@ronvalencia: This guy just sticks his fingers in his ears and screams when you try and say some empirically factual stuff, not sure why you bother any more at this point lol, Fury X has 4GB of VRAM for all itself... PS Pro has a 5.5GB shared pool for CPU/GPU, that leaves about 2.5GB/3.0GB for game each for CPU/GPU to use, so simple
BF1 on ultra at 1440p uses about 4.5 GB-5GB VRAM and 6-8GB of system RAM for the game on my system
Rainbow Six Siege ultra texture pack uses 5-6GB of VRAM on my RX 480
This is more available RAM then PS4/Pro total for it's CPU/GPU for games
PS4 games has zero copy advantage i.e. don't apply Windows PC's memory model on PS4.
CPU side AI logic and physics are not large.
http://www.dualshockers.com/2014/04/02/how-infamous-second-son-used-the-ps4s-8-4-5-gb-of-ram-cpu-and-gpu-compute-to-make-our-jaws-drop/
It's mostly VRAM related.
http://www.eurogamer.net/articles/digitalfoundry-ps3-system-software-memory
PS4 has 5 GB for game.
For Windows 10 DX12 and Vulkan, notice the following
1. Data uploaded into system memory pool. On PS4, there's only one memory pool.
2. GPU reads data from system memory pool e,g, to populate GPU's VRAM. This step is not necessary for PS4 ie. there's only one memory pool.
3. CPU's lacking direct access to GPU's memory pool indicates non-HSA i.e. no memory address GPU and CPU merging.
Blame non-HSA GPUs.
http://wccftech.com/intel-amd-nvidia-future-industry-hsa/2/
Nvidia introduced the capability to share virtual memory between the CPU and GPU with CUDA 6 and its Maxwell graphics architecture. The technology doesn’t offer a hardware-level unified memory access like AMD’s hUMA, which was introduced with Kaveri APUs and the GCN 1.1 graphics architecture. What it does is simplify the method by which programmers address CPU and GPU memory. What it does not do, however, is allow the processors to share data via pointers. Which is an essential part in addressing the most significant bottleneck in a heterogeneous design. Without reducing the huge latency dictated by data copies no tangible performance or efficiency gains can be realized from a a heterogeneous processor. So despite the memory pool being “virtually” unified in software, performance exhaustive data copies still have to be made between the CPU and GPU.
Windows 10 DX12 and Vulkan was designed for lowest common dominator i.e. non-unified address space hardware capable GPUs.
Radeon HD 7970's unified address space hardware.
AMD GCN's memory handling exceeds DX12 and Vulkan.
Log in to comment