Deep Learning at Novatech - Deep Learning GPUs / Graphics Cards

Deep learning is a complex field that requires intense computational power in order to run a powerful system. The correct GPU's are fundamental in determining what kind of deep learning system you are empowering and how you utilise its abilities for your industry. As a user, a GPU-accelerated system will run much faster, providing the results that you have been looking for.

With our intensive array of powerful graphics processors, you can power a deep learning network to run experiments and advanced analyses without delay. What once took months is now whittled down to a few days; days have been transformed into easy hours. With the right GPU for your system, hours of computational power are now dissolved into a few minutes.

Finding the right GPU is critical. Luckily, Novatech is here for you every step of the way. Find out which GPU is right for you here:

Popular Deep Learning GPUs

GPU Product	Titan XP	Titan V	Quadro GP100	Tesla K40	Tesla M40	Tesla P100	Tesla V100
GPU	GP102 (Pascal)	GV100 Volta	GP100 (Pascal)	GK110 (Kepler)	GM200 (Maxwell)	GP100 (Pascal)	GV100 Volta
GPU Boost Clock	1582 MHz	1455 MHz	1480 MHz	810/875 MHz	1114 MHz	1480 MHz	1455 MHz
Peak Half-Precision Performance (FP16)	0.2 teraFLOPs	30 TBC	21 teraFLOPs	N/A	N/A	21 teraFLOPs	125 teraFLOPs
Peak Single-Precision Performance (FP32)	12 teraFLOPs	15 teraFLOPs	10 teraFLOPs	5.04 teraFLOPs	6.8 teraFLOPs	10.6 teraFLOPs	15.7 teraFLOPs
Peak Double-Precision Performance (FP64)	0.4 teraFLOPs	7.5 teraFLOPs	5.2 teraFLOPs	1.4 teraFLOPs	0.2 teraFLOPs	5.4 teraFLOPs	7.8 teraFLOPs
Memory Interface	384-bit GDDR5X	3072-bit HBM2	4096-bit HBM2	384-bit GDDR5	384-bit GDDR5	4096-bit HBM2	4096-bit HBM2
Memory Bandwidth	548GB/sec	653GB/sec	717GB/sec	288GB/sec	288GB/sec	Max 720GB/sec	900GB/sec
Memory Size	Max 12GB	Max 12GB	Max 16GB	Max 12GB	Max 24GB	16GB	16GB
TDP	250 Watts	250 Watts	235 Watts	235 Watts	250 Watts	300 Watts	300 Watts
Transistors	21.1 Billion	21.1 Billion	15.3 Billion	7.1 Billion	8 Billion	15.3 Billion	21.1 Billion
GPU Die Size	471 mm2	815 mm2	610 mm2	551 mm2	601 mm2	610 mm2	815 mm2
Manufacturing Process	TSMC 16nm + FinFET	12nm +FFN	16nm + FinFET	28 nm	28 nm	16 nm + FinFET	12 nm FFN

Quadro P400

Download Datasheet

256 Cuda cores
Single Precision : Up to 0.641 TFLOPs Floating point performance
2 GB GDDR5 / 64-bit / Up to 32 GB sec
Power consumption 30W Max
3x mDP Connections

Quadro P600

Download Datasheet

384 Cuda cores
Single Precision : Up to 1.195 TFLOPs Floating point performance
2 GB GDDR5 / 128-bit / Up to 64GB sec
Power consumption 40W Max
4x mDP Connections

Quadro P1000

Download Datasheet

640 Cuda cores
Single Precision : Up to 1.894 TFLOPs Floating point performance
4 GB GDDR5 / 128-bit / Up to 80GB sec
Power consumption 47W Max
4x mDP Connections

Quadro P2000

Download Datasheet

1024 Cuda cores
Single Precision : Up to 3 TFLOPs Floating point performance
5 GB GDDR5 / 160-bit / Up to 140GB sec
Power consumption 75W Max
DP 1.4 (4) Connections

Quadro P4000

Download Datasheet

1792 Cuda cores
Single Precision : Up to 5.3 TFLOPs Floating point performance
8 GB GDDR5 / 256-bit / Up to 243GB sec
Power consumption 105W Max
DP 1.4 (4) Optional Stereo (1) Connections

Quadro P5000

Download Datasheet

2560 Cuda cores
Single Precision : Up to 8.9 TFLOPs Floating point performance
16 GB GDDR5 / 256-bit / Up to 288GB sec
Power consumption 180W Max
DP 1.4 (4) + DVI-D DL (1) + Stereo Connections

Quadro P6000

Download Datasheet

3840 Cuda cores
Single Precision : Up to 12 TFLOPs Floating point performance
24 GB GDDR5 / 384-bit / Up to 432GB sec
Power consumption 250W Max
DP 1.4 (4) + DVI-D DL (1) + Stereo Connections

Quadro GP100

Download Datasheet

3584 Cuda cores
Single Precision : 10.3 TFLOPS / Double Precision : 5.2 TFLOPS / Peak Half Precision : 20.7 TFLOPS Floating point performance
16GB HBM2 / 4096-bit / Up to 717GB sec
Power consumption 235W Max
DP 1.4 (4) + DVI-D DL (1) + Stereo Connections

Quadro GV100

Download Datasheet

5120 Cuda cores
640 Tensor cores
Single Precision 14.8TFLOPS / Double Precision 7.4TFLOPS Floating point performance
32GB HBM2 - 870Gb/s memory bandwidth
Power consumption 250W Max
4x DP1.4 Connections

Tesla P4

Download Datasheet

2560 Cuda cores
Single Precision : 5.5TFLOPS Floating point performance
22 TOPS (Tera-Operations per Second)
8GB / 256Bit / 192GB sec
Power consumption 50W/75W Max

Tesla P40

Download Datasheet

3840 Cuda cores
Single Precision : 12TFLOPS Floating point performance
47TOPS (Tera-Operations Per Second)
24GB / 384bit / 346GB sec
Power consumption 250W Max

Tesla P100

Download Datasheet

3584 Cuda cores
Single Precision : 9.3TFLOPS / Double Precision : 4.7TFLOPS / Half Precision : 18.7TFLOPS Floating point performance
12 or 16 GB HBM2 - 4096-bit (16GB 732GB/s – 12GB 549GB/s)
Power consumption 250W Max

Tesla V100

Download Datasheet

5120 Cuda cores
640 Tensor cores
Single Precision : 14TFLOPS / Double Precision : 7TFLOPS / Tensor Performance : 112TFLOPS Floating point performance
16 GB HBM2 / 900 GB sec
Power consumption 250W Max

Tesla M60

Download Datasheet

4096 (2048 per GPU) Cuda cores
Single Precision : 8 Tflops (GPU Boost Clocks) Floating point performance
16 GB GDDR5 (8 GB per board) / 256-bit x 2 / 160 Gb/s x 2
Power consumption 300W Max

Tesla M40

Download Datasheet

3072 Cuda cores
7 Tflops (GPU Boost Clocks) Floating point performance
12 GB GDDR5 / 384-bit / 288 GB/s
Power consumption 250W Max

Tesla M10

Download Datasheet

2560 (640 per GPU) Cuda cores
32 GB GDDR5 Memory (8 GB per GPU)
Power consumption 225W Max

Tesla K40

Download Datasheet

2880 Cuda cores
4.29 teraFLOPs Floating point performance
12 GB GDDR5
Power consumption 235W Max

Titan XP

Download Datasheet

3840 Cuda cores
12 teraFLOPs Floating point performance
12 GB GDDR5X
Power consumption 250W Max
DP 1.43, HDMI 2.0b Connections

Titan V

5120 Cuda cores
15 teraFLOPs Floating point performance
12GB HBM2
Power consumption 250W Max
3 x DisplayPort, 1 x HDMI Connections