Caporegime
- Joined
- 9 Nov 2009
- Posts
- 25,767
- Location
- Planet Earth
At least the RTX3080 is based on the larger die this generation. Which is a good move.
TBF,if you compared the Nvidia flagships when they move to a new node,there should be around a 70% improvement.
IIRC,the GTX780TI was around that much faster than the GTX580,and the GTX1080TI was similar compared to a GTX980TI.
Also the GTX670 was faster than the GTX580,and the GTX1070 faster than the GTX980TI.
So going by that,it would make sense the RTX3070 is faster than a RTX2080TI.
Edit!!
Also,the fact is a RTX2080TI is 42% faster than a 40CU RX5700XT at qHD according to TPU.
So if AMD doubles the CU count,etc then realistically they should easily beat an RTX2080TI.
Instead of perfect 2X scaling,lets say its 1.8X,so that should make an 80CU part at least 30% faster than an RTX2080TI,and that assumes no IPC improvements or clockspeed improvements.
If an 80CU RDNA2 part cannot beat an RTX2080TI by at least 30% something has gone wrong.
I was hoping the new Nvidia cards would be a disappointment so AMD can wreak havoc with big navi and take back market share. I'm worried for them.
TBF,if you compared the Nvidia flagships when they move to a new node,there should be around a 70% improvement.
IIRC,the GTX780TI was around that much faster than the GTX580,and the GTX1080TI was similar compared to a GTX980TI.
Also the GTX670 was faster than the GTX580,and the GTX1070 faster than the GTX980TI.
So going by that,it would make sense the RTX3070 is faster than a RTX2080TI.
Edit!!
Also,the fact is a RTX2080TI is 42% faster than a 40CU RX5700XT at qHD according to TPU.
So if AMD doubles the CU count,etc then realistically they should easily beat an RTX2080TI.
Instead of perfect 2X scaling,lets say its 1.8X,so that should make an 80CU part at least 30% faster than an RTX2080TI,and that assumes no IPC improvements or clockspeed improvements.
If an 80CU RDNA2 part cannot beat an RTX2080TI by at least 30% something has gone wrong.
Last edited: