By using this site, you agree to our Privacy Policy and our Terms of Use. Close
Ashadian on 07 October 2013
Edit Thread
Adinnieken said:
iamdeath said:

yOU DON'T KNOW WHAT YOU ARE TALKING ABOUT. wHAT CERNY SAYS ACTUALLY COUNTERACTS WHAT YOU ARE SAYING lmao.

 

You have been trying to spin hardware for months in the x1's favour and everything you have ever posted has been factually wrong. Just like your scret sauce, extra GPU threads.

 

That ahs been disproven, numerous times/.

Actually, he is agreeing with me.  When he says "It has a little be more ALU than it would if you were strictly thinking about graphics... ." (emphasis added)  It implies the GPU would be used for GPGPU purposes.  Not saying that the additional four CUs can't be used for rendering, simply that at 14 CUs you're saturating the memory bandwidth.  You only have 176GB/s that you can pipe data through.  That's peak efficiency, something not realized in the realworld.  At 80% efficiency you get 140GB/s (80% of the PS4's 176GB/s) through that pipe.  Since the PS4 has 50% more CUs than the Xbox One's GPU, and at 12 CUs the Xbox One is generating 146GB/s, then it stands to reason that at 50% more power the PS4 would saturate the 176GB/s if all 18 CUs are running (219GB/s).   

Again, doesn't mean you can't, just you're not doing yourself any favors if you do.   You're going to run into a bottleneck.

If it has, as you've said, been disproven please provide a link.

If you have some magical way where you can fit 219GB into a 140GB hole, than please tell us how you do it.  I'm sure there are a lot of people would love to know how it can be done.


Are you for real? The Radeon 7870 GPU has 20 CU's and only a 154Gb peak bus bandwidth yet it's rated at roughly 2.5 Teraflops. So going by your maths how fast would the bus bandwidth be at 80% effeminacy? So what magic are AMD using? Maybe they have a DGPU secretly hidden or even stacked? But you have to wait till the 29th of Nov for the NDA to lift.

As far as I'm concerned your just trolling as anyone with any common sense knows that GPU CU's are the same. They can all be used for graphics or physics or a combination of both. Also you convieniantly forgot to add the quote where Cerny states that when certain parts of the GPU are idle or not running at 100% they can be utilised for GPGPU work!