Today NVIDIA announced two professional graphics cards based on the Maxwell Graphics core, the Quadro M5000 and M4000 Quadro, both graphics cards share the Maxwell GM204 although with minor differences.

NVIDIA launches Quadro M5000 and Quadro M4000

DWReal

The NVIDIA Quadro M5000 uses Maxwell GM204 with CUDA Cores 2048 , 128 TMUs and 64 ROPs at a frequency of1050 MHz  accompanied  8GB GDDR5 @ 6.60 GHz memory interface coupled with 256 bits  which gives a bandwidth of 211 GB / s  with floating point performance of 4.3 TFLOPS  leading to a consumption of 150W.

Furthermore, the  NVIDIA Quadro M4000 uses a cut down / harvested version of Maxwell GM204 with CUDA Cores 1664 , 104 TMUs and 56 ROPsat a frequency of 800 MHz, which keeps 8 GB GDDR5 but @ 6.00 GHz together with memory interface  256 bits  gives a bandwidth of 192 GB / s  with floating point performance of 2.6 TFLOPS.

The cut down Graphics core help it in lowering the power consumption to the 120W, allowing the use of a cooling system that occupies a single PCI expansion slot  (Vs two of the M5000). The price of both professional GPUs is not announced.

NVIDIA Quadro Specification Comparison
Quadro M5000Quadro M4000Quadro K5200Quadro K4200
CUDA Cores2048166423041344
Boost Clock~1050MHz~780MHz650MHz780MHz
Memory Clock6.6GHz GDDR56GHz GDDR56GHz GDDR55.4GHz GDDR5
Memory Bus Width256-bit256-bit256-bit256-bit
VRAM8GB8GB8GB4GB
FP641/321/321/241/24
TDP150W120W150W105W
GPUGM204?GM204GK110GK104
ArchitectureMaxwell 2Maxwell 2KeplerKepler
SizeDouble-SlotSingle-SlotDouble-SlotSingle-Slot
4Kp60 Displays Supported4422