In the wake of the recent reviews on the just released GTX 1080, the full specifications of the GTX 1070 have been posted on NVIDIA’s website. Just as suspected, the card will be a substantial update over the GTX 970 and will not have the same VRAM allocation issue that plagued the 970.
GTX 1070 is three times faster than its predecessor
The 1070 has solid specifications at first glance. 8GB of ‘normal’ GDDR5 working at near the highest rate possible for standard GDDR5 feeding 1920 Pascal-based CUDA cores. Interestingly, it is completely compatible with the new HB SLI bridge that was introduced at their event. It was alluded that the new SLI bridge would be reserved and compatible only with the GTX 1080, that’s untrue as according to the specifications it works perfectly.
NVIDIA GeForce GTX 1070 |
||
---|---|---|
GeForce GTX 1070 | GeForce GTX 970 | |
Architecture | Pascal | Maxwell |
CUDA Cores | 1920 | 1664 |
VRAM | 8GB GDDR5 | 4GB GDDR5 |
VRAM Speed | 8Gbps | 7Gbps |
Base Clock | 1506MHz | 1050MHz |
Boost Clock | 1683MHz | 1178MHz |
TDP | 150W | 145W |
Price | $379 | $329 |
As the top-end of the mid-range it’s also claimed to be 10% faster than the Titan X in some workloads. If we extrapolate based off of the performance numbers of the GTX 1080, then it might not be so possible, unless the difference in performance between the two new cards is small. Nonetheless, it should still be on par with a 980 Ti, if not slightly faster in most situations.
NVIDIA themselves quote the GTX 1070 being almost 1.75 times faster than the GTX 970 in Rise of the Tomb Raider and The Witcher 3: Wild Hunt, which is not an easy feat. The lack of complete asynchronous compute in the new Pascal architecture is still somewhat worrisome, though initial results show a small increase in framerate in Ashes of the Singularity while showing a slight drop in performance when switching to DirectX 12 in Hitman. Regardless, the actual improvements and price could make it one of the hottest selling GPUs of this generation.