By using this site, you agree to our Privacy Policy and our Terms of Use. Close
JEMC said:
se7en7thre3 said:

this was 10k's no filter updated quote from the dev, which he is still leaning on despite being a lone source:

"So far I know this: Polaris architecture as a base, mixed with Fiji, low power consumption, raw power numbers are modest increase, performance for optimized games for NX would be easily 2x PS4 performance."

If you look at r9 nano (Fiji XT), it has a core config of 4096:256:64 = 8.192 TF @ 175w.

For the tech heads here, is it possible Nintendo could get a custom gpu based of that card, lets say Nintendo wants 1/4 of that so 1024:64:16, how much could that possibly drop the tdp, on either 14 or 28nm?

I'm not exactly a tech head, but I'll answer you.

AMD doesn't have a GPU with that configuration, 1024:64:16, but it has something very similar in the GPU inside the R7 370 with a config of 1024:64:32. That sounds better, right? Well, no it doesn't.

It turns out that that GPU is worse than the one inside the PS4, that has a config of 1152:72:32.

To have 2x specs of the PS4, the NX would need something like the R9 380X which is 2048:128:32, but the reality is that that only works on paper. Here is a comparison between the R7 370 and the R9 380X to show you what I mean: http://www.anandtech.com/bench/product/1590?vs=1592

To have twice the performance of a PS4 you would need something like the R9 390X, and that's too expensive and uses too much power for a console.

I definitely don't think its 2x PS4, that talk got out of hand when 14nm got mentioned.  I bring up Fiji as that's the updated quote.  For performance, the r9 nano's 8.192 tf @ 175w seems to be the best at 28nm.  I was wondering if Nintendo could get a cut down/customized version, and if the TDP would drop accordingly (by half or more?).  i.e. how much can TDP be reduced from 175w, with a lower core config that compares more/less or between XB1 - PS4.

edit @JEMC

I do look at that comparison chart, and compared Fiji Pro to Fiji XT.