By using this site, you agree to our Privacy Policy and our Terms of Use. Close
BlueFalcon on 26 May 2016
Edit Thread
Intrinsic said:
Soundwave said:

Could be a custom order scaled up Polaris 11.

Polaris 11 churns out 2.5 TFLOPS at under 50 watts with 16 CUs.

Bump that up to 40 CUs ... 6.25 TFLOPS at about 120-130 watts.

Or just a vanilla Polaris 10 is already 5.5 TFLOPS at sub-150 watts for $299.99. 

To help put things in perspective for you. When the PS4/XB1 launched, their APUs (CPU+GPU) costs under $100usd. 

Thats how the console business works. NO ONE is going to put a $250/$300/$350 GPU alone in a console. 

You are quoting retail prices. Add-in board partners and retailers add mark-ups, logistics costs, etc. Buying the chips, GDDR5 directly shouldn't cost more than $80-100 for the small Polaris 10 chip. Besides, AMD will offer a good discount for volumes and to have DX12/GCN Eco-system in place to benefit their PC business via GCN-optimized console ports to PC. Performing well in those titles helps to sell AMD's PC graphics cards. The Division, Need for Speed, Far Cry Primal, Hitman, etc. all run very well on AMD cards because they were optimized for consoles' GCN architectures. For that reason it's in AMD's best interest to get Polaris into both the Neo and XB2. It also costs too much $$$ ($100-120 million) to do a new shrink of existing 28nm parts to 14nm. Might as well use 14nm GPUs already designed for the new node. Since GCN 4.0 (Polaris 10) is backwards compatible with all previous GCN parts, using Polaris 10 ensures both backwards and forward compatibility for XB3/PS5. 

The the biggest issue here is not hardware but software related. Sony already said they will not allow exclusive Neo title as, more or less making the extra hardware only viable to increase frame rates or some graphical effects on what are going to be base PS4 games. Pretty meh concept. For XB2, MS would need to allow developers to make exclusives or give a lot more free room to make much better looking games or otherwise the major gap between PS4/XB1 and XB2, combined with 80-100 million of those "old" consoles userbase will chain developers to optimize for the old, not new consoles. For the strategy to work, MS would need to drop XB1 entirely. Othwrwise, there will be a massive 2-3 year lag between 2017's XB2 and when developers start using that hardware. This is why many are sceptical of the new hardware release strategy.