| Eddie_Raja said: I won't doubt that the Fury X is an unbalanced card, because it clearly is. But the PS4 is equal to about an R9 370, and that card has VRAM that is about aws fast as the PS4 - and it needs it buddy. The X1 is massively bandwidth starved compared to its closest desktop counterpart (Performance wise) - the 7770. |
There is no such thing as the R9 370, it's R7 370. :) You are right on the memory bandwidth though.
R7 370 = 256-bit bus @ 5600mhz GDDR5 = 179.2GB/sec total memory bandwidth
http://www.guru3d.com/articles-pages/asus-radeon-r7-370-strix-review,1.html
PS4's GPU = 256-bit bus @ 5500mhz GDDR5 = 176 GB/sec total memory bandwidth.
PS4's GPU also has 3 key optimizations - 8 Asynchronoous Compute engines with up to 64 Command Ques (8 ACEs), while Xbox 1's has 2 ACEs. It has a 20GB/sec bus connecting the GPU directly to the system memory - that's faster than PCIe 3.0 x16. It also has volatile bits (tags for cache lines) for reducing the overhead of the graphics card writh respect to context switching between graphics and compute:
http://www.gamasutra.com/view/feature/191007/inside_the_playstation_4_with_mark_.php?print=1
Xbox One's GPU also has major deficiencies with 1/2 the ROPs of PS4's GPU and only 48 TMUs vs. 72 in PS4. The problem is most 3rd party developers don't spend extra time to take full advantage of the GPU in PS4 because in theory it has up to 50% more performance. We'd have to wait for 1st parties like Naughty Dog to be able to extract maximum performance to see what the PS4 is truly capable of. The other issue is PS4 (and XB1) do run into CPU bottlenecks because 8-core Jaguar clocked at 1.6-1.75Ghz are still very slow compared to modern Core i3/i5/i7s or FX9590 series, etc.
But even the issue with ARM SoCs is that their GPU are absolute dogs compared to what Nintendo could technically acquire from AMD in 2016. AMD's GPU performance scales with how much $ Nintendo is willing to pay for it, while ARM SOC's maximum GPU performance scales between dog slow and snail slow.
If we look at R7 265 and 270 cards, they are 36-45% faster than the GTX750.
http://www.computerbase.de/2014-02/nvidia-geforce-gtx-750-ti-maxwell-test/5/#chart-groups-47830
In turn, GTX750 trashes the fastest Intel graphics currently available for sale:
The graphics capabilities of Surface Pro 3 are not even as good as the best integrated grapics of the top Intel chips and yet GTX750 smokes them, but in turn R7 265/270 smoke the 750. But the Surface Pro 3 itself has graphics performance in 3DMark that's nearly 75% faster than Apple 6S's SOC.
There is no way any 2016 ARM graphics will have the capability of coming anywhere close to the performance of modern Xbox 1/PS4 in modern games.
This sound like some fanboy fantasy for people who hate the idea of an AMD-powered Nintendo NX.
