How many SM's in full fat TU-102 and TU-104 ?

[Meaker]

So where are they hiding in the die shot?

 

[Dirk Diggler]

I'm confused by all the negative speculation around the 2080Ti regarding its relative performance compared to the 1080Ti...

...I watched the Nvidia launch and heard Jensen Huang saying there is a new architecture for these chips and the SMs will have significantly improved performance over Pascal.

Maybe you would be less confused if he had released actual benchmark comparison figures, as opposed to all the cryptic Gigaray patter.

 

[4K8KW10]

754 sq. mm chip is so large, it is crazy. Why do nvidia even feel the need to release it, when their old Pascal midrange chip was just 200 sq. mm?!
AMD will completely demolish them with a much smaller chip.

 

[orbitalwalsh]

754 sq. mm chip is so large, it is crazy. Why do nvidia even feel the need to release it, when their old Pascal midrange chip was just 200 sq. mm?!
AMD will completely demolish them with a much smaller chip.

Smaller nodes doesn't always mean smaller chips . Means you can cram more into a set area but when you need to cram a lot in.. it's going to get big

If they cut the RT with 2060/50, should be a nice smaller die

 

[Kaapstad]

So where are they hiding in the die shot?

You are correct, it looks like 4608 is all you get on TU 102 chip.

Having said that going by NVidias naming system there will be a TU 100 chip lurking around somewhere waiting to be launched and it will almost definitely have SP 5120 cores or even more.

https://www.techpowerup.com/247039/nvidia-tu102-rt-core-and-tensor-core-counts-revealed

 

[Mauller]

You are correct, it looks like 4608 is all you get on TU 102 chip.

Having said that going by NVidias naming system there will be a TU 100 chip lurking around somewhere waiting to be launched and it will almost definitely have SP 5120 cores or even more.

https://www.techpowerup.com/247039/nvidia-tu102-rt-core-and-tensor-core-counts-revealed

Yeah, like I mentioned there likely won't be a TU100 but a TU200 on 7nm with a respin of the rest of the series on 7nm as well. I don't see the 12/16nm Turing series being long lived. With 7nm they will also likely be able to increase clocks by 25% giving a performance uplift with the refresh.