Associate
- Joined
- 22 Jun 2018
- Posts
- 1,763
- Location
- Doon the watah ... Scotland
I can’t help but think that a lot of gpu die space has been given over to an as yet unused feature in graphics which for a lot of the time may be going unused in general.
Why not have the RT cores on a separate PCIe card to the GPU so that if you wanted rt compute, you use that, leaving the gpu cards more space to improve over previous gen.
The quadro announce discussed how 2 cards could be paired to share resource which implies the data can be passed between card in the necessary timeframes. so surely that idea could be the means to add rt to gpu separately?
Why not have the RT cores on a separate PCIe card to the GPU so that if you wanted rt compute, you use that, leaving the gpu cards more space to improve over previous gen.
The quadro announce discussed how 2 cards could be paired to share resource which implies the data can be passed between card in the necessary timeframes. so surely that idea could be the means to add rt to gpu separately?