Sign in to follow this  
Followers 0


1 post in this topic

ALU is the same... seems odd, but, promises efficency per watt in only promise, compare 28nm to 16FF and you get huge efficient.

Add HMB reduces memory power significantly makes the rest of it.

Hopefully untrue?

Clarification: ALUs on Maxwell v. on Pascal are near the same.  Pascal's addition perf per watt comes from node jumps and a decrease in power thanks to use HBM instead of GDDR5.

The speed increase if any will mostly come from adding more ALU + more bandwidth.

This is something I have heard from someone who rarely says things unless he thinks they can be supported, and, he has not 100% confirmed this.

That being said, Keeping Maxwell ALUs seems like a very good thing, considering the performance and the low power they use, shrinking it should make it better.  As Nvidia has to deal with finFETs and also with HBM costs/R&D this should allow Pascal chips to come faster while bringing performance increase without breaking NVidia's budget or the consumers budget.



Share this post

Link to post
Share on other sites

Please sign in to comment

You will be able to leave a comment after signing in

Sign In Now
Sign in to follow this  
Followers 0