution with 1,664 CUDA cores. It also has 8GB of GDDR5 memory on a 256-bit bus, but it's clocked a 6,000MHz for 192GB/s of memory bandwidth. It has a boost clockspeed of around 780MHz and 120W TDP. FP32 performance is listed at 2.6 TFLOPs.
There's software-based error correction code (ECC) for the memory on the M5000 model for added reliability. Both cards have four DisplayPort 1.2 connectors and can support up to four 4K displays (the M5000 also adds a DVI-I connector to the mix).
Nvidia didn't say how much the cards will cost or when they'll be available.
Follow Paul on Google+, Twitter, and Facebook
From maximumpc
from http://bit.ly/1JaJqHc
