We have the confirmed specifications of upcoming GeForce RTX 2080 series.
NVIDIA RTX 2080 Series: 4352 and 2944 CUDA cores
What we know: GeForce RTX 2080 Ti packs 11GB and 352-bit memory (just as GTX 1080 Ti). What is new is the GDDR6 memory clocked at 14 Gbps. This increased the maximum theoretical bandwidth to 616 Gb/s (that’s 132 Gb/s more).
We can confirm that GeForce RTX 2080 Ti features 4352 CUDA cores, that’s fewer than Quadro RTX 6000/8000, which both feature 4608 unified shaders.
The GeForce RTX 2080 is an 8GB 256-bit model. It is equipped with 2944 CUDA cores, not 3072 as previously speculated. This means GeForce RTX 2080 lost 128 CUDA cores from TU104 GPU compared to Quadro RTX 5000.
We also have clocks and TDP values, but since the origin of the data may suggest these are non-reference models, we will release this information later.
| NVIDIA RTX Series | ||||
|---|---|---|---|---|
| CUDA Cores | Memory | Memory Bus | Memory Bandwidth | |
| Quadro RTX 6000 | ||||
| GeForce RTX 2080 Ti | ||||
| GeForce GTX 1080 Ti | ||||
| Quadro RTX 5000 | ||||
| GeForce RTX 2080 | ||||
| GeForce GTX 1080 | ||||
Here’s another model you did not see before. This is not AORUS card, so likely just called GAMING (OC):
