NVIDIA announces first Pascal-based graphics card with staggering 150 billion transistors (the full package).
TESLA P100 has 150 billion transistors
While 150 billion transistors seems like a large number, this is not just the GPU. Total number of transistors comes from GPU, HBM and interposer. The GPU itself only has 15.3 billion transistors..
TESLA P100 becomes the first Pascal GPU and the first HBM2 based graphics card.
NVIDIA decided to go full throttle on new FinFET 16nm architecture by designing the biggest possible chip.
TESLA P100 has die size of 600m2, which is exactly how I predicted in my last article about Pascal. It’s the most complex project NVIDIA has ever built.
The chip has some amazing specs and performance.
- 21.2 TFLOPs FP16
- 10.6 TFLOPs FP32
- 5.3 TFLOPs FP64
Jen-Hsun Huang announced that P100 is in volume production today. This chip will not reach OEM market sooner than Q1 2017, so (G)P100 based might not launch as soon as we expect.
The exact specifications were not revealed: CUDA Cores, clock speed, memory capacity. Tesla P100 has 3584 CUDA cores, which gives us 10.6 TFLOP FP32 performance.
| NVIDIA Pascal | ||
|---|---|---|
| Graphics Processor | NVIDIA GP100 (Tesla P100) | NVIDIA GM200 |
| Die Size | 610 m2 | 601 mm2 |
| FP32 Computing Performance | 10.6 TF | 6.1 TF |
| CUDA Cores | 3584 | 3072 |
| Memory Type | up to 16GB HBM2 | up to 24GB GDDR5 |
